Overhead transmission tower lightning current monitoring system and method

Abstract

The invention discloses an overhead transmission tower lightning current monitoring system and method, and the method comprises the following steps: installing a multi-mode sensor module on a tower, and collecting a lightning current signal and structure response data; utilizing a local sparse dynamic window optimization algorithm to process the data segment by segment, extracting dynamic features and generating a multi-scale feature map; modeling a lightning stroke propagation rule through a space-time generative adversarial network, and generating a space-time characteristic matrix; constructing a spatial correlation map by combining a hypergeometric map embedding algorithm, and identifying a lightning stroke influence range and a diffusion path; dynamically adjusting a communication path and frequency distribution by adopting a Bayesian game optimization model; analyzing the potential influence of lightning stroke on the structure based on a multi-task self-supervised learning model, and predicting a time evolution trend; and comprehensively evaluating real-time and historical data through the remote monitoring terminal, and generating a risk prediction report and a protection reinforcement strategy. According to the invention, the problems of real-time performance, accuracy and environmental adaptability of lightning current monitoring are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and particularly to an overhead transmission tower lightning current monitoring system and method. Background Art

[0002] With the continuous expansion of the power transmission network and the wide application of high-voltage power transmission technology, as an important part of the power transmission system, the operation safety of overhead transmission towers directly affects the stability and reliability of the power grid. Overhead transmission towers are vulnerable to lightning strikes under high-voltage operating conditions, which not only causes damage to power transmission equipment but also may trigger large-scale power outages. Although there has been some research on lightning current monitoring in the prior art, there are still many deficiencies.

[0003] Traditional lightning current monitoring technologies have great limitations in terms of real-time performance and accuracy. Existing methods mostly rely on a single type of sensor, such as electromagnetic induction sensors or current transformers. These sensors have limited capabilities in capturing the amplitude, waveform, and duration of complex lightning current signals. The prior art usually cannot accurately perceive the structural response characteristics of the tower during lightning strikes, which results in the inability of the lightning current monitoring system to provide comprehensive characteristic data in practical applications and seriously affects the accuracy of the monitoring results.

[0004] Existing lightning strike monitoring methods lack in-depth analysis of the propagation laws of lightning strike events. Current technologies mostly use static models or simple algorithms to evaluate the lightning strike influence range and cannot dynamically capture the propagation path and influence area of lightning strikes in the tower network. This static evaluation method is difficult to meet the real-time requirements in complex lightning strike events, resulting in large errors in lightning strike risk identification and diffusion path prediction and affecting the accuracy of fault rapid location and early warning.

[0005] Existing lightning strike monitoring technologies have deficiencies in data processing and analysis methods. Traditional methods mostly use basic signal processing and statistical analysis means, fail to fully utilize the potential of multi-modal data, lack effective means for joint modeling of lightning current signals and tower structure response data, the fusion processing and feature extraction of data are not deep enough, and it is difficult to build a complete dynamic risk assessment model. The prior art has not been fully optimized in terms of long-term data storage and environmental adaptability. Especially in high electromagnetic interference and bad weather conditions, the stability and accuracy of the monitoring system are difficult to guarantee.

[0006] The dynamic adaptation ability of existing wireless communication modules is weak and cannot effectively support the real-time transmission and remote monitoring of large-scale lightning strike data. Traditional communication path optimization strategies mainly rely on static allocation models and have insufficient adaptability to complex environmental variables, resulting in data transmission delays and unreasonable frequency allocation and affecting the overall efficiency of the monitoring system.

[0007] In the subsequent analysis and decision support of lightning strike events, the existing technologies lack a systematic risk assessment and protection strategy generation mechanism. Most traditional methods are based on fixed rules or preset models, making it difficult to dynamically adjust lightning protection schemes and reinforcement strategies, resulting in the inability to provide targeted decision support in different operating scenarios.

[0008] Therefore, how to provide a lightning current monitoring system and method for overhead transmission towers is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0009] An object of the present invention is to propose a lightning current monitoring system and method for overhead transmission towers, which makes full use of dynamic optimization algorithms and spatio-temporal modeling technologies, and details the entire process from lightning current signal acquisition to analysis prediction and protection strategy generation, with the advantages of high real-time performance, strong accuracy, excellent environmental adaptability, and strong remote decision support capabilities.

[0010] A lightning current monitoring system and method for overhead transmission towers according to an embodiment of the present invention includes the following steps:

[0011] S1. Install a multi-modal sensor module at key parts of the overhead transmission tower, including a Rogowski coil current sensor, a vibration sensor, and an FPGA-based high-speed data acquisition module, for real-time acquisition of lightning current amplitude, waveform, duration, and tower structure response data;

[0012] S2. Based on the data collected by the multi-modal sensor, use the local sparse dynamic window optimization algorithm to segment the lightning current signal, and at the same time extract the dynamic features in different time periods to generate a multi-scale feature map of the lightning current;

[0013] S3. Input the multi-scale feature map of the lightning current and the tower structure response data into a spatio-temporal adversarial generation network, and model the dynamic law of the propagation of the lightning strike event through the adversarial learning mechanism in the network to generate a spatio-temporal characteristic matrix of the lightning strike event;

[0014] S4. Combine the spatio-temporal characteristic matrix, and use the hypergeometric spectral embedding algorithm to construct a spatial correlation map of the lightning strike event, and identify the lightning strike influence range and diffusion path in the tower network;

[0015] S5. Based on the spatial correlation map and real-time environmental variables, adopt a Bayesian game optimization model to dynamically adjust the transmission path and frequency allocation of the wireless communication module;

[0016] S6. In the edge computing unit, jointly model the transmitted data, and use a multi-task self-supervised learning algorithm to analyze the potential damage degree of the lightning current to the tower structure, and predict the time evolution trend of the lightning strike event;

[0017] S7. Through the remote monitoring terminal, comprehensively evaluate the analysis results of the edge computing unit and the historical data to generate a risk prediction report for lightning strike events, and at the same time put forward targeted lightning protection and structural reinforcement suggestions.

[0018] Optionally, the S1 includes the following steps:

[0019] S11. Install a multi-modal sensor module at the top and key stress-bearing parts of the overhead transmission tower, and the module combines a Rogowski coil current sensor and a vibration sensor;

[0020] S12. Convert the lightning current signal into a voltage signal through the Rogowski coil, and optimize its acquisition accuracy through anti-interference design to reduce the impact of environmental electromagnetic interference on data acquisition;

[0021] S13. Adopt a high-speed data acquisition module based on FPGA to perform analog-to-digital conversion and high-speed processing on the analog signal output by the sensor, and generate time series data related to the lightning current;

[0022] S14. Filter the voltage signal and perform pre-peak data protection in the signal conditioning circuit to avoid data loss or pre-peak distortion of the lightning current waveform in the high-frequency band;

[0023] S15. Through the software and hardware combination method of the sensor module, dynamically adjust the signal acquisition threshold to capture the complete characteristics of the lightning current waveform;

[0024] S16. Combine the output data of the vibration sensor, and synchronously fuse the characteristics of the lightning current signal and the tower structure stress state data through a specific algorithm to generate a multi-dimensional data set including the lightning current amplitude, waveform, duration and tower response characteristics;

[0025] S17. Adopt a low-power wireless communication method based on LoRa communication and 4G transmission module to upload the processed multi-dimensional feature data to the edge computing unit.

[0026] Optionally, the S2 includes the following steps:

[0027] S21. Perform time serialization processing on the lightning current signal data I(t), waveform function f(t), duration ΔT and tower structure response data R(t) collected by the multi-modal sensor to generate a signal matrix X = [I(t), f(t), ΔT, R(t)], where I(t) is the lightning current time series amplitude, f(t) is the lightning current waveform function, ΔT is the lightning current duration, and R(t) is the tower structure response time series;

[0028] S22. Use the local sparse dynamic window optimization algorithm to segment the signal matrix X, and the dynamic adjustment formula of the time window L w is:

[0029]

[0030] Among them, L base is the basic window length, α is the time-varying weight, represents the lightning current gradient, σ 2 is the standard deviation of signal change, H(t) is the signal sparse entropy value, defined as H(t) = -∑p(t)logp(t), and p(t) is the signal normalized probability density;

[0031] S23. For the signals in each time window, adopt the sparse coding optimization method to extract signal features, and the optimization objective function is:

[0032]

[0033] Among them, X w represents the signal segment within the time window, D is the sparse dictionary, F w is the sparse representation feature vector, and λ and γ are the sparse regularization and smoothing regularization coefficients;

[0034] S24. Fuse the window feature vector with the tower response data to generate the dynamic feature matrix F = [F1, F2,..., F n , where F i is the multi-modal feature representation of the i-th time window;

[0035] S25. Decompose the dynamic feature matrix through the multi-scale mapping method to generate the multi-scale feature map T = [T1, T2,..., T m ;

[0036] S26. Store the generated lightning current multi-scale feature map as a standard format data file.

[0037] Optionally, the S3 includes the following steps:

[0038] S31. Fuse the lightning current multi-scale feature map T = [T1, T2,..., T m with the tower structure response data matrix R = [R1, R2,..., R n to generate the spatio-temporal characteristic input data set X input = [T, R];

[0039] S32. Construct a spatio-temporal adversarial generation network, the network includes a generator and a discriminator, and the generator learns the dynamic law of lightning strike event propagation through the input data set X input to generate the lightning strike propagation characteristic matrix;

[0040] S33. The discriminator performs comparative learning based on the input real lightning strike characteristic matrix and the generated characteristic matrix, and continuously optimizes the output characteristic matrix of the generator;

[0041] S34. Through the alternating training of the generator and the discriminator, the generated lightning strike propagation characteristic matrix X is finally obtained output ;

[0042] S35. Decompose the generated lightning strike propagation characteristic matrix, and extract the time dimension X time and the space dimension X space , which respectively represent the propagation time law and the space diffusion path of the lightning strike event;

[0043] S36. Recombine the characteristics of the time dimension and the space dimension to generate the spatio-temporal characteristic matrix X of the lightning strike event containing the complete spatio-temporal propagation law spatio-temporal ;

[0044] S37. Use the generated spatio-temporal characteristic matrix as the input for subsequent dynamic modeling, and provide high-confidence basic data for the risk assessment and path analysis of lightning strike events.

[0045] Optionally, the S4 includes the following steps:

[0046] S41. Extract the space characteristic matrix X spatio-temporal = [X time , X space in the spatio-temporal characteristic matrix X of the lightning strike event as the input, and combine it with the initial adjacency graph G = (V, ε) of the tower network, where V is the set of tower nodes and ε is the set of edges connecting the tower nodes. W = [w space is the edge weight matrix, and its calculation formula is: ij

[0047]

[0048] Among them, X i and X j are the characteristic vectors of tower nodes i and j, σ is the control scale factor, and deg(i) is the degree of node i;

[0049] S42. Construct the initial adjacency matrix A, and perform normalization processing in combination with the space characteristic weight matrix W;

[0050] S43. Based on the hypergeometric spectral graph embedding algorithm, perform high-order characteristic embedding on the initial adjacency matrix to generate a high-order adjacency relationship matrix. The embedding formula is:

[0051]

[0052] Among them, A k is the high-order embedding matrix,​ Represents the high-order propagation operation of the adjacency matrix. D is the degree matrix, and γ i is the weight coefficient for each order embedding, used to dynamically adjust the influence of high-order characteristics;

[0053] S44. Based on the high-order adjacency relationship matrix A k calculate the spatial correlation matrix M spatial where the matrix element m ij represents the correlation strength between nodes i and j;

[0054] S45. Combine the real-time environmental variable matrix E and optimize the spatial correlation matrix through the dynamic weight adjustment model;

[0055] S46. Utilize the dynamic weight matrix W * and apply the path search algorithm to identify the lightning strike influence range and diffusion path in the tower network;

[0056] S47. Output the diffusion path result, including the set of tower nodes within the lightning strike influence range and the correlation strength of the path.

[0057] Optionally, the S5 includes the following steps:

[0058] S51. Take the spatial correlation matrix M spatial = [m ij and the real-time environmental variable matrix E = [e ij as inputs, where m ij represents the spatial correlation strength between tower nodes i and j, and e ij represents the influence factor of the environmental variable on the communication path between nodes i and j;

[0059] S52. Construct a wireless communication network topology graph G = (V, ε, W * ), where V is the set of communication nodes, ε is the set of communication paths between nodes, is the dynamic path weight matrix;

[0060] S53. Adopt the Bayesian game optimization model to dynamically optimize the selection and frequency allocation of communication paths. The objective function of the model is:

[0061]

[0062] where P is the set of paths, and f ij is the current frequency occupancy rate of communication paths i and j;

[0063] S54. In the Bayesian game optimization, through the prior probability distribution and the posterior probability distribution

[0064] Iteratively update the path weights and dynamically adjust the path selection strategy;

[0065] S55. Use the optimized communication path set P opt , select the transmission path with the minimum signal interference and the maximum path weight to achieve the dynamic optimization of communication data;

[0066] S56. Combine the optimized path set and allocate the best transmission frequency for the wireless communication module where, β is the frequency adjustment factor and F is the set of available frequencies;

[0067] S57. Upload the optimized communication path and frequency allocation scheme to the remote monitoring center to complete the dynamic adjustment of the wireless communication module and provide efficient support for subsequent data transmission.

[0068] Optionally, the S6 includes the following steps:

[0069] S61. Receive the spatial correlation matrix M spatial , the real-time environmental variable matrix E, the tower structure response data matrix R, and the lightning current multi-scale feature map T, and integrate them into the joint dataset D = {M spatial , E, R, T};

[0070] S62. Design a multi-task self-supervised learning model The model includes a main task network and an auxiliary task network, which process the spatial correlation and structural response in the dataset, as well as the time evolution characteristics respectively;

[0071] S63. Input R and M spatial to the main task network, generate the tower structure damage characteristic vector S, perform sparse optimization on the characteristic vector, and extract the key structural response characteristics;

[0072] S64. Input T and E to the auxiliary task network, generate the time evolution characteristic matrix T evolve , and smooth and optimize the time dimension change trend;

[0073] S65. Jointly process the structure damage characteristic vector S and the time evolution characteristic matrix T evolve to generate the comprehensive evaluation matrix E comprehensive , which represents the comprehensive state of the tower and the event evolution;

[0074] S66. Upload the comprehensive evaluation matrix to complete the damage analysis and evolution trend prediction of the lightning strike event, and output the structure state evaluation and lightning strike risk results.

[0075] Optionally, the S7 includes the following steps:

[0076] S71. Receive the evaluation results generated by the edge computing unit, and fuse them with the pole tower status data in the historical database to form a combined dataset for evaluation;

[0077] S72. Analyze the combined dataset, identify the damaged characteristics of the pole tower structure and the time evolution characteristics of lightning strike events, and generate a risk assessment matrix in combination with historical data;

[0078] S73. Based on the risk assessment matrix and the current environmental variable data, design an optimized lightning protection strategy to form protection suggestions for each pole tower node;

[0079] S74. Combine the reinforcement records in the historical database and the real-time evaluation data to generate a structural reinforcement strategy matrix and provide specific reinforcement suggestions;

[0080] S75. Integrate the risk assessment matrix, the lightning protection strategy and the structural reinforcement strategy to generate a standardized output report, and upload the standardized output report to the remote monitoring terminal to generate a standardized output report.

[0081] Optionally, it includes the following modules:

[0082] Multimodal sensor module: Installed at key parts of the overhead transmission pole tower. The module includes a Rogowski coil current sensor and a vibration sensor, and is used to collect the amplitude, waveform, duration of the lightning current signal and the structural response data of the pole tower in real time;

[0083] Signal processing unit: Connected to the multimodal sensor module, configured to process the collected lightning current signal and structural response data, extract dynamic features using the local sparse dynamic window optimization algorithm, and generate a multi-scale feature map of the lightning current;

[0084] Edge computing unit: Connected to the signal processing unit, configured to jointly model the multi-scale feature map of the lightning current and the structural response data based on the spatio-temporal adversarial generation network, generate a spatio-temporal characteristic matrix of the lightning strike event, and construct a spatial association map of the lightning strike event through the hypergeometric spectral embedding algorithm;

[0085] Wireless communication module: Connected to the edge computing unit, dynamically adjust the transmission path and frequency allocation based on the Bayesian game optimization model, and transmit the analysis results to the remote monitoring terminal in real time;

[0086] Remote monitoring terminal: Connected to the wireless communication module, configured to receive the analysis results of the edge computing unit, conduct a comprehensive evaluation in combination with the pole tower status data in the historical database, generate a risk prediction report for the lightning strike event, and put forward lightning protection and structural reinforcement suggestions.

[0087] The beneficial effects of the present invention are:

[0088] The present invention utilizes a local sparse dynamic window optimization algorithm to perform dynamic segmentation processing on lightning current signals, extract dynamic features in different time periods, and generate a multi-scale feature map of lightning current, solving the problems of lag and insufficient resolution in the segmented analysis of lightning current signals by traditional methods, and providing reliable data support for the accurate modeling of lightning strike events.

[0089] The present invention uses a spatio-temporal adversarial generation network to model the lightning strike propagation law, generates a spatio-temporal characteristic matrix, and constructs a spatial association map of lightning strike events through a hypergeometric spectral embedding algorithm, accurately identifying the lightning strike impact range and diffusion path, overcoming the problem of insufficient dynamic capture ability of the existing technology for the lightning strike propagation law, and providing technical guarantee for the accurate assessment of lightning strike risks.

[0090] The present invention introduces a Bayesian game optimization model, optimizes the data transmission strategy by dynamically adjusting the transmission path and frequency allocation of the wireless communication module, improves the communication efficiency and data stability of the monitoring system in a large-scale complex environment, and solves the problems of delay and insufficient adaptability of traditional communication schemes in complex environments.

[0091] Based on a multi-task self-supervised learning algorithm, the present invention combines multi-modal data for joint modeling, not only deeply analyzes the potential damage degree of the lightning current to the tower structure, but also predicts the lightning strike risk trend in advance through a time evolution characteristic prediction model, providing a reliable basis for prediction and analysis for the safe operation of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention. In the drawings:

[0093] Figure 1 is a flowchart of a lightning current monitoring system and method for an overhead transmission tower proposed by the present invention;

[0094] Figure 2 is a schematic diagram of the lightning current signal segmented processing by the local sparse dynamic window optimization algorithm proposed by the present invention;

[0095] Figure 3 is a schematic diagram of constructing a spatial association map of lightning strike events by the hypergeometric spectral embedding algorithm proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0096] Now, the present invention will be further described in detail with reference to the drawings. These drawings are all simplified schematic diagrams, only showing the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0097] Refer to Figures 1 - 3, an overhead transmission tower lightning current monitoring system and method, including the following steps:

[0098] S1. Install a multi-modal sensor module at key parts of the overhead transmission tower, including a Rogowski coil current sensor, a vibration sensor, and an FPGA-based high-speed data acquisition module, for real-time acquisition of lightning current amplitude, waveform, duration, and tower structure response data;

[0099] S2. Based on the data collected by the multi-modal sensors, use the local sparse dynamic window optimization algorithm to segment the lightning current signal, and at the same time extract the dynamic features in different time periods to generate a multi-scale feature map of the lightning current;

[0100] S3. Input the multi-scale feature map of the lightning current and the tower structure response data into the spatio-temporal adversarial generation network, and model the dynamic law of the lightning strike event propagation through the adversarial learning mechanism in the network to generate a spatio-temporal characteristic matrix of the lightning strike event;

[0101] S4. Combine the spatio-temporal characteristic matrix, and use the hypergeometric graph embedding algorithm to construct a spatial correlation graph of the lightning strike event, and identify the lightning strike influence range and diffusion path in the tower network;

[0102] S5. Based on the spatial correlation graph and real-time environmental variables, adopt the Bayesian game optimization model to dynamically adjust the transmission path and frequency allocation of the wireless communication module;

[0103] S6. In the edge computing unit, jointly model the transmitted data, and use the multi-task self-supervised learning algorithm to analyze the potential damage degree of the lightning current to the tower structure, and predict the time evolution trend of the lightning strike event;

[0104] S7. Through the remote monitoring terminal, comprehensively evaluate the analysis results of the edge computing unit and the historical data to generate a risk prediction report of the lightning strike event, and at the same time put forward targeted lightning protection and structural reinforcement suggestions.

[0105] In this embodiment, S1 includes the following steps:

[0106] S11. Install a multi-modal sensor module at the top and key stress parts of the overhead transmission tower, and the module combines a Rogowski coil current sensor and a vibration sensor;

[0107] S12. Convert the lightning current signal into a voltage signal through the Rogowski coil, and optimize its acquisition accuracy through anti-interference design to reduce the influence of environmental electromagnetic interference on data acquisition;

[0108] S13. Adopt an FPGA-based high-speed data acquisition module to perform analog-to-digital conversion and high-speed processing on the analog signal output by the sensor to generate time series data related to the lightning current;

[0109] S14. Filter the voltage signal and protect the leading-edge data in the signal conditioning circuit to avoid data loss or leading-edge distortion of the lightning current waveform in the high-frequency band;

[0110] S15. Dynamically adjust the signal acquisition threshold through the combination of software and hardware of the sensor module to capture the complete characteristics of the lightning current waveform;

[0111] S16. Combine the output data of the vibration sensor, and synchronously fuse the characteristics of the lightning current signal with the tower force state data through a specific algorithm to generate a multi-dimensional data set including the lightning current amplitude, waveform, duration, and tower response characteristics;

[0112] S17. Adopt a low-power wireless communication method based on LoRa communication and 4G transmission module to upload the processed multi-dimensional feature data to the edge computing unit.

[0113] In this embodiment, S2 includes the following steps:

[0114] S21. Perform time serialization processing on the lightning current signal data I(t), waveform function f(t), duration ΔT, and tower structure response data R(t) collected by the multi-modal sensor to generate a signal matrix X = [I(t), f(t), ΔT, R(t)], where I(t) is the lightning current time series amplitude, f(t) is the lightning current waveform function, ΔT is the lightning current duration, and R(t) is the tower structure response time series;

[0115] S22. Segment the signal matrix X using the local sparse dynamic window optimization algorithm, and the dynamic adjustment formula of the time window L w is:

[0116]

[0117] where L base is the basic window length, α is the time change weight, represents the lightning current gradient, σ 2 is the standard deviation of signal change, and H(t) is the signal sparse entropy value, defined as H(t) = -∑p(t)logp(t), where p(t) is the signal normalized probability density;

[0118] S23. For the signal in each time window, use the sparse coding optimization method to extract the signal features, and the optimization objective function is:

[0119]

[0120] where X w represents the signal segment within the time window, D is the sparse dictionary, and F wIt is a sparse representation feature vector, and λ and γ are sparse regularization and smoothing regularization coefficients;

[0121] S24. Integrate the window feature vector with the tower response data to generate a dynamic feature matrix F = [F1, F2, …, F n , where F i is the multi-modal feature representation of the i-th time window;

[0122] S25. Decompose the dynamic feature matrix through a multi-scale mapping method to generate a multi-scale feature map T = [T1, T2, …, T m ;

[0123] S26. Store the generated multi-scale feature map of lightning current as a standard format data file.

[0124] In this embodiment, S3 includes the following steps:

[0125] S31. Integrate the multi-scale feature map of lightning current T = [T1, T2, …, T m with the tower structure response data matrix R = [R1, R2, …, R n to generate a spatio-temporal characteristic input data set X input = [T, R];

[0126] S32. Construct a spatio-temporal adversarial generation network, which includes a generator and a discriminator. The generator learns the dynamic law of lightning strike event propagation through the input data set X input and generates a lightning strike propagation characteristic matrix;

[0127] S33. The discriminator performs comparative learning based on the input real lightning strike characteristic matrix and the generated characteristic matrix, and continuously optimizes the output characteristic matrix of the generator;

[0128] S34. Through the alternating training of the generator and the discriminator, finally obtain the generated lightning strike propagation characteristic matrix X output ;

[0129] S35. Decompose the generated lightning strike propagation characteristic matrix, and extract the time dimension X time and the space dimension X space , which respectively represent the propagation time law and the space diffusion path of the lightning strike event;

[0130] S36. Recombine the characteristics of the time dimension and the space dimension to generate a lightning strike event spatio-temporal characteristic matrix X spatio-temporal containing the complete spatio-temporal propagation law;

[0131] S37. Use the generated spatio-temporal characteristic matrix as the input for subsequent dynamic modeling, and provide high-confidence basic data for the risk assessment and path analysis of lightning strike events.

[0132] In this embodiment, S4 includes the following steps:

[0133] S41. Take the spatial feature matrix X spatio-temporal = [X time , X space in the spatio-temporal feature matrix of lightning strike events as the input, and combine it with the initial adjacency graph G=(V, ε) of the tower network, where V is the set of tower nodes and ε is the set of edges connecting the tower nodes. W = [w space is the edge weight matrix, and its calculation formula is: ij

[0134]

[0135] where X i and X j are the feature vectors of tower nodes i and j, σ is the control scale factor, and deg(i) is the degree of node i;

[0136] S42. Construct the initial adjacency matrix A and perform normalization processing in combination with the spatial feature weight matrix W;

[0137] S43. Based on the hypergeometric spectral embedding algorithm, perform high-order feature embedding on the initial adjacency matrix to generate a high-order adjacency relationship matrix. The embedding formula is:

[0138]

[0139] where A k is the high-order embedding matrix, represents the high-order propagation operation of the adjacency matrix, D is the degree matrix, and γ i is the weight coefficient for each order of embedding, which is used to dynamically adjust the influence of high-order features;

[0140] S44. Based on the high-order adjacency relationship matrix A k , calculate the spatial correlation matrix M spatial , where the matrix element m ij represents the correlation strength between nodes i and j;

[0141] S45. Combine the real-time environmental variable matrix E and optimize the spatial correlation matrix through the dynamic weight adjustment model;

[0142] S46. Use the dynamic weight matrix W * , apply the path search algorithm to identify the lightning strike influence range and diffusion path in the tower network;

[0143] S47. Output the diffusion path result, including the set of tower nodes within the lightning strike influence range and the correlation strength of the path. ​

[0144] In this embodiment, S5 includes the following steps:

[0145] S51. Using the spatial correlation matrix M spatial =[m ij and the real-time environmental variable matrix E = [e ij as inputs, where m ij represents the spatial correlation strength between pole tower nodes i and j, and e ij represents the influence factor of environmental variables on the communication path between nodes i and j;

[0146] S52. Construct a wireless communication network topology graph G=(V, ε, W * ), where V is the set of communication nodes, ε is the set of communication paths between nodes, is the dynamic path weight matrix;

[0147] S53. Adopt a Bayesian game optimization model to dynamically optimize the selection of communication paths and frequency allocation. The objective function of the model is:

[0148]

[0149] where P is the set of paths, and f ij is the current frequency occupancy rate of communication paths i and j;

[0150] S54. In the Bayesian game optimization, the path weights are iteratively updated through the prior probability distribution and the posterior probability distribution

[0151] to dynamically adjust the path selection strategy;

[0152] S55. Using the optimized communication path set P opt , select the transmission path with the minimum signal interference and the maximum path weight to achieve the dynamic optimization of communication data;

[0153] S56. Combining the optimized path set, allocate the best transmission frequency for the wireless communication module where, β is the frequency adjustment factor, and F is the set of available frequencies;

[0154] S57. Upload the optimized communication path and frequency allocation scheme to the remote monitoring center to complete the dynamic adjustment of the wireless communication module and provide efficient support for subsequent data transmission.

[0155] In this embodiment, S6 includes the following steps:

[0156] S61. Receive the spatial correlation matrix M spatial, the real-time environmental variable matrix E, the tower structure response data matrix R, and the lightning current multi-scale feature map T are integrated into the joint dataset D = {M spatial , E, R, T};

[0157] S62. Design a multi-task self-supervised learning model The model includes a main task network and an auxiliary task network, which process the spatial association and structural response in the dataset, as well as the time evolution characteristics respectively;

[0158] S63. Input R and M spatial to the main task network, generate the tower structure damage characteristic vector S, perform sparse optimization on the characteristic vector, and extract the key structural response characteristics;

[0159] S64. Input T and E to the auxiliary task network, generate the time evolution characteristic matrix T evolve , and smooth and optimize the change trend in the time dimension;

[0160] S65. Jointly process the structure damage characteristic vector S and the time evolution characteristic matrix T evolve to generate the comprehensive evaluation matrix E comprehensive , representing the comprehensive state of the tower and the evolution of events;

[0161] S66. Upload the comprehensive evaluation matrix to complete the damage analysis and evolution trend prediction of the lightning strike event, and output the structure state evaluation and lightning strike risk results.

[0162] In this embodiment, S7 includes the following steps:

[0163] S71. Receive the evaluation results generated by the edge computing unit, and fuse them with the tower state data in the historical database to form a joint dataset for evaluation;

[0164] S72. Analyze the joint dataset, identify the tower structure damage characteristics and the time evolution characteristics of the lightning strike event, and generate a risk assessment matrix in combination with historical data;

[0165] S73. Based on the risk assessment matrix and the current environmental variable data, design an optimized lightning protection strategy to form protection suggestions for each tower node;

[0166] S74. Combine the reinforcement records in the historical database and the real-time evaluation data to generate a structure reinforcement strategy matrix and provide specific reinforcement suggestions;

[0167] S75. Integrate the risk assessment matrix, the lightning protection strategy, and the structure reinforcement strategy to generate a standardized output report, and upload the standardized output report to the remote monitoring terminal to generate a standardized output report.

[0168] In this embodiment, the following modules are included:

[0169] Multimodal sensor module: Installed at key positions of overhead transmission towers. The module includes a Rogowski coil current sensor and a vibration sensor, and is used to collect the amplitude, waveform, duration of lightning current signals and the structural response data of the towers in real time.

[0170] Signal processing unit: Connected to the multimodal sensor module, configured to process the collected lightning current signals and structural response data, extract dynamic features using the local sparse dynamic window optimization algorithm, and generate a multi-scale feature map of lightning current.

[0171] Edge computing unit: Connected to the signal processing unit, configured to jointly model the multi-scale feature map of lightning current and the structural response data based on the spatio-temporal adversarial generation network, generate a spatio-temporal characteristic matrix of lightning strike events, and construct a spatial association map of lightning strike events through the hypergeometric spectral embedding algorithm.

[0172] Wireless communication module: Connected to the edge computing unit, dynamically adjusts the transmission path and frequency allocation based on the Bayesian game optimization model, and transmits the analysis results to the remote monitoring terminal in real time.

[0173] Remote monitoring terminal: Connected to the wireless communication module, configured to receive the analysis results of the edge computing unit, comprehensively evaluate by combining the tower state data in the historical database, generate a risk prediction report of lightning strike events, and propose lightning protection and structural reinforcement suggestions.

[0174] Embodiment 1:

[0175] In a simulation environment, a simulation test of lightning strike events was carried out on a section of overhead transmission line. The transmission line consists of 30 towers, and the impacts of multiple-frequency and high-intensity lightning strike events under thunderstorm weather were simulated to verify the effectiveness of the lightning current monitoring method for overhead transmission towers proposed by the present invention. Traditional lightning current monitoring systems are difficult to capture lightning strike events in real time and accurately in complex environments, and lack dynamic evaluation of the lightning strike impact range, diffusion path and tower structure damage. This embodiment verifies the improvement of the present invention in terms of real-time performance, accuracy and adaptability through simulation tests.

[0176] Install a multimodal sensor module at the key stress positions of each tower, including a Rogowski coil current sensor and a vibration sensor, to collect the amplitude, waveform, duration of lightning current signals and the structural response data of the tower respectively. These sensors transmit the multimodal data to the edge computing unit in real time through a low-power wireless communication module.

[0177] In the simulation environment, lightning strike events with lightning current signal amplitude ranges from 10 kA to 200 kA are introduced to dynamically simulate the impacts of lightning strikes with different intensities on the tower. The edge computing unit uses the local sparse dynamic window optimization algorithm to segment the lightning current signal, extract feature data, and generate a multi-scale feature map of the lightning current. Subsequently, the spatio-temporal adversarial generation network is used to model the propagation law of lightning strikes, generate a spatio-temporal characteristic matrix, and combine with the hypergeometric spectral embedding algorithm to construct a spatial association map of lightning strike events, accurately identifying the lightning strike impact range and diffusion path.

[0178] The simulation further adopts the Bayesian game optimization model to dynamically adjust the wireless communication path and frequency allocation to ensure data transmission stability under high interference conditions. Combining with the multi-task self-supervised learning algorithm, analyze the potential damage degree of lightning strike events to the tower and predict the time evolution trend. Through the remote monitoring terminal, combined with historical simulation data, generate a risk prediction report and protection and reinforcement suggestions for lightning strike events.

[0179] Simulation Data and Effect Verification

[0180] Table 1 Performance Comparison Data between the Present Invention and Traditional Methods

[0181] Test indicators Traditional monitoring system System of the present invention Improvement effect Average monitoring delay (seconds) 15.8 4.2 Reduced by 73.4% Lightning strike event capture rate 82.10％ 97.80％ Improved by 15.7% Data transmission success rate 79.50％ 96.20％ Improved by 16.7% Diffusion path prediction accuracy rate No relevant data 93.60％ New capabilities Structural damage prediction coincidence rate No relevant data 88.20％ New capabilities Damage rate after protection (simulated) 13.90％ 3.70％ Reduced by 10.2%

[0182] In the simulation environment, a total of 60 lightning strike events are simulated. The test results show that the present invention is significantly superior to traditional methods in various performance indicators; Real-time test: The average acquisition and analysis time of the lightning current signal is 4.2 seconds, significantly better than 15.8 seconds of the traditional monitoring system, with a delay reduction of 73.4%. Accuracy verification: The capture rate reaches 97.8%, much higher than 82.1% of the traditional system. Diffusion path prediction: The prediction accuracy of the lightning strike impact range and diffusion path reaches 93.6%, providing reliable support for quickly locating faults. Structural damage prediction: In the simulated 15 severe lightning strike events, the coincidence rate between the predicted damage degree by the system and the actual situation is 88.2%. Communication adaptability: Under high electromagnetic interference conditions, the data transmission success rate of the system reaches 96.2%, significantly higher than 79.5% of traditional methods. Protection effect: After implementing the generated protection suggestions, the damage rate of the simulated protected towers drops to 3.7%, while the damage rate of unprotected towers is 13.9%.

[0183] The present invention significantly improves the real-time performance and accuracy of lightning current monitoring in the simulation environment. Especially under complex high-interference conditions, the data transmission and diffusion path prediction capabilities of the system show high reliability. By analyzing the generated protection and reinforcement suggestions, the damage probability of the tower after lightning strikes is effectively reduced. The present invention has high technical adaptability and practicability, providing reliable technical guarantees for lightning strike event monitoring and power grid operation safety.

[0184] The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent substitutions or changes should be covered within the protection scope of the present invention.

Claims

1. A method for monitoring lightning current of an overhead transmission tower, characterized in that: The steps include: S1. Install multi-modal sensor modules at key locations of overhead transmission towers, including Rogowski coil current sensors, vibration sensors, and FPGA-based high-speed data acquisition modules, to collect lightning current amplitude, waveform, duration, and tower structure response data in real time; S2. Based on the data collected by multimodal sensors, the lightning current signal is processed in segments using a local sparse dynamic window optimization algorithm, and the dynamic features in different time periods are extracted to generate a multi-scale feature map of the lightning current; S3, inputting the lightning current multi-scale characteristic map and the tower structure response data into the spatiotemporal adversarial generation network, modeling the dynamic law of lightning event propagation through the adversarial learning mechanism in the network, and generating the spatiotemporal characteristic matrix of lightning events; S4. Combined with the spatiotemporal characteristic matrix, the hypergeometric graph embedding algorithm is used to construct the spatial correlation graph of lightning strike events, and the lightning strike impact range and diffusion path are identified in the tower network; S5. Based on the spatial correlation graph and real-time environmental variables, the transmission path and frequency allocation of the wireless communication module are dynamically adjusted using the Bayesian game optimization model; S6. In the edge computing unit, the transmitted data is jointly modeled, and a multi-task self-supervised learning algorithm is used to analyze the potential damage of lightning current to the tower structure and predict the time evolution trend of lightning strike events; S7. Through the remote monitoring terminal, the analysis results of the edge computing unit are comprehensively evaluated with historical data to generate a risk prediction report for lightning strikes, and targeted lightning protection and structural reinforcement suggestions are put forward.

2. The method for monitoring lightning current of an overhead transmission tower according to claim 1, characterized in that: The S1 comprises the following steps: S11, installing a multi-modal sensor module at the top and key stress-bearing parts of the overhead transmission tower, wherein the module combines a Rogowski coil current sensor and a vibration sensor; S12, converting lightning current signals into voltage signals through Rogowski coils, and optimizing their acquisition accuracy through anti-interference design to reduce the impact of environmental electromagnetic interference on data acquisition; S13, using a high-speed data acquisition module based on FPGA to perform analog-to-digital conversion and high-speed processing on the analog signal output by the sensor to generate time series data related to the lightning current; S14. Filter the voltage signal and protect the front peak data in the signal conditioning circuit to avoid data loss or front peak distortion of the lightning current waveform in the high frequency band; S15, through the software and hardware combination method of the sensor module, dynamically adjust the signal acquisition threshold to capture the complete characteristics of the lightning current waveform; S16. Combine the output data of the vibration sensor, and use a specific algorithm to synchronously integrate the characteristics of the lightning current signal with the stress state data of the tower, to generate a multi-dimensional data set including the lightning current amplitude, waveform, duration and tower response characteristics; S17, using a low-power wireless communication method based on LoRa communication and 4G transmission module to upload the processed multi-dimensional feature data to the edge computing unit.

3. The method for monitoring lightning current of an overhead transmission tower according to claim 1, characterized in that: The S2 comprises the following steps: S21, performing time series processing on the lightning current signal data I(t), waveform function f(t), duration ΔT and tower structure response data R(t) collected by the multimodal sensor to generate a signal matrix X=[I(t), f(t), ΔT, R(t)], where I(t) is the amplitude of the lightning current time series, f(t) is the lightning current waveform function, ΔT is the lightning current duration, and R(t) is the tower structure response time series; S22, using the local sparse dynamic window optimization algorithm to process the signal matrix X in segments, the time window L w The dynamic adjustment formula is: Among them, L base is the basic window length, α is the time-varying weight, represents the lightning current gradient, σ 2 is the standard deviation of the signal change, H(t) is the signal sparse entropy value, defined as H(t) = -∑p(t)logp(t), p(t) is the normalized probability density of the signal; S23. For the signal in each time window, a sparse coding optimization method is used to extract signal features, and the optimization objective function is: Among them, X w represents the signal fragment within the time window, D is the sparse dictionary, and F w is the sparse representation feature vector, λ and γ are the sparse regularization and smooth regularization coefficients; S24, fuse the window feature vector with the tower response data to generate a dynamic feature matrix F=[F1,F2,…,F n ], where F i is the multimodal feature representation of the i-th time window; S25, decompose the dynamic feature matrix through a multi-scale mapping method to generate a multi-scale feature map T = [T1, T2, ..., T m ]; S26. Store the generated lightning current multi-scale characteristic map as a standard format data file.

4. The method for monitoring lightning current of an overhead transmission tower according to claim 1, characterized in that: The S3 comprises the following steps: S31, multi-scale characteristic graph of lightning current T=[T1,T2,…,T m ] and the tower structure response data matrix R = [R1, R2, …, R n ] to perform data fusion and generate the spatiotemporal characteristic input dataset X input =[T,R]; S32, build a spatiotemporal adversarial generation network, the network includes a generator and a discriminator, the generator is input with the data set X input Learn the dynamic laws of lightning propagation and generate a lightning propagation characteristic matrix; S33, the discriminator performs comparative learning based on the input real lightning characteristic matrix and the generated characteristic matrix, and continuously optimizes the output characteristic matrix of the generator; S34, through the alternating training of the generator and the discriminator, the generated lightning propagation characteristic matrix X is finally obtained output ; S35, decompose the generated lightning propagation characteristic matrix and extract the time dimension X time and the spatial dimension X space , respectively represent the propagation time law and spatial diffusion path of lightning events; S36, recombining the characteristics of the time dimension and the space dimension to generate a lightning event time-space characteristic matrix X containing the complete time-space propagation law spatio-temporal ; S37. The generated spatiotemporal characteristic matrix is ​​used as input for subsequent dynamic modeling to provide high-confidence basic data for risk assessment and path analysis of lightning strike events.

5. The method for monitoring lightning current of an overhead transmission tower according to claim 1, characterized in that: The S4 comprises the following steps: S41, the spatiotemporal characteristic matrix X of the lightning event spatio-temporal =[X time ,X space ] in the spatial characteristic matrix X space Extract as input, combined with the initial adjacency graph of the tower network G = (V, ε), where V is the set of tower nodes, ε is the set of edges connecting tower nodes, and W = [w ij ] is the edge weight matrix, and its calculation formula is: Among them, X i and X j is the characteristic vector of tower nodes i and j, σ is the control scale factor, and deg(i) is the degree of node i; S42, constructing an initial adjacency matrix A, and performing normalization processing in combination with the spatial characteristic weight matrix W; S43. Based on the hypergeometric graph embedding algorithm, the initial adjacency matrix is ​​embedded with high-order features to generate a high-order adjacency relationship matrix. The embedding formula is: Among them, A k is a high-order embedding matrix, represents the high-order propagation operation of the adjacency matrix, D is the degree matrix, γ i is the weight coefficient embedded at each order, which is used to dynamically adjust the influence of high-order characteristics; S44, based on the high-order adjacency matrix A k , calculate the spatial incidence matrix M spatial , where the matrix element m ij represents the strength of association between nodes i and j; S45, combining the real-time environmental variable matrix E, optimizing the spatial correlation matrix through a dynamic weight adjustment model; S46. Using dynamic weight matrix W * , the path search algorithm is applied to identify the lightning strike impact range and diffusion path in the tower network; S47. Output the diffusion path results, including the tower node set within the lightning strike influence range and the correlation strength of the path.

6. The method for monitoring lightning current of an overhead transmission tower according to claim 1, characterized in that: The S5 comprises the following steps: S51, taking the spatial correlation matrix M of lightning strike events spatial =[m ij ] and the real-time environment variable matrix E = [e ij ] as input, where m ij represents the spatial correlation strength between tower nodes i and j, e ij represents the impact factor of environmental variables on the communication path between nodes i and j; S52, construct a wireless communication network topology graph G = (V, ε, W * ), where V is the set of communication nodes, ε is the set of communication paths between nodes, is the dynamic path weight matrix; S53. A Bayesian game optimization model is used to dynamically optimize the selection of communication paths and frequency allocation. The model objective function is: Among them, P is the path set, f ij is the current frequency occupancy of communication paths i and j; S54. In Bayesian game optimization, the prior probability distribution and the posterior probability distribution Iteratively update the path weights and dynamically adjust the path selection strategy; S55, using the optimized communication path set P opt , select the transmission path with the least signal interference and the largest path weight to achieve dynamic optimization of communication data; S56: Combine the optimized path set and assign the best transmission frequency to the wireless communication module in, β is the frequency adjustment factor, F is the available frequency set; S57. Upload the optimized communication path and frequency allocation plan to the remote monitoring center to complete the dynamic adjustment of the wireless communication module and provide efficient support for subsequent data transmission.

7. The method for monitoring lightning current of an overhead transmission tower according to claim 1, characterized in that: The S6 comprises the following steps: S61, receiving space correlation matrix M spatial , real-time environmental variable matrix E, tower structure response data matrix R, lightning current multi-scale characteristic map T, integrated into a joint data set D = {M spatial ,E,R,T}; S62. Design a multi-task self-supervised learning model The model includes a main task network and an auxiliary task network, which respectively process the spatial correlation and structural response of the data set, as well as the temporal evolution characteristics; S63, input R and M spatial To the main task network, generate the damaged characteristic vector S of the tower structure, perform sparse optimization on the characteristic vector, and extract the key structural response characteristics; S64, input T and E to the auxiliary task network to generate the time evolution characteristic matrix T evolve , smooth optimization of the changing trend of the time dimension; S65, the structural damage characteristic vector S and the time evolution characteristic matrix T evolve Perform joint processing to generate a comprehensive evaluation matrix E comprehensive , represents the comprehensive state of the tower and the evolution of events; S66. Upload the comprehensive assessment matrix, complete the damage analysis and evolution trend prediction of the lightning strike event, and output the structural status assessment and lightning strike risk results.

8. The method for monitoring lightning current of an overhead transmission tower according to claim 1, characterized in that: The S7 comprises the following steps: S71, receiving the evaluation result generated by the edge computing unit, and fusing it with the tower status data in the historical database to form a joint data set for evaluation; S72. Analyze the joint data set to identify the damage characteristics of the tower structure and the time evolution characteristics of the lightning event, and generate a risk assessment matrix based on historical data; S73. Based on the risk assessment matrix and current environmental variable data, design a lightning protection optimization strategy and form protection recommendations for each tower node; S74. Combine the reinforcement records in the historical database with the real-time assessment data to generate a structural reinforcement strategy matrix and provide specific reinforcement recommendations; S75. Integrate the risk assessment matrix, lightning protection strategy and structural reinforcement strategy, generate a standardized output report, upload the standardized output report to the remote monitoring terminal, and generate a standardized output report.

9. An overhead transmission tower lightning current monitoring system, characterized in that: Includes the following modules: Multimodal sensor module: installed at key locations of overhead transmission towers, the module includes a Rogowski coil current sensor and a vibration sensor, and is used to collect the amplitude, waveform, duration of lightning current signals and structural response data of the tower in real time; Signal processing unit: connected to the multimodal sensor module, configured to process the collected lightning current signal and structural response data, extract dynamic features using a local sparse dynamic window optimization algorithm, and generate a lightning current multi-scale feature map; Edge computing unit: connected to the signal processing unit, configured to jointly model the lightning current multi-scale characteristic graph and structural response data based on the spatiotemporal adversarial generation network, generate the spatiotemporal characteristic matrix of the lightning event, and construct the spatial correlation map of the lightning event through the hypergeometric map embedding algorithm; Wireless communication module: connected to the edge computing unit, dynamically adjusts the transmission path and frequency allocation based on the Bayesian game optimization model, and transmits the analysis results to the remote monitoring terminal in real time; Remote monitoring terminal: connected to the wireless communication module, configured to receive the analysis results of the edge computing unit, conduct a comprehensive assessment based on the tower status data in the historical database, generate a risk prediction report for lightning strikes, and propose suggestions for lightning protection and structural reinforcement.

Images