Transmission line galloping risk early warning system based on multiphysics field fusion

Abstract

This invention discloses a transmission line galloping risk early warning and system based on multiphysics field fusion. The method includes: first, acquiring a multiphysics field dataset of the transmission line galloping process, generating a dynamic three-dimensional point cloud sequence through spatiotemporal alignment and three-dimensional attitude reconstruction; then, extracting galloping energy distribution features, frequency spectrum features, and spatial trajectory features from the dataset and fusing them to obtain a fused feature vector; subsequently, inputting this vector into a physically constrained enhanced bidirectional long short-term memory network to predict the galloping evolution trend and outputting a joint prediction matrix of galloping amplitude and frequency; next, inputting this matrix into a pre-trained risk propagation inference model to calculate multi-level risk quantification indicators; finally, generating a risk early warning prompt based on the indicators and sending it to the corresponding risk management equipment to achieve transmission line galloping risk early warning. This invention can accurately predict the evolution trend of transmission line galloping, thereby accurately warning of transmission line galloping risks.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a transmission line galloping risk early warning and system based on multi-physics field fusion. Background Art

[0002] During the operation of transmission lines, conductor galloping, a self-excited vibration phenomenon induced by wind, poses a significant threat to the safe and stable operation of the power grid due to its low-frequency and high-amplitude characteristics. Current technologies typically employ single-physical-field monitoring methods or simple multi-parameter combined analysis approaches for risk assessment. Specifically, traditional technologies rely primarily on a limited number of sensors installed on the transmission lines (such as accelerometers, anemometers, and thermometers) to collect local data, combined with meteorological parameters provided by weather stations, to analyze the galloping state of the lines through threshold judgments or simple statistical models. Some methods use video monitoring to capture images of line vibration and extract galloping features through image processing techniques, or utilize traditional machine learning algorithms (such as support vector machines and decision trees) to model and analyze historical monitoring data to predict galloping risks. These methods often only consider information from a single time or spatial dimension in data processing, lacking in-depth fusion analysis of multi-physical-field data, and the prediction models are mostly data-driven, failing to fully consider the physical mechanisms constraining transmission line galloping.

[0003] Therefore, the existing technical solutions have the following problems: the lack of in-depth analysis of multi-physics data leads to inaccurate prediction of the evolution trend of transmission line galloping, thus making it impossible to accurately warn of transmission line galloping risks. Summary of the Invention

[0004] This invention provides a transmission line galloping risk early warning and system based on multi-physics field fusion, which can accurately predict the evolution trend of transmission line galloping, thereby accurately warning of transmission line galloping risks.

[0005] An embodiment of the present invention provides a method for early warning of transmission line galloping risk based on multi-physics field fusion, comprising:

[0006] Obtain a multiphysics dataset of the galloping process of power transmission lines;

[0007] Spatiotemporal alignment and 3D pose reconstruction are performed on the multiphysics dataset to generate a dynamic 3D point cloud sequence of the power transmission line galloping.

[0008] The three features of dancing energy distribution, frequency spectrum and spatial trajectory are extracted from the dynamic three-dimensional point cloud sequence, and the three features are fused to obtain a fused feature vector.

[0009] The fused feature vector is input into a physically constrained enhanced bidirectional long short-term memory network to predict the galloping evolution trend, and the joint prediction matrix of the galloping amplitude and frequency of the transmission line in the future time is output.

[0010] The fused feature vector and the joint prediction matrix of the dancing amplitude and frequency are input into the pre-trained risk propagation inference model to calculate the multi-level risk quantification index of the transmission line. The multi-level risk quantification index includes the probability of conductor dynamic tension exceeding the limit and the probability of insulation gap breakdown.

[0011] Based on the multi-level risk quantification indicators, corresponding risk warning prompts are generated and sent to the corresponding risk management devices.

[0012] As an improvement to the above scheme, the step of performing spatiotemporal alignment and 3D pose reconstruction on the multiphysics dataset to generate a dynamic 3D point cloud sequence of the transmission line galloping includes the following sub-steps:

[0013] Phase synchronization calibration based on the dancing resonant frequency is performed on the motion acceleration data and video stream data in the multiphysics dataset to obtain a time-aligned dataset; the multiphysics dataset includes motion acceleration data, ambient wind speed data, and video stream data;

[0014] Based on the time-aligned dataset, the rigid body dynamics constraint model of the conductor is applied to perform spatial coordinate compensation on the environmental wind speed data to obtain spatially corrected data.

[0015] The spatial correction data is input into a three-dimensional reconstruction algorithm that integrates conductor bending stiffness parameters to generate a dynamic three-dimensional point cloud sequence of the transmission line galloping.

[0016] As an improvement to the above scheme, the step of extracting the dancing energy distribution features, frequency spectrum features, and spatial trajectory features from the dynamic three-dimensional point cloud sequence, and fusing the three features to obtain a fused feature vector, includes the following sub-steps:

[0017] Based on the coupling equation of the aerodynamic structure of the conductor, the dynamic three-dimensional point cloud sequence is subjected to vortex-induced vibration mode separation to obtain the galloping energy distribution characteristics.

[0018] The frequency spectrum features are obtained by applying a dancing dominant frequency adaptive filter to the dynamic three-dimensional point cloud sequence.

[0019] The trajectory curvature of the dynamic three-dimensional point cloud sequence is corrected based on the catenary equation of the conductor to obtain spatial trajectory features.

[0020] The energy distribution characteristics, frequency spectrum characteristics, and spatial trajectory characteristics of the dancing motion are fused to obtain the fused feature vector.

[0021] As an improvement to the above scheme, the step of inputting the fused feature vector into a physically constrained enhanced bidirectional long short-term memory network for galloping evolution trend prediction, and outputting a joint prediction matrix of the galloping amplitude and frequency of the transmission line, includes the following sub-steps:

[0022] The fused feature vectors are reorganized in a temporal dimension to generate a feature tensor sequence with a fixed time window length.

[0023] The feature tensor sequence is synchronously input into the forward and reverse computation layers of the bidirectional long short-term memory network, and the forward dancing evolution hidden state sequence and the reverse dancing evolution hidden state sequence are output respectively.

[0024] A physical rule base is constructed based on the conductor suspension point displacement constraint equation and aerodynamic torque balance condition. The forward gobling evolution hidden state sequence and the reverse gobling evolution hidden state sequence are dynamically weighted and fused to generate a gobling evolution feature matrix with physical constraint correction.

[0025] The physical constraint-corrected dancing evolution feature matrix is ​​mapped to the joint prediction matrix of dancing amplitude and frequency.

[0026] As an improvement to the above scheme, the joint prediction matrix of the galloping amplitude and frequency is input into a pre-trained risk propagation inference model to calculate the multi-level risk quantification index of the transmission line. The multi-level risk quantification index includes the probability of conductor dynamic tension exceeding the limit and the probability of insulation gap breakdown. This includes the following sub-steps:

[0027] The amplitude-frequency parameter is decoupled from the joint prediction matrix of the dancing amplitude and frequency to separate the dancing amplitude component sequence and the dancing frequency component sequence;

[0028] The galloping amplitude component sequence is input into a pre-trained risk propagation inference model to calculate the dynamic tension of the conductor, and combined with the static initial tension value of the transmission line, the dynamic tension time history data of the transmission line is calculated.

[0029] Based on the calculated dynamic tension time history data and the preset conductor tension safety threshold, the percentage of tension exceeding the limit duration is statistically analyzed to obtain the probability of the conductor dynamic tension exceeding the limit.

[0030] The amplitude and frequency component sequences of the galloping motion are input into a pre-trained risk propagation inference model to perform electric field analysis on the insulation gap. Combined with the inherent insulation configuration parameters of the transmission line, the minimum air gap distance variation and local electric field intensity distribution are calculated. The probability of insulation gap breakdown risk is obtained based on the calculation results.

[0031] As an improvement to the above solution, the step of generating corresponding risk warning prompts based on the multi-level risk quantification indicators and sending them to the corresponding risk management devices includes the following sub-steps:

[0032] Based on the probability of conductor dynamic tension exceeding the limit and the probability of insulation gap breakdown risk in the multi-level risk quantification indicators, the risk type, risk location and risk level are determined.

[0033] A structured risk warning is generated based on the risk level, and the risk warning includes risk type, risk location, and risk level information.

[0034] The risk warning will be dynamically sent to the corresponding risk management device based on the risk location.

[0035] Another embodiment of the present invention provides a transmission line galloping risk early warning system based on multiphysics field fusion, comprising:

[0036] The acquisition module is used to acquire multiphysics datasets of the galloping process of transmission lines;

[0037] The generation module is used to perform spatiotemporal alignment and three-dimensional attitude reconstruction on the multiphysics dataset to generate a dynamic three-dimensional point cloud sequence of the power transmission line galloping.

[0038] The fusion module is used to extract the dancing energy distribution features, frequency spectrum features and spatial trajectory features from the dynamic three-dimensional point cloud sequence, and to fuse the three features to obtain a fused feature vector.

[0039] The prediction module is used to input the fused feature vector into a physically constrained enhanced bidirectional long short-term memory network to predict the galloping evolution trend and output a joint prediction matrix of the galloping amplitude and frequency of the transmission line.

[0040] The calculation module is used to input the joint prediction matrix of the galloping amplitude and frequency into the pre-trained risk propagation inference model to calculate the multi-level risk quantification index of the transmission line.

[0041] The early warning module is used to generate corresponding risk warning prompts based on the multi-level risk quantification indicators and send them to the corresponding risk management devices.

[0042] As an improvement to the above solution, the generation module is specifically used for:

[0043] Phase synchronization calibration based on the dancing resonant frequency is performed on the motion acceleration data and video stream data in the multiphysics dataset to obtain a time-aligned dataset; the multiphysics dataset includes motion acceleration data, ambient wind speed data, and video stream data;

[0044] Based on the time-aligned dataset, the rigid body dynamics constraint model of the conductor is applied to perform spatial coordinate compensation on the environmental wind speed data to obtain spatially corrected data.

[0045] The spatial correction data is input into a three-dimensional reconstruction algorithm that integrates conductor bending stiffness parameters to generate a dynamic three-dimensional point cloud sequence of the transmission line galloping.

[0046] As an improvement to the above solution, the fusion module is specifically used for:

[0047] Based on the coupling equation of the aerodynamic structure of the conductor, the dynamic three-dimensional point cloud sequence is subjected to vortex-induced vibration mode separation to obtain the galloping energy distribution characteristics.

[0048] The frequency spectrum features are obtained by applying a dancing dominant frequency adaptive filter to the dynamic three-dimensional point cloud sequence.

[0049] The trajectory curvature of the dynamic three-dimensional point cloud sequence is corrected based on the catenary equation of the conductor to obtain spatial trajectory features.

[0050] The energy distribution characteristics, frequency spectrum characteristics, and spatial trajectory characteristics of the dancing motion are fused to obtain the fused feature vector.

[0051] As an improvement to the above scheme, the prediction module is specifically used for:

[0052] The fused feature vectors are reorganized in a temporal dimension to generate a feature tensor sequence with a fixed time window length.

[0053] The feature tensor sequence is synchronously input into the forward and reverse computation layers of the bidirectional long short-term memory network, and the forward dancing evolution hidden state sequence and the reverse dancing evolution hidden state sequence are output respectively.

[0054] A physical rule base is constructed based on the conductor suspension point displacement constraint equation and aerodynamic torque balance condition. The forward gobling evolution hidden state sequence and the reverse gobling evolution hidden state sequence are dynamically weighted and fused to generate a gobling evolution feature matrix with physical constraint correction.

[0055] The physical constraint-corrected dancing evolution feature matrix is ​​mapped to the joint prediction matrix of dancing amplitude and frequency.

[0056] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0057] By acquiring a multiphysics dataset of the transmission line galloping process, multidimensional physical information related to galloping is collected. The multiphysics dataset is spatiotemporally aligned and three-dimensional attitude reconstructed to generate a dynamic three-dimensional point cloud sequence, eliminating spatiotemporal biases. Galloping energy distribution features, frequency spectrum features, and spatial trajectory features are extracted from the dynamic three-dimensional point cloud sequence and fused to obtain a fused feature vector. This multidimensional feature vector comprehensively characterizes the energy, frequency, and spatial motion characteristics of the galloping, avoiding the limitations of single features. The fused feature vector is input into a physically constrained bidirectional long short-term memory network (LSTM). The LSTM captures forward and backward correlation information of the time-series data, and the enhanced physical constraints ensure that the prediction conforms to actual physical laws, thereby better predicting the galloping evolution trend and outputting a joint prediction matrix of galloping amplitude and frequency. The joint prediction matrix of galloping amplitude and frequency is input into a pre-trained risk propagation inference model, transforming the predicted galloping features into graded and quantified risk indicators, achieving multi-level risk quantification. Finally, a risk warning is generated based on the multi-level risk quantification indicators and sent to the corresponding risk management equipment to complete the warning. In summary, the embodiments of the present invention can accurately predict the galloping evolution trend of transmission lines by acquiring multi-physics field data, integrating the data and extracting and fusing multi-dimensional features, using a physically constrained enhanced bidirectional long short-term memory network to accurately predict the galloping evolution trend, and quantifying multi-level risks through a risk propagation inference model, thereby accurately predicting the galloping risk of transmission lines. Attached Figure Description

[0058] Figure 1 This is a flowchart illustrating a method for early warning of transmission line galloping risk based on multi-physics field fusion, provided in an embodiment of the present invention.

[0059] Figure 2 This is a schematic diagram of the structure of a transmission line galloping risk early warning system based on multi-physics field fusion, provided by an embodiment of the present invention. Detailed Implementation

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0061] See Figure 1 This is a flowchart illustrating a method for early warning of transmission line galloping risks based on multi-physics field fusion, according to an embodiment of the present invention. The method includes:

[0062] S10, Obtain the multiphysics dataset of the galloping process of the transmission line;

[0063] S11, perform spatiotemporal alignment and three-dimensional attitude reconstruction on the multiphysics dataset to generate a dynamic three-dimensional point cloud sequence of the power transmission line galloping.

[0064] S12, extract the dancing energy distribution features, frequency spectrum features and spatial trajectory features from the dynamic three-dimensional point cloud sequence, and perform feature fusion on the three features to obtain a fused feature vector;

[0065] S13, input the fused feature vector into the physically constrained enhanced bidirectional long short-term memory network to predict the galloping evolution trend, and output the joint prediction matrix of the galloping amplitude and frequency of the transmission line in the future time.

[0066] S14, input the fused feature vector and the joint prediction matrix of the dancing amplitude and frequency into the pre-trained risk propagation inference model to calculate the multi-level risk quantification index of the transmission line. The multi-level risk quantification index includes the probability of conductor dynamic tension exceeding the limit and the probability of insulation gap breakdown.

[0067] S15, generate corresponding risk warning prompts based on the multi-level risk quantification indicators, and send them to the corresponding risk management device.

[0068] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0069] By acquiring a multiphysics dataset of the transmission line galloping process, multidimensional physical information related to galloping is collected. The multiphysics dataset is spatiotemporally aligned and three-dimensional attitude reconstructed to generate a dynamic three-dimensional point cloud sequence, eliminating spatiotemporal biases. Galloping energy distribution features, frequency spectrum features, and spatial trajectory features are extracted from the dynamic three-dimensional point cloud sequence and fused to obtain a fused feature vector. This multidimensional feature vector comprehensively characterizes the energy, frequency, and spatial motion characteristics of the galloping, avoiding the limitations of single features. The fused feature vector is input into a physically constrained bidirectional long short-term memory network (LSTM). The LSTM captures forward and backward correlation information of the time-series data, and the enhanced physical constraints ensure that the prediction conforms to actual physical laws, thereby better predicting the galloping evolution trend and outputting a joint prediction matrix of galloping amplitude and frequency. The joint prediction matrix of galloping amplitude and frequency is input into a pre-trained risk propagation inference model, transforming the predicted galloping features into graded and quantified risk indicators, achieving multi-level risk quantification. Finally, a risk warning is generated based on the multi-level risk quantification indicators and sent to the corresponding risk management equipment to complete the warning. In summary, the embodiments of the present invention can accurately predict the galloping evolution trend of transmission lines by acquiring multi-physics field data, integrating the data and extracting and fusing multi-dimensional features, using a physically constrained enhanced bidirectional long short-term memory network to accurately predict the galloping evolution trend, and quantifying multi-level risks through a risk propagation inference model, thereby accurately predicting the galloping risk of transmission lines.

[0070] As one example of the above scheme, the step of performing spatiotemporal alignment and 3D pose reconstruction on the multiphysics dataset to generate a dynamic 3D point cloud sequence of the transmission line galloping includes the following sub-steps:

[0071] Phase synchronization calibration based on the dancing resonant frequency is performed on the motion acceleration data and video stream data in the multiphysics dataset to obtain a time-aligned dataset; the multiphysics dataset includes motion acceleration data, ambient wind speed data, and video stream data;

[0072] Based on the time-aligned dataset, the rigid body dynamics constraint model of the conductor is applied to perform spatial coordinate compensation on the environmental wind speed data to obtain spatially corrected data.

[0073] The spatial correction data is input into a three-dimensional reconstruction algorithm that integrates conductor bending stiffness parameters to generate a dynamic three-dimensional point cloud sequence of the transmission line galloping.

[0074] In this embodiment, firstly, phase synchronization calibration is performed on the motion acceleration data and video stream data in the multi-physics dataset based on the inherent resonant frequency characteristics of transmission line galloping, achieving precise alignment of multi-source heterogeneous data in the time dimension. Then, a rigid body dynamics constraint model of the conductor is used to compensate for the spatial coordinates of the environmental wind speed data, correcting the spatial position deviation of the conductor under wind load and eliminating environmental interference. Finally, the spatiotemporally aligned and spatially corrected data is input into a 3D reconstruction algorithm that fuses the conductor bending stiffness parameters. By introducing constraints based on the mechanical properties of the conductor material, a dynamic 3D point cloud sequence that accurately reflects the deformation characteristics during transmission line galloping is generated. Therefore, this embodiment solves the problem of inconsistent time references among multi-source data through resonant frequency phase synchronization, eliminates spatial positioning errors caused by environmental wind loads through rigid body dynamics constraint compensation, and ensures that the point cloud sequence conforms to the physical deformation laws of the conductor through bending stiffness fusion reconstruction, thereby constructing a high-fidelity dynamic 3D model of galloping.

[0075] Specifically, the working process of this embodiment is illustrated below:

[0076] First, the multiphysics dataset was acquired as follows: motion acceleration data was collected by triaxial accelerometers deployed at the suspension points and midpoints of the conductor, with a sampling frequency of 200Hz, recording instantaneous acceleration in the x (along the conductor), y (perpendicular to the conductor), and z (vertical) directions; video stream data was collected by cameras near the conductor; and environmental wind speed data was collected by a 3D ultrasonic anemometer next to the cameras, recording instantaneous wind speed, direction, and turbulence intensity. Due to differences in hardware response delay and sampling triggering mechanisms between the accelerometers and cameras, there is a time phase deviation between the two types of data, requiring phase synchronization calibration based on the galloping resonant frequency. Specifically, a Fast Fourier Transform (FFT) was first performed on the motion acceleration data to extract the galloping principal resonant frequency. (Typically 0.1-5Hz, determined by conductor type, span, and wind speed); then with Using this as a baseline, calculate the initial phase difference between the acceleration data and the video stream data. (Assume the acceleration data phase) Video stream data phase ,but );according to The timestamp correction for the video stream is calculated using the following formula: , The original timestamp of the video stream. This is the corrected timestamp. The corrected video stream data, original motion acceleration data, and environmental wind speed data are integrated according to a unified timestamp to obtain a time-aligned dataset.

[0077] Based on a time-aligned dataset, key feature points such as wire suspension points and span midpoints are first extracted from the video stream data using the YOLOv8 algorithm. These features are then combined with camera intrinsic parameters (focal length). Principal point coordinates ( Using external parameters, calculate the real-time 3D coordinates of each feature point in the world coordinate system. Then, by integrating the motion acceleration data twice, the instantaneous velocity and displacement of the midpoint of the conductor are obtained. Combined with the rigid body motion characteristics of the conductor (no significant deformation during galloping, only rigid body motion), a constraint model is constructed to inversely deduce the real-time spatial attitude parameters of the conductor (midpoint displacement). Deflection angle Pitch angle Based on these attitude parameters, the environmental wind speed data (three components in the world coordinate system) is... Convert to components in the local coordinate system of the conductor (along the conductor axis) , vertical conductor radial Vertical direction This compensates for wind speed deviations caused by the spatial distance between the anemometer and the conductor, as well as the conductor's orientation, to form spatial correction data.

[0078] Extracting consecutive frames from the video stream from the spatially corrected data, and using the SIFT feature matching algorithm to initially generate the original 3D point cloud of the conductor. ( Original coordinates Simultaneously, obtain the physical parameters of the conductor (such as the elastic modulus of the conductor). Moment of inertia of cross section Calculate bending stiffness ), combined with motion acceleration data Directional acceleration The static theoretical shape of the conductor is obtained through quadratic integration and the catenary model, serving as a deformation correction benchmark. A point cloud correction model incorporating bending stiffness is constructed, based on beam bending theory and considering the aerodynamic moment acting on the conductor. The original point cloud is corrected using the following formula: , , For the corrected point cloud coordinates, for Orientation deformation correction amount for Directional unit vector, The coordinates of the point cloud along the conductor axis are given. After correcting each frame of the original point cloud, they are arranged in chronological order to form a dynamic 3D point cloud sequence with time intervals consistent with the video stream frame rate, thus fully recording the spatial morphology and bending deformation of the conductor's movement.

[0079] The method involves considering three features: frequency spectrum features, spatial trajectory features, and spatial characteristics, and fusing these three features to obtain a fused feature vector. This includes the following sub-steps:

[0080] Based on the coupling equation of the aerodynamic structure of the conductor, the dynamic three-dimensional point cloud sequence is subjected to vortex-induced vibration mode separation to obtain the galloping energy distribution characteristics.

[0081] The frequency spectrum features are obtained by applying a dancing dominant frequency adaptive filter to the dynamic three-dimensional point cloud sequence.

[0082] The trajectory curvature of the dynamic three-dimensional point cloud sequence is corrected based on the catenary equation of the conductor to obtain spatial trajectory features.

[0083] The energy distribution characteristics, frequency spectrum characteristics, and spatial trajectory characteristics of the dancing motion are fused to obtain the fused feature vector.

[0084] In this embodiment, vortex-induced vibration modes are first separated from the dynamic three-dimensional point cloud sequence based on the aerodynamic coupling equation of the conductor. By analyzing the vibration mode characteristics of the conductor under wind load, energy features reflecting the galloping energy level distribution are extracted. Then, an adaptive filtering algorithm based on the dominant frequency of galloping is applied to the point cloud sequence to dynamically identify and extract key frequency components during the galloping process, forming a frequency spectrum feature characterizing the periodicity of vibration. Simultaneously, the curvature of the point cloud trajectory is corrected using the catenary equation of the conductor to eliminate static deformation interference caused by gravity and restore the true spatial trajectory characteristics of the galloping. Finally, the three types of physical constraint features are fused in multiple dimensions to construct a fused feature vector containing energy, frequency, and spatial elements. In summary, this embodiment improves the characterization accuracy and physical consistency of the galloping energy distribution, frequency characteristics, and spatial trajectory features through a physical constraint feature extraction method involving vortex-induced vibration mode separation, adaptive filtering of the dominant frequency, and catenary trajectory correction.

[0085] As an example, the working process of this embodiment is as follows:

[0086] From the dynamic 3D point cloud sequence, first extract discrete points along the span of the conductor at every 0.5m in the main direction of the troika ( Based on displacement time history data (in the direction of displacement), combined with the conductor's mass per unit length (calculated from conductor density and cross-sectional area), damping coefficient, bending stiffness, and aerodynamic parameters, an improved aerodynamic-structural coupling equation for the conductor is constructed. This equation optimizes the traditional linear model's characterization of the coupling relationship between aerodynamic loads and structural vibrations by introducing a nonlinear lift coefficient term that dynamically varies with displacement and velocity. The equation expression is as follows: , Mass per unit length of conductor The damping coefficient is... For bending stiffness, air density, For wind speed, The diameter of the wire. The dynamic lift coefficient. For the first The displacement time histories of discrete points are used. The Galerkin method is used to perform modal decomposition on the equation to obtain the natural frequencies and mode shapes of the first and second order vortex-induced vibration modes. Then, the energy (including kinetic energy and elastic potential energy) of each mode at different discrete points is calculated by combining the displacement time histories data. The energy values ​​are mapped to the entire length of the conductor according to the discrete point positions to form the galloping energy distribution characteristics, which can clearly distinguish the energy proportion of different modes at different positions of the conductor.

[0087] Displacement time-history data of the midpoint of the conductor in a dynamic 3D point cloud sequence (with the largest galloping amplitude and most significant frequency characteristics) were selected. First, a continuous wavelet transform was used to obtain the time-frequency energy spectrum, identifying frequency ranges with an energy proportion exceeding 60% as candidate dominant frequency ranges. An adaptive filter was constructed based on this range, with the filter passband frequency dynamically adjusted according to the energy spectrum: when the energy of a certain frequency accounts for ≥10% of the total energy, it is included in the passband; otherwise, it is considered noise and filtered out. Power spectral density analysis was performed on the filtered displacement time-history data to extract the peak frequency (dominant frequency), 3dB bandwidth, and peak power. Simultaneously, power spectral data were collected at the conductor suspension point and the 1 / 4 span point, calculating the dominant frequency deviation at each location, and integrating these data to form frequency spectral features.

[0088] Extract the spatial point set of the traverse at a certain moment from the dynamic 3D point cloud sequence, fit the original trajectory curve using the least squares method, and then calculate the static theoretical trajectory based on the traverse catenary equation (no need to list the equation, directly use the existing catenary model parameters for calculation). Compare the original trajectory with the theoretical trajectory to obtain the deviation at each point, and correct points where the deviation exceeds 1 / 5 of the traverse diameter. The correction formula is as follows: , , , The corrected coordinates, , Original coordinates This is the deviation amount. , The coordinates are the theoretical trajectory coordinates, and 0.9 is a correction coefficient. Based on the corrected trajectory, the curvature and torsion are calculated, and the coordinates of the highest and lowest points of the trajectory, as well as the amplitude of the swing, are extracted to form the spatial trajectory features.

[0089] The three types of features are standardized: the dancing energy distribution features (two-dimensional matrix) are reduced to vectors using PCA. The frequency spectrum features (containing 6 parameters) are organized into a vector. The spatial trajectory features (including 5 parameters) are organized into a vector. The random forest algorithm is used to determine the feature weights. , , Through the weighted fusion formula: ,right , Perform zero-fill to After calculating the dimension, the final fused feature vector is obtained. This vector encompasses core information such as energy, frequency, and space, meeting the input requirements of subsequent prediction models.

[0090] As one example of the above scheme, the step of inputting the fused feature vector into a physically constrained enhanced bidirectional long short-term memory network to predict the galloping evolution trend and outputting a joint prediction matrix of the galloping amplitude and frequency of the transmission line includes the following sub-steps:

[0091] The fused feature vectors are reorganized in a temporal dimension to generate a feature tensor sequence with a fixed time window length.

[0092] The feature tensor sequence is synchronously input into the forward and reverse computation layers of the bidirectional long short-term memory network, and the forward dancing evolution hidden state sequence and the reverse dancing evolution hidden state sequence are output respectively.

[0093] A physical rule base is constructed based on the conductor suspension point displacement constraint equation and aerodynamic torque balance condition. The forward gobling evolution hidden state sequence and the reverse gobling evolution hidden state sequence are dynamically weighted and fused to generate a gobling evolution feature matrix with physical constraint correction.

[0094] The physical constraint-corrected dancing evolution feature matrix is ​​mapped to the joint prediction matrix of dancing amplitude and frequency.

[0095] In this embodiment, the fused feature vectors are reorganized temporally to construct a feature tensor sequence with a fixed time window length, thus standardizing the temporal input format. Subsequently, the feature tensor sequence is synchronously input into the forward and backward computation layers of a Bidirectional Long Short-Term Memory (BiLSTM) network. Through bidirectional temporal learning, the historical dependencies and future trends of the dancing evolution are captured, generating complementary forward and backward dancing evolution hidden state sequences. Then, a physical rule base is constructed based on the conductor suspension point displacement constraint equation and aerodynamic torque balance conditions. The forward and backward hidden state sequences are dynamically weighted and fused, and the prediction bias of the data-driven model is corrected through physical mechanisms, generating a dancing evolution feature matrix with enhanced physical consistency. Finally, the corrected feature matrix is ​​mapped to a joint prediction matrix of dancing amplitude and frequency, achieving coupled output of multi-dimensional evolutionary trends. This embodiment of the invention uses a physical rule base to dynamically weight and fuse the hidden states of the Bidirectional Long Short-Term Memory (BiLSTM) network, which improves the physical consistency of the joint prediction matrix of dancing amplitude and frequency and the accuracy of capturing evolutionary trends.

[0096] The working process of this embodiment is illustrated below:

[0097] First, the acquired fused feature vectors are reorganized according to their temporal dimensions. Initially, the fused feature vectors are one-dimensional vectors, with the dimension corresponding to the integration of energy, frequency, and spatial features. Considering the temporal correlation of galloping evolution, a fixed time window length needs to be set based on the historical periodic characteristics of transmission line galloping (e.g., the temporal dependency of the past 10-30 sampling periods). The fused feature vectors from the continuous time series are truncated by sliding along a time window, and each truncated vector group is reassembled into a structure with dimension [missing information]. The feature tensor (where To fuse the dimension of the feature vector, multiple tensors are arranged in chronological order to generate a sequence of feature tensors, ensuring that the subsequent Bi-LSTM can capture the forward and backward correlation information in time.

[0098] The feature tensor sequence is synchronously input into a physically constrained enhanced Bi-LSTM network, which includes forward and backward computation layers. The forward computation layer processes the feature tensor sequence chronologically from past to present, learning the evolution of the dance from historical states to the current state through hidden units. The output dimension is... The positive dance evolution hidden state sequence ( (This refers to the number of hidden units in the Bi-LSTM layer). The reverse computation layer processes the feature tensor sequence in reverse chronological order from the present to the past, learning the evolutionary pattern of the dance from the future (the subsequent states of the current moment) to the current state. The output dimension is also [missing information]. The reverse dancing evolution of hidden state sequence During the two-layer computation process, the activation function and state update logic of the hidden layer units follow the standard LSTM mechanism to ensure that long-term temporal dependencies can be effectively captured.

[0099] A physical rule base is constructed based on the conductor suspension point displacement constraint equation and aerodynamic moment balance condition to achieve dynamic weighted fusion of forward and reverse hidden state sequences. In the physical rule base, the conductor suspension point displacement constraint equation stipulates that the suspension point has no displacement during galloping (because the suspension point is fixed to the tower, the displacement...). The aerodynamic moment equilibrium condition stipulates that the sum of the aerodynamic moment and the structural resistance moment at any cross-section of the conductor is 0. Based on this rule base, the conformity between the forward and reverse hidden state sequences and physical constraints at each time step is calculated (such as the deviation rate between the predicted suspension point displacement and the constraint value, and the torque imbalance). The conformity is then converted into dynamic weighting coefficients. and ( For time step, and (The higher the consistency of the hidden state sequence, the greater its weight). The forward and reverse hidden state sequences are fused using an improved dynamic weighted fusion formula, which is: , For the first The hidden state vector after physical constraint correction for each time step; the hidden state vector of all time steps. Arranged in order, the generation dimension is... The physical constraint-corrected dance evolution feature matrix.

[0100] Finally, the physically constrained dance evolution feature matrix is ​​input into a fully connected mapping layer. This layer uses linear transformation and an activation function (such as ReLU) to reduce the dimension of the feature matrix from... Convert to ( The preset time step for predicting the galloping evolution trend is 2, which corresponds to the galloping amplitude and frequency respectively. The converted data is organized according to the dimension of "prediction time step - (amplitude + frequency)" to form a joint prediction matrix of the galloping amplitude and frequency of the transmission line. Each row in the matrix corresponds to the predicted value of the galloping amplitude and the predicted value of the frequency for one prediction time step, and the core quantitative indicators of the galloping evolution trend are fully output.

[0101] As one example of the above scheme, the step of inputting the joint prediction matrix of the galloping amplitude and frequency into a pre-trained risk propagation inference model to calculate the multi-level risk quantification index of the transmission line, wherein the multi-level risk quantification index includes the probability of conductor dynamic tension exceeding the limit and the probability of insulation gap breakdown, includes the following sub-steps:

[0102] The amplitude-frequency parameter is decoupled from the joint prediction matrix of the dancing amplitude and frequency to separate the dancing amplitude component sequence and the dancing frequency component sequence;

[0103] The galloping amplitude component sequence is input into a pre-trained risk propagation inference model to calculate the dynamic tension of the conductor, and combined with the static initial tension value of the transmission line, the dynamic tension time history data of the transmission line is calculated.

[0104] Based on the calculated dynamic tension time history data and the preset conductor tension safety threshold, the percentage of tension exceeding the limit duration is statistically analyzed to obtain the probability of the conductor dynamic tension exceeding the limit.

[0105] The amplitude and frequency component sequences of the galloping motion are input into a pre-trained risk propagation inference model to perform electric field analysis on the insulation gap. Combined with the inherent insulation configuration parameters of the transmission line, the minimum air gap distance variation and local electric field intensity distribution are calculated. The probability of insulation gap breakdown risk is obtained based on the calculation results.

[0106] In this embodiment, the joint prediction matrix of galloping amplitude and frequency is decoupled, separating independent galloping amplitude component sequences and galloping frequency component sequences, achieving precise decomposition of multi-dimensional prediction parameters. Then, the amplitude component sequence is input into the risk propagation inference model, and dynamic tension is calculated in conjunction with the static initial tension value of the transmission line, generating time history data reflecting real-time tension changes. The probability of conductor dynamic tension exceeding limits is quantified by statistically analyzing the proportion of tension exceeding the limit duration. Simultaneously, the amplitude and frequency components are input into the risk propagation inference model, and insulation gap electric field analysis is performed in conjunction with line insulation configuration parameters to calculate the minimum air gap distance change and local electric field intensity distribution, thereby deriving the probability of insulation gap breakdown risk. Therefore, this embodiment achieves precise quantification of the probability of conductor tension exceeding limits and the probability of insulation gap breakdown risk through dual-channel physical modeling of dynamic tension time history analysis and insulation gap electric field calculation.

[0107] For ease of understanding, the working process of this embodiment is described in the following example:

[0108] First, the amplitude-frequency parameter decoupling is performed on the joint prediction matrix of the dancing amplitude and frequency. The dimension of this joint prediction matrix is... Parameter decoupling is achieved by extracting elements from the first column of the matrix: all elements in the first column are extracted and arranged in order of prediction time steps, resulting in a matrix of length [length missing]. The sequence of dancing amplitude components ( For the first (Dancing amplitude at time step); extract all elements from the second column of the matrix, arrange them in order of the same time step, and obtain a length of... Dancing frequency component sequence ( For the first The time step's dancing frequency ensures that the two types of parameters can be used separately for the calculation of different risk indicators in the future.

[0109] In addition, the pre-trained risk propagation inference model is trained as follows: the model is an improved inference model based on graph neural networks (GNN) combined with physical rules, which aims to characterize the propagation path and quantification relationship of transmission line galloping risk in the "conductor-insulator-tower" system. The model's input layer is defined as "risk source parameters" (such as galloping amplitude and frequency) and "system physical parameters" (such as conductor characteristics and insulation configuration), while the output layer is defined as "multi-level risk quantification indicators" (such as dynamic tension over-limit probability and insulation gap breakdown probability). In the network structure, nodes are divided into "risk unit nodes" (including conductor segment nodes, insulator nodes, and suspension point nodes) and "propagation relationship nodes" (representing the force / electric field transmission relationship between adjacent risk units). The edge weights are determined through joint training using historical risk data and physical rules. The training process uses "actual galloping accident data and simulated galloping data" as the training set (including the galloping-risk correspondence under different conductor types, spans, and wind speeds). A hybrid loss function of "physical loss and probability loss" is adopted (physical loss ensures that the inference results conform to the laws of mechanics / electric fields, while probability loss optimizes the accuracy of risk probability calculation). Iterative training is performed until the loss function converges, ultimately yielding a risk propagation inference model with generalization capabilities that meet engineering requirements.

[0110] The dance amplitude component sequence Input the "Dynamic Tension Calculation Module" of the risk propagation inference model. This module has a built-in dynamic tension calculation model for conductors, combined with the static initial tension value of the transmission line. (Determined by conductor type, span, and installation conditions, serving as known input parameters for the model), dynamic tension time history data is calculated by considering the coupling effect of galloping amplitude and frequency on tension fluctuations. An improved formula is used for calculating dynamic tension: , For the first Dynamic tension of the conductor at the time step. It is the amplitude-tension coupling coefficient (reflecting the degree of influence of galloping amplitude on tension fluctuation, determined by the conductor's elastic modulus and span). This is the frequency-tension correction factor (reflecting the influence of the galloping frequency on the average tension deviation, determined by the conductor damping characteristics). This is the dancing phase term (characterizing the temporal periodicity of tension fluctuations). The model calculates the dynamic tension step-by-step using this formula, obtaining a length of... Dynamic tension time history data .

[0111] Based on dynamic tension time history data and preset conductor tension safety threshold (Determined by the conductor's rated tensile strength and safety factor, e.g., 80% of the rated tensile strength), calculate the probability of dynamic tension exceeding the limit. The specific process is: iterate through each value in the dynamic tension time history data and determine... Is it greater than Statistical satisfaction Number of time steps The number of time steps exceeding the limit is compared with the total prediction time step. The ratio of these values ​​is used as the probability of the conductor's dynamic tension exceeding the limit, i.e.: , This represents the probability of the conductor's dynamic tension exceeding the limit. The larger this value, the higher the risk of the conductor breaking strands or breaking due to excessive tension.

[0112] The dance amplitude component sequence With frequency component sequence The "Insulation Gap Electric Field Analysis Module" is a common input to the risk propagation inference model. This module incorporates the inherent insulation configuration parameters of the transmission line (including insulator string length). Insulator string swing angle limit Rated creepage distance (etc., as known input parameters of the model), first calculate the minimum air gap distance change through geometric analysis: based on the dancing amplitude. Based on the swing characteristics of the insulator string, the minimum air gap between the conductor and the tower / ground is derived. (The greater the amplitude of the dance, the better) (The smaller the value); then, through electric field simulation, the local electric field intensity distribution is calculated, focusing on extracting... Maximum electric field strength at the corresponding location Based on the above calculation results, an improved formula for the probability of insulation gap breakdown is adopted: , This represents the probability of insulation gap breakdown. The electric field-gap joint influence coefficient (corrected for air humidity and air pressure). The breakdown field strength of air is approximately 30 kV / cm under standard atmospheric pressure, which is corrected for actual environment in the model. This is the rated minimum air gap (determined by the line voltage level). The model calculates step-by-step. Take all time steps The maximum value is taken as the final probability of insulation gap breakdown risk. The larger the value, the higher the risk of flashover of the insulator string due to gap breakdown.

[0113] Ultimately, the risk propagation inference model outputs the probability of the conductor's dynamic tension exceeding the limit. Probability of insulation gap breakdown Together, these two constitute a multi-level risk quantification index for transmission lines.

[0114] As one example of the above solution, the step of generating a corresponding risk warning based on the multi-level risk quantification indicators and sending it to the corresponding risk management device includes the following sub-steps:

[0115] Based on the probability of conductor dynamic tension exceeding the limit and the probability of insulation gap breakdown risk in the multi-level risk quantification indicators, the risk type, risk location and risk level are determined.

[0116] A structured risk warning is generated based on the risk level, and the risk warning includes risk type, risk location, and risk level information.

[0117] The risk warning will be dynamically sent to the corresponding risk management device based on the risk location.

[0118] In this embodiment, the risk type, location and level are located by analyzing multi-level risk quantification indicators (probability of conductor dynamic tension exceeding the limit and probability of insulation gap breakdown), generating risk warning prompts containing structured information, and dynamically sending them to the corresponding risk management equipment based on the risk location, thus realizing the accurate and timely push of risk information.

[0119] Specifically, the working process of this embodiment is illustrated as follows: First, the probability of conductor dynamic tension exceeding the limit and the probability of insulation gap breakdown risk in the multi-level risk quantification indicators are used as the core judgment criteria. Combined with the risk source tracing data output by the risk propagation inference model, key information is determined: if the probability of conductor dynamic tension exceeding the limit is higher than the preset threshold (e.g., 60%), the risk type is determined to be "conductor dynamic tension exceeding the limit risk". If the probability of insulation gap breakdown risk is higher than the threshold, it is determined to be "insulation gap breakdown risk". At the same time, the specific location of the risk occurrence is locked through the transmission line topology data (including span number, tower coordinates, and insulator installation position) associated with the model (e.g., "middle section of conductor in spans #28-#29 of 110kV East Line 3" and "insulator string of tower #45 of 220kV South Line 5"). The risk level is divided according to the probability range (60%-75% is level 3 risk, 76%-90% is level 2 risk, and 91% and above is level 1 risk). Subsequently, structured risk warning prompts are generated based on the determined risk levels. Key content is integrated using a fixed information module format. For example, a level 3 risk warning includes "Risk type: Over-limit dynamic tension of conductor; Risk location: Middle section of conductor in spans #28-#29 of 110kV East Line 3; Risk level: Level 3; Risk probability: 68%". Level 2 and above risk warnings additionally include "Risk impact: May cause conductor fatigue damage / insulator flashover", ensuring information completeness and conforming to maintenance reading habits. Finally, based on the risk location, a preset equipment association mapping table is queried (this table pre-stores the maintenance terminal number, monitoring center receiving address, and on-site warning device communication protocol for each line segment and tower). The structured risk warning prompts are dynamically sent to the corresponding risk management devices through a dedicated power communication network (such as the distribution automation system communication link). For example, a level 3 risk is pushed to the mobile inspection APP of the maintenance team, while level 2 and above risks are simultaneously pushed to the main system of the transmission monitoring center and the audible and visual alarm devices of the on-site towers. This achieves differentiated and accurate transmission of different risk levels, supporting maintenance personnel to quickly carry out risk handling.

[0120] See Figure 2 This is a schematic diagram of a transmission line galloping risk early warning system based on multi-physics field fusion, provided in an embodiment of the present invention. The transmission line galloping risk early warning system based on multi-physics field fusion includes:

[0121] Module 10 is used to acquire a multiphysics dataset of the galloping process of transmission lines;

[0122] The generation module 11 is used to perform spatiotemporal alignment and three-dimensional attitude reconstruction on the multiphysics dataset to generate a dynamic three-dimensional point cloud sequence of the power transmission line galloping.

[0123] The fusion module 12 is used to extract the dancing energy distribution features, frequency spectrum features and spatial trajectory features from the dynamic three-dimensional point cloud sequence, and to fuse the three features to obtain a fused feature vector.

[0124] Prediction module 13 is used to input the fused feature vector into a physically constrained enhanced bidirectional long short-term memory network to predict the galloping evolution trend and output a joint prediction matrix of the galloping amplitude and frequency of the transmission line.

[0125] Calculation module 14 is used to input the joint prediction matrix of the galloping amplitude and frequency into a pre-trained risk propagation inference model to calculate the multi-level risk quantification index of the transmission line.

[0126] The early warning module 15 is used to generate corresponding risk warning prompts based on the multi-level risk quantification indicators and send them to the corresponding risk management device.

[0127] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0128] By acquiring a multiphysics dataset of the transmission line galloping process, multidimensional physical information related to galloping is collected. The multiphysics dataset is spatiotemporally aligned and three-dimensional attitude reconstructed to generate a dynamic three-dimensional point cloud sequence, eliminating spatiotemporal biases. Galloping energy distribution features, frequency spectrum features, and spatial trajectory features are extracted from the dynamic three-dimensional point cloud sequence and fused to obtain a fused feature vector. This multidimensional feature vector comprehensively characterizes the energy, frequency, and spatial motion characteristics of the galloping, avoiding the limitations of single features. The fused feature vector is input into a physically constrained bidirectional long short-term memory network (LSTM). The LSTM captures forward and backward correlation information of the time-series data, and the enhanced physical constraints ensure that the prediction conforms to actual physical laws, thereby better predicting the galloping evolution trend and outputting a joint prediction matrix of galloping amplitude and frequency. The joint prediction matrix of galloping amplitude and frequency is input into a pre-trained risk propagation inference model, transforming the predicted galloping features into graded and quantified risk indicators, achieving multi-level risk quantification. Finally, a risk warning is generated based on the multi-level risk quantification indicators and sent to the corresponding risk management equipment to complete the warning. In summary, the embodiments of the present invention can accurately predict the galloping evolution trend of transmission lines by acquiring multi-physics field data, integrating the data and extracting and fusing multi-dimensional features, using a physically constrained enhanced bidirectional long short-term memory network to accurately predict the galloping evolution trend, and quantifying multi-level risks through a risk propagation inference model, thereby accurately predicting the galloping risk of transmission lines.

[0129] As one example of the above scheme, the generation module is specifically used for:

[0130] Phase synchronization calibration based on the dancing resonant frequency is performed on the motion acceleration data and video stream data in the multiphysics dataset to obtain a time-aligned dataset; the multiphysics dataset includes motion acceleration data, ambient wind speed data, and video stream data;

[0131] Based on the time-aligned dataset, the rigid body dynamics constraint model of the conductor is applied to perform spatial coordinate compensation on the environmental wind speed data to obtain spatially corrected data.

[0132] The spatial correction data is input into a three-dimensional reconstruction algorithm that integrates conductor bending stiffness parameters to generate a dynamic three-dimensional point cloud sequence of the transmission line galloping.

[0133] As one example of the above solution, the fusion module is specifically used for:

[0134] Based on the coupling equation of the aerodynamic structure of the conductor, the dynamic three-dimensional point cloud sequence is subjected to vortex-induced vibration mode separation to obtain the galloping energy distribution characteristics.

[0135] The frequency spectrum features are obtained by applying a dancing dominant frequency adaptive filter to the dynamic three-dimensional point cloud sequence.

[0136] The trajectory curvature of the dynamic three-dimensional point cloud sequence is corrected based on the catenary equation of the conductor to obtain spatial trajectory features.

[0137] The energy distribution characteristics, frequency spectrum characteristics, and spatial trajectory characteristics of the dancing motion are fused to obtain the fused feature vector.

[0138] As one example of the above scheme, the prediction module is specifically used for:

[0139] The fused feature vectors are reorganized in a temporal dimension to generate a feature tensor sequence with a fixed time window length.

[0140] The feature tensor sequence is synchronously input into the forward and reverse computation layers of the bidirectional long short-term memory network, and the forward dancing evolution hidden state sequence and the reverse dancing evolution hidden state sequence are output respectively.

[0141] A physical rule base is constructed based on the conductor suspension point displacement constraint equation and aerodynamic torque balance condition. The forward gobling evolution hidden state sequence and the reverse gobling evolution hidden state sequence are dynamically weighted and fused to generate a gobling evolution feature matrix with physical constraint correction.

[0142] The physical constraint-corrected dancing evolution feature matrix is ​​mapped to the joint prediction matrix of dancing amplitude and frequency.

[0143] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0144] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for early warning of transmission line galloping risk based on multi-physics field fusion, characterized in that, include: Obtain a multiphysics dataset of the galloping process of power transmission lines; Spatiotemporal alignment and 3D pose reconstruction are performed on the multiphysics dataset to generate a dynamic 3D point cloud sequence of the power transmission line galloping. The three features of dancing energy distribution, frequency spectrum and spatial trajectory are extracted from the dynamic three-dimensional point cloud sequence, and the three features are fused to obtain a fused feature vector. The fused feature vector is input into a physically constrained enhanced bidirectional long short-term memory network to predict the galloping evolution trend, and the joint prediction matrix of the galloping amplitude and frequency of the transmission line is output. The joint prediction matrix of the galloping amplitude and frequency is input into a pre-trained risk propagation inference model to calculate the multi-level risk quantification index of the transmission line. Based on the multi-level risk quantification indicators, corresponding risk warning prompts are generated and sent to the corresponding risk management devices; The process of performing spatiotemporal alignment and 3D pose reconstruction on the multiphysics dataset to generate a dynamic 3D point cloud sequence of the power transmission line galloping includes the following sub-steps: Phase synchronization calibration based on the dancing resonant frequency is performed on the motion acceleration data and video stream data in the multiphysics dataset to obtain a time-aligned dataset; the multiphysics dataset includes motion acceleration data, ambient wind speed data, and video stream data; Based on the time-aligned dataset, the rigid body dynamics constraint model of the conductor is applied to perform spatial coordinate compensation on the environmental wind speed data to obtain spatially corrected data. The spatial correction data is input into a three-dimensional reconstruction algorithm that integrates conductor bending stiffness parameters to generate a dynamic three-dimensional point cloud sequence of the transmission line galloping. The step of inputting the fused feature vector into a physically constrained enhanced bidirectional long short-term memory network for gobbling evolution trend prediction, and outputting a joint prediction matrix of the gobbling amplitude and frequency of the transmission line, includes the following sub-steps: The fused feature vectors are reorganized in a temporal dimension to generate a feature tensor sequence with a fixed time window length. The feature tensor sequence is synchronously input into the forward and reverse computation layers of the bidirectional long short-term memory network, and the forward dancing evolution hidden state sequence and the reverse dancing evolution hidden state sequence are output respectively. A physical rule base is constructed based on the conductor suspension point displacement constraint equation and aerodynamic torque balance condition. The forward gobling evolution hidden state sequence and the reverse gobling evolution hidden state sequence are dynamically weighted and fused to generate a gobling evolution feature matrix with physical constraint correction. The physical constraint-corrected dancing evolution feature matrix is ​​mapped to the joint prediction matrix of dancing amplitude and frequency.

2. The method for early warning of transmission line galloping risk based on multi-physics field fusion as described in claim 1, characterized in that, The process of extracting the dancing energy distribution features, frequency spectrum features, and spatial trajectory features from the dynamic 3D point cloud sequence, and fusing these three features to obtain a fused feature vector, includes the following sub-steps: Based on the coupling equation of the aerodynamic structure of the conductor, the dynamic three-dimensional point cloud sequence is subjected to vortex-induced vibration mode separation to obtain the galloping energy distribution characteristics. The frequency spectrum features are obtained by applying a dancing dominant frequency adaptive filter to the dynamic three-dimensional point cloud sequence. The trajectory curvature of the dynamic three-dimensional point cloud sequence is corrected based on the catenary equation of the conductor to obtain spatial trajectory features. The energy distribution characteristics, frequency spectrum characteristics, and spatial trajectory characteristics of the dancing motion are fused to obtain the fused feature vector.

3. The method for early warning of transmission line galloping risk based on multiphysics field fusion as described in claim 2, characterized in that, The step of inputting the joint prediction matrix of the galloping amplitude and frequency into a pre-trained risk propagation inference model to calculate the multi-level risk quantification index of the transmission line, which includes the probability of conductor dynamic tension exceeding the limit and the probability of insulation gap breakdown, includes the following sub-steps: The amplitude-frequency parameter is decoupled from the joint prediction matrix of the dancing amplitude and frequency to separate the dancing amplitude component sequence and the dancing frequency component sequence; The galloping amplitude component sequence is input into a pre-trained risk propagation inference model to calculate the dynamic tension of the conductor, and combined with the static initial tension value of the transmission line, the dynamic tension time history data of the transmission line is calculated. Based on the calculated dynamic tension time history data and the preset conductor tension safety threshold, the percentage of tension exceeding the limit duration is statistically analyzed to obtain the probability of the conductor dynamic tension exceeding the limit. The amplitude and frequency component sequences of the galloping motion are input into a pre-trained risk propagation inference model to perform electric field analysis on the insulation gap. Combined with the inherent insulation configuration parameters of the transmission line, the minimum air gap distance variation and local electric field intensity distribution are calculated. The probability of insulation gap breakdown risk is obtained based on the calculation results.

4. The method for early warning of transmission line galloping risk based on multiphysics field fusion as described in claim 3, characterized in that, The step of generating corresponding risk warning prompts based on the multi-level risk quantification indicators and sending them to the corresponding risk management devices includes the following sub-steps: Based on the probability of conductor dynamic tension exceeding the limit and the probability of insulation gap breakdown risk in the multi-level risk quantification indicators, the risk type, risk location and risk level are determined. A structured risk warning is generated based on the risk level, and the risk warning includes risk type, risk location, and risk level information. The risk warning will be dynamically sent to the corresponding risk management device based on the risk location.

5. A transmission line galloping risk early warning system based on multi-physics field fusion, characterized in that, include: The acquisition module is used to acquire multiphysics datasets of the galloping process of transmission lines; The generation module is used to perform spatiotemporal alignment and three-dimensional attitude reconstruction on the multiphysics dataset to generate a dynamic three-dimensional point cloud sequence of the power transmission line galloping. The fusion module is used to extract the dancing energy distribution features, frequency spectrum features and spatial trajectory features from the dynamic three-dimensional point cloud sequence, and to fuse the three features to obtain a fused feature vector. The prediction module is used to input the fused feature vector into a physically constrained enhanced bidirectional long short-term memory network to predict the galloping evolution trend and output a joint prediction matrix of the galloping amplitude and frequency of the transmission line. The calculation module is used to input the joint prediction matrix of the galloping amplitude and frequency into the pre-trained risk propagation inference model to calculate the multi-level risk quantification index of the transmission line. The early warning module is used to generate corresponding risk early warning prompts based on the multi-level risk quantification indicators and send them to the corresponding risk management devices. Specifically, the generation module is used for: Phase synchronization calibration based on the dancing resonant frequency is performed on the motion acceleration data and video stream data in the multiphysics dataset to obtain a time-aligned dataset; the multiphysics dataset includes motion acceleration data, ambient wind speed data, and video stream data; Based on the time-aligned dataset, the rigid body dynamics constraint model of the conductor is applied to perform spatial coordinate compensation on the environmental wind speed data to obtain spatially corrected data. The spatial correction data is input into a three-dimensional reconstruction algorithm that integrates conductor bending stiffness parameters to generate a dynamic three-dimensional point cloud sequence of the transmission line galloping. The prediction module is specifically used for: The fused feature vectors are reorganized in a temporal dimension to generate a feature tensor sequence with a fixed time window length. The feature tensor sequence is synchronously input into the forward and reverse computation layers of the bidirectional long short-term memory network, and the forward dancing evolution hidden state sequence and the reverse dancing evolution hidden state sequence are output respectively. A physical rule base is constructed based on the conductor suspension point displacement constraint equation and aerodynamic torque balance condition. The forward gobling evolution hidden state sequence and the reverse gobling evolution hidden state sequence are dynamically weighted and fused to generate a gobling evolution feature matrix with physical constraint correction. The physical constraint-corrected dancing evolution feature matrix is ​​mapped to the joint prediction matrix of dancing amplitude and frequency.

6. The transmission line galloping risk early warning system based on multiphysics field fusion as described in claim 5, characterized in that, The fusion module is specifically used for: Based on the coupling equation of the aerodynamic structure of the conductor, the dynamic three-dimensional point cloud sequence is subjected to vortex-induced vibration mode separation to obtain the galloping energy distribution characteristics. The frequency spectrum features are obtained by applying a dancing dominant frequency adaptive filter to the dynamic three-dimensional point cloud sequence. The trajectory curvature of the dynamic three-dimensional point cloud sequence is corrected based on the catenary equation of the conductor to obtain spatial trajectory features. The energy distribution characteristics, frequency spectrum characteristics, and spatial trajectory characteristics of the dancing motion are fused to obtain the fused feature vector.

Images