Power transmission line anti-nesting self-adaptive protection method and system

Abstract

The invention discloses an anti-nesting self-adaptive protection method and system for a power transmission line, and relates to the technical field of power transmission line protection. The method comprises the following steps: deploying multiple sensors at key parts of a power transmission line tower, and constructing a sensing network to detect bird activity data flow in real time; calling the nesting threat evaluation dual-channel through an edge calculation unit, carrying out the recognition evaluation of the data flow, and outputting a threat parameter; analyzing the threat parameters based on a bird nesting protection strategy library, and determining a target protection strategy; and according to the strategy parameters, a nesting blocking and expelling linkage device is controlled, and self-adaptive feedback protection is carried out on the tower. The technical problems of short circuit, tripping and other potential safety hazards caused by bird nesting to the safety of the power transmission line are solved, bird activities are accurately detected in real time through a multi-mode sensing network, nesting behavior recognition and threat degree quantification are completed in combination with edge calculation dual-channel evaluation, the risk of short circuit tripping of the line is reduced, and the safety of the power transmission line is improved. And the operation reliability and the intelligent operation and maintenance level of the power transmission line are improved.

Description

Technical Field

[0001] This invention relates to the field of power transmission line protection technology, specifically to an adaptive protection method and system for preventing nesting on power transmission lines. Background Technology

[0002] Transmission lines, as the backbone of the power system, are crucial for safe and stable operation. However, birds nesting on critical locations such as crossarms of towers and above insulator strings has become one of the main external causes of line short circuits, tripping, and even large-scale power outages. Nesting materials (such as twigs and wires) that fall or become damp can easily lead to air gap breakdown, and the birds' own activities can also cause discharges between conductors or between conductors and towers, seriously threatening the reliability of the power grid.

[0003] Currently, protective measures to address this issue mainly fall into two categories: First, passive physical barriers, such as installing bird spikes, baffles, and insulating sleeves on poles and towers, attempting to prevent birds from landing and nesting through physical barriers. These methods have significant limitations: the protection is delayed, often implemented only after bird nests have already caused hazards or malfunctions; the devices themselves may alter the electric field distribution, triggering new corona or discharge problems; and long-term exposure leads to aging and damage, requiring frequent maintenance. Second, active bird deterrence, such as installing sound, light, and electrical bird deterrents. These methods may be effective initially, but birds easily develop tolerance, and the deterrent effect diminishes over time; simultaneously, continuous and indiscriminate deterrence may disrupt the local ecological balance and generate noise and light pollution. Therefore, how to construct an active protection system capable of real-time and accurate perception, intelligent threat assessment and intervention, and achieve a shift from "static defense" to "dynamic intelligent defense" has become an urgent problem to be solved in the safety protection of transmission lines. Summary of the Invention

[0004] This application provides an adaptive protection method and system for preventing bird nesting on power transmission lines, which solves the technical problems of short circuits, tripping and other safety hazards caused by bird nesting on power transmission lines.

[0005] The first aspect of this application provides an adaptive protection method for preventing basting in transmission lines, the method comprising: Key components of transmission line towers are identified and multiple sensors are deployed to construct a multimodal sensing network. This network is used to detect multimodal bird activity data streams in real time. An edge computing unit invokes a dual-channel nesting threat assessment mechanism, consisting of a nesting behavior recognition branch channel and a threat level assessment branch channel connected in series. Based on this dual-channel assessment, the multimodal bird activity data streams are identified and evaluated, and nesting threat level parameters are output. A bird nesting protection strategy library is built, and the nesting threat level parameters are analyzed using this library to determine target nesting protection strategy parameters. A nesting blocking and repelling linkage device is acquired, and based on the target nesting protection strategy parameters, the device is controlled to perform anti-nesting feedback protection on the transmission line towers.

[0006] A second aspect of this application provides an adaptive protection system for transmission lines against caching, the system comprising: Real-time detection module: Identifies key parts of transmission line towers and deploys multiple sensors to construct a multimodal perception network. This network is used to detect real-time multimodal bird activity data streams. Threat identification module: Utilizes an edge computing unit to access a dual-channel nesting threat assessment system. This system consists of a nesting behavior identification branch channel and a threat level assessment branch channel connected in series. Based on this dual-channel system, it identifies and assesses the multimodal bird activity data streams and outputs nesting threat level parameters. Strategy parsing module: Builds a bird nesting protection strategy library. Using this library, it analyzes the nesting threat level parameters to determine target nesting protection strategy parameters. Anti-nesting module: Acquires information about nesting blocking and removal linkage devices. Based on the target nesting protection strategy parameters, it controls these devices to perform anti-nesting feedback protection on the transmission line towers.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, a multimodal sensing network is constructed by identifying and deploying various sensors on key components of transmission line towers to monitor and capture bird activity data in real time. Then, edge computing units analyze this data to assess the threat level of bird nesting behavior and output nesting threat parameters. Next, based on these parameters, a bird nesting protection strategy library is consulted to analyze the optimal protection strategy and determine the corresponding protection parameters. Finally, based on these protection strategy parameters, blocking and repelling equipment is controlled to implement feedback protection measures against bird nesting, thereby ensuring the safety of the transmission lines. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a schematic diagram of the adaptive protection method for preventing nesting in transmission lines provided in the embodiments of this application.

[0010] Figure 2 This is a schematic diagram of the structure of the adaptive protection system for preventing nesting in power transmission lines provided in an embodiment of this application.

[0011] Figure labeling: Real-time detection module 11, Threat identification module 12, Strategy analysis module 13, Anti-nesting module 14. Detailed Implementation

[0012] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0013] Example 1, as Figure 1 As shown, this application provides an adaptive protection method for preventing burying in transmission lines, wherein the method includes: Key parts of transmission line towers are identified and multiple sensors are deployed to construct a multimodal sensing network. The multimodal sensing network is then used to detect real-time multimodal bird activity data streams.

[0014] In this embodiment, to effectively monitor and prevent birds from nesting on power transmission line towers, the key parts of the towers are first identified and multiple sensors are deployed. This process involves scanning the towers using 3D laser scanning or point cloud data acquisition to obtain spatial point cloud data. A 3D model of the tower is then constructed using denoising and 3D reconstruction techniques. Based on this 3D model and the behavioral characteristics of bird nesting, potential key nesting sites on the towers are identified, such as beams, supports, and conductor junctions—areas where birds are likely to nest. After identifying these key sites, suitable sensors are selected based on the tower's structural characteristics, such as high-definition cameras, infrared sensors, temperature and humidity sensors, acoustic sensors, and vibration sensors. These sensors are then deployed across the towers to construct a multimodal sensing network. This network connects all sensors wirelessly, transmitting information streams collected from different sensors in real time. These information streams are aggregated to form a multimodal bird activity data stream, providing data support for subsequent behavior recognition and threat assessment. This enables intelligent and real-time response in protection efforts, effectively ensuring the safe operation of the power transmission lines.

[0015] Taking the multi-modal bird activity data stream of a key part of a pole within the same time window as an example, at timestamp T=2025-05-18 09:32:15, a high-definition camera located at the crossarm of the pole captured a frame of image data. Its metadata includes: image resolution of 1920×1080, frame number Frame_102345, target detection results showing the presence of one bird target, target bounding box coordinates of (x1=420, y1=310, x2=610, y2=520), target category of "bird", and confidence level of 0.93; at the same time, the infrared sensor collected an infrared radiation intensity value of 37.8℃ in the corresponding area, which is 3.2℃ higher than the background ambient temperature, indicating the presence of birds. During the activity of birds near biological heat sources, vibration sensors installed at the same key location detected micro-vibration signals with an acceleration amplitude of 0.018g and vibration frequencies concentrated in the 8-12Hz range, consistent with the vibration characteristics of birds perching or preparing nesting materials. Acoustic sensors simultaneously acquired environmental sound signals, and their spectral analysis showed a continuous bird call characteristic frequency band in the 2.5kHz-4.2kHz range, with a sound pressure level of 48dB. Additionally, environmental sensors recorded temperature 29.4℃, relative humidity 61%, and wind speed 2.1m / s. The data from visual, infrared, vibration, acoustic, and environmental sensors were aligned with a unified timestamp and combined in a structured form into a multimodal bird activity data stream: {timestamp, image feature vector, target detection result, infrared thermal feature value, vibration amplitude and frequency characteristics, acoustic spectral characteristics, environmental parameters}. This multimodal activity data stream was transmitted as input to the edge computing unit for subsequent bird nesting behavior identification and threat level assessment.

[0016] Furthermore, constructing a multimodal sensing network includes: The process involves scanning to acquire a spatial point cloud dataset of the transmission line towers, performing denoising preprocessing and 3D reconstruction based on the dataset to generate a 3D point cloud model of the towers. Key components of the 3D point cloud model are identified based on bird nesting behavior characteristics, yielding information on N key tower components. Monitoring requirements are analyzed for these N key components, and sensor selection and deployment coverage analysis are performed based on the monitoring requirements to determine multi-sensor model specifications and deployment strategy parameters. Finally, multi-sensor deployment is implemented on the transmission line towers based on these parameters to construct a multimodal sensing network.

[0017] Preferably, a high-precision scan of the transmission line towers is first performed using a 3D laser scanner, such as a LiDAR device, to accurately measure the position of each point on the tower and generate a spatial point cloud dataset. This dataset provides a digital representation of the detailed spatial structure of the tower, typically including the tower's height, crossbeams, supports, conductors, and other components' precise 3D coordinates. The obtained spatial point cloud data usually contains noise points, such as erroneous data caused by abnormal reflected signals or environmental interference. Therefore, before 3D reconstruction, these invalid points need to be removed using Gaussian filtering. After preprocessing, based on the normal vector information of each point in the point cloud data, an implicit function containing the point cloud surface is constructed by solving the Poisson equation. This implicit function is then used for 3D reconstruction to generate a 3D point cloud model of the tower. This 3D point cloud model of the tower will present the overall geometric shape of the tower and can display its detailed features, including the structure of the supports, the arrangement of the crossbeams, and the layout of the conductors. Based on a 3D point cloud model, the system identifies high-frequency areas by combining the characteristics of bird nesting behavior with the frequency of transmission line faults. It then filters these high-frequency areas to identify potential nesting sites, such as the intersection of beams and supports, conductor supports, and other protruding or platform-like components. Next, based on the identified key tower components, a monitoring requirements analysis is performed. This involves matching available sensors to a sensor database based on the needs of each key component, and calculating the suitability of each sensor for different locations according to its technical specifications, such as detection range, sensitivity, and resolution. For example, high-definition cameras are selected for areas requiring high-resolution image monitoring; infrared sensors are selected for nighttime monitoring; and vibration sensors are selected for detecting tower vibration.

[0018] After sensor selection, the system reads the model specifications of the selected sensors, including detection radius, field of view, detection direction, resolution threshold, and effective sensing distance, and constructs the corresponding theoretical sensing cone in 3D space. Next, several candidate installation points are generated on the surface of the tower's 3D point cloud model. For each candidate installation point, a virtual sensor deployment is performed, and spatial geometric calculations are used to determine the intersection relationship between the theoretical sensing cone and spatial targets at key locations, calculating whether the key locations fall within the effective sensing range. Based on this, an occlusion detection algorithm is introduced, and ray tracing analysis of the sensor's line of sight is performed in conjunction with the tower's own structure to eliminate coverage results that are blocked by tower components and cannot be effectively sensed. Then, for each key location, the number of times it is covered by different sensors, the coverage angle, and the coverage redundancy are calculated to form a coverage matrix. The system optimizes the coverage matrix according to preset coverage constraints, such as each key location being effectively covered by at least one or more sensors and the coverage overlap rate not exceeding a set threshold, selecting sensor combinations that meet coverage integrity requirements and their corresponding installation locations. Finally, the system outputs the multi-sensor deployment coverage results and generates multi-sensor deployment strategy parameters to guide actual sensor installation and network construction.

[0019] Finally, the sensors are deployed according to their specifications and deployment strategies. For example, high-definition cameras are installed in locations with wide fields of view and easy monitoring to capture images of birds nesting or roosting. Infrared sensors are installed on high-frequency sections of poles, especially in key areas where nesting may occur, to ensure monitoring capabilities at night or in low-visibility conditions. All sensors are connected to the system via a wireless communication network for real-time data transmission. Once deployed, all sensors together form a multimodal sensing network. This network can acquire and transmit various types of data from different sensors in real time, such as image data, temperature and humidity data, vibration signals, and sound signals, providing comprehensive data support for subsequent bird behavior analysis, threat assessment, and protection decisions.

[0020] Furthermore, information on N key components of the towers is obtained, including: DBSCAN was used to perform point cloud clustering analysis on the three-dimensional point cloud model of the tower to obtain the three-dimensional point cloud clustering analysis results. Based on the three-dimensional point cloud clustering analysis results, structural features were extracted from the three-dimensional point cloud model of the tower to obtain the structural feature set of the tower area. High-frequency areas of the structural feature set of the tower area were identified according to the fault frequency of the transmission line to obtain information on multiple high-frequency parts of the tower. Based on the nesting behavior characteristics of birds, key parts of the multiple high-frequency parts of the tower were screened to obtain information on multiple key parts of the tower.

[0021] Optionally, the point cloud in the 3D point cloud model of the tower is first normalized to unify it to a fixed spatial scale and coordinate system, thus eliminating calculation biases caused by different tower sizes. Then, the DBSCAN density clustering algorithm is invoked to perform cluster analysis on the point cloud data. Specifically, the number of neighboring points within a given neighborhood radius ε for each point is calculated and compared with the minimum density threshold MinPts. Points meeting the density condition are identified as core points, and clustering regions are gradually expanded outwards. Points with insufficient spatial density and unable to belong to any cluster are marked as noise and removed. This process automatically divides the tower point cloud into multiple geometrically consistent clustering regions in spatial structure, thus obtaining the 3D point cloud clustering analysis results. After obtaining the clustering results, the system performs structural feature extraction using each point cloud clustering region as a basic unit. That is, it calculates the geometric and spatial feature parameters of each clustering region, including but not limited to the 3D bounding box size, height distribution range, principal direction of the point cloud normal vector, surface curvature statistical characteristics, and spatial connectivity with adjacent clustering regions. Based on the above parameters, the system determines the structure type of clustered regions through rule matching. For example, when the angle between the main axis direction and the gravity direction of a clustered region is less than a preset threshold α, and its height dimension is significantly greater than its horizontal dimension and its aspect ratio is greater than a first ratio threshold, the clustered region is determined to be a column-type structure. When the angle between the main axis direction and the horizontal direction of a clustered region is less than a preset threshold β, and its length dimension is greater than its height dimension and it has a stable connection relationship with a column-type structure, the clustered region is determined to be a crossarm-type structure. When the geometric shape of a clustered region exhibits an inclined distribution, its main axis direction does not meet the determination conditions for either column-type or crossarm-type, and its spatial position is located between a column and a crossarm, exhibiting obvious connection or reinforcement characteristics, it is determined to be a support connection structure. When a clustered region is located high on the periphery of a tower, its point cloud distribution exhibits a slender linear characteristic, and its spatial height is consistent with the end of the crossarm or the connection position of the insulator, the clustered region is determined to be a conductor connection structure. Through the aforementioned rule matching, the system can automatically complete the refined identification and labeling of tower structure types without manual intervention, thereby forming a corresponding tower area structural feature set. Then, historical fault data of transmission lines is introduced, and the coordinates of fault locations in the historical fault records are mapped onto the 3D point cloud model of the tower. Spatial matching is used to assign fault points to the corresponding structural feature regions. The number of historical fault occurrences for each type of structural region is counted, and the regional fault frequency is calculated by dividing this number by the total number of faults. When the fault frequency of a certain region exceeds a preset threshold, that region is automatically marked as a high-frequency location on the tower, thus obtaining information on multiple high-frequency locations on the tower.Then, combining the characteristics of bird nesting behavior, such as structural areas located at higher positions, structurally stable areas with small vibration amplitude and certain support conditions, and structural areas with certain shielding conditions, the information of high-frequency parts of the tower is screened for key parts to obtain target areas that have both high failure risk and conform to the characteristics of bird nesting behavior. These areas are identified as key parts of the tower for subsequent key monitoring and protection deployment.

[0022] The nesting threat assessment dual channel is invoked through the edge computing unit. The nesting threat assessment dual channel is composed of a nesting behavior recognition branch channel and a threat level assessment branch channel connected in series. Based on the nesting threat assessment dual channel, the multi-modal activity data stream of birds is identified and assessed, and nesting threat level parameters are output.

[0023] In one embodiment, the edge computing unit first receives real-time transmitted multimodal bird activity data streams from the multimodal sensing network and performs time alignment and formatting on the data stream, uniformly encapsulating data from different sensors such as vision, infrared, vibration, acoustics, and environment into standardized input samples. Subsequently, the edge computing unit invokes the dual-channel model for nesting threat assessment and inputs the multimodal bird activity data streams into the nesting behavior recognition branch channel according to a preset data flow order. The nesting behavior recognition branch channel performs feature extraction and fusion analysis on the input data stream, outputting dynamic fault data and behavioral confidence information of nesting behavior within the corresponding time window to characterize whether birds are currently exhibiting nesting-related behavioral characteristics such as nest building, twig gathering, or accumulating. After completing the nesting behavior recognition, the edge computing unit uses the output of the nesting behavior recognition branch channel as input to further transmit it to the threat level assessment branch channel. The threat level assessment branch channel performs a comprehensive threat assessment calculation on the identified nesting behavior based on the bird nesting threat assessment system, generates continuous numerical threat assessment results corresponding to the safety risks of transmission lines, and outputs them as nesting threat level parameters. This is used to characterize the real-time risk level posed by current bird activities to the operational safety of transmission lines, and provides a basis for subsequent protection strategy analysis and nesting prevention linkage control.

[0024] Furthermore, the dual-channel approach to nesting threat assessment is invoked through edge computing units, including: Historical datasets of nesting faults on power transmission lines are collected and standardized to obtain a dataset of available nesting faults on power transmission lines. The available datasets are then subjected to structured classification and behavior recognition training to generate a nesting behavior recognition branch channel. Threat assessment training is performed on the dynamic fault dataset of nesting behavior output by the nesting behavior recognition branch channel to construct a threat level assessment branch channel. The nesting behavior recognition branch channel and the nesting behavior recognition branch channel are then connected in series to form a dual-channel nesting threat assessment system, which is then stored in the edge computing unit.

[0025] Optionally, firstly, multi-source fault data related to bird nesting are collected from the transmission line operation and maintenance platform, defect management system, online monitoring system, and historical inspection data to form a historical dataset of transmission line nesting faults. This historical dataset of transmission line nesting faults should at least include fields such as fault occurrence time, tower number, fault phase, current / voltage change, tripping type, video file path, acoustic raw waveform data, vibration raw time sequence data, and infrared temperature data. Then, these fields are standardized, including time alignment according to timestamps, mapping spatial location descriptions from different sources to tower three-dimensional coordinates, rule completion for missing fields, statistical threshold removal for outliers, and normalization of data from different ranges, thereby obtaining a usable dataset of transmission line nesting faults. Subsequently, structured classification and behavior recognition training are performed on the available transmission line nesting fault dataset. Specifically, the data is automatically divided into a visual transmission line nesting fault dataset and a non-visual transmission line nesting fault dataset based on data modal attributes. Visual nesting behavior recognition branches and non-visual nesting behavior recognition branches are then trained based on these two branches. A nesting behavior recognition branch channel is constructed by weighted fusion of the two branches. After training the nesting behavior recognition branch channel, the inference results of this branch channel on training samples are correlated with the corresponding historical fault consequences, automatically generating a dynamic nesting behavior fault dataset. This dynamic dataset includes bird species, behavior types, behavior confidence, occurrence location, duration, repetition frequency, and nesting material characteristics. Based on this dynamic fault dataset, the system jointly models the relevant behavioral features with a bird nesting threat assessment system. Through threat assessment training, validation, and optimization, the system learns the mapping relationship between behavioral features and line risks, constructing a stable and usable threat level assessment branch channel. Finally, following a fixed data interface specification, the nesting behavior recognition branch channel and the threat level assessment branch channel are serially encapsulated, and a unified data input structure is defined. This allows multimodal features to first enter the nesting behavior recognition branch channel, and its output dynamic fault data of nesting behavior is directly transmitted to the threat level assessment branch channel, forming an end-to-end dual-channel model for nesting threat assessment. This dual-channel model, after lightweight processing, is encapsulated into an edge deployment version and stored in the edge computing unit. During operation, the edge computing unit can continuously receive multimodal bird activity data streams, complete behavior recognition and threat assessment in the dual-channel sequence, and output nesting threat level parameters in real time, providing accurate data for subsequent protection strategy analysis and coordinated control.

[0026] Furthermore, the generation of nest-building behavior recognition branch channels includes: The available power transmission line nesting fault dataset is structured and classified to obtain a visual power transmission line nesting fault dataset and a non-visual power transmission line nesting fault dataset. Bird activity behavior is labeled on both datasets to obtain a visual power transmission line nesting behavior sample set and a non-visual power transmission line nesting behavior sample set. Behavior recognition training is performed based on these samples to obtain a visual nesting behavior recognition branch and a non-visual nesting behavior recognition branch. The visual and non-visual nesting behavior recognition branches are then weighted, fused, validated, and optimized to generate a nesting behavior recognition branch channel.

[0027] Optionally, the available transmission line nesting fault dataset is first read line by line, and modality is determined based on the data field types and file formats contained in the read data. When a sample contains image file paths or video frame sequence fields, the sample is automatically labeled as a visual sample; when a sample only contains acoustic time-series data, vibration time-series data, infrared temperature, etc., the sample is labeled as a non-visual sample. After classification, visual transmission line nesting fault datasets and non-visual transmission line nesting fault datasets are constructed separately, and independent data indexes, timestamp indexes, and tower spatial location indexes are created for each dataset to support subsequent behavior annotation and training calls. After completing the dataset partitioning, bird activity behavior annotation processing is performed on the visual transmission line nesting fault dataset and the non-visual transmission line nesting fault dataset respectively. For visual data, the system uses video timelines or image timestamps as units, associating each time slice with corresponding fault records and on-site descriptions to automatically generate a set of candidate behavior labels. It also annotates the bounding box position, occurrence time, duration (frames), and behavior category label of bird targets in image or video frames, thus forming a visual transmission line nesting behavior sample set containing {image / video features, spatial location, behavior label, and temporal parameters}. For non-visual data, the system extracts feature segments from acoustic, vibration, and infrared time-series signals using a sliding time window, matching each feature segment with fault labels and behavior descriptions within the corresponding time period to generate a non-visual transmission line nesting behavior sample set containing {temporal feature vector, time window, behavior label, and frequency of occurrence}. Through this process, all samples possess clear input features and standardized behavior labels. Subsequently, behavior recognition training is performed based on both the visual and non-visual transmission line nesting behavior sample sets.

[0028] For the visual nesting behavior sample set of power transmission lines, the system selects a CNN-LSTM-based temporal behavior recognition model as the visual nesting behavior recognition branch. This model is used to simultaneously learn the spatial features of bird appearance and the temporal evolution features of nesting behavior. Before training, the original image or video data is uniformly scaled, such as to 224×224, and the video is sampled frame by frame at fixed time intervals to form a continuous image sequence of length L, such as 16 frames. At the same time, the bird target region is cropped for each frame to reduce background interference. In terms of model structure, the visual nesting behavior recognition branch includes a convolutional feature extraction layer, a temporal modeling layer, and a fully connected classification output layer. Among them, the convolutional feature extraction layer adopts a convolutional neural network with a preset number of layers, for example, 4 convolutional units. Each convolutional unit contains a convolutional layer, a batch normalization layer, and a ReLU activation layer, which is used to extract bird appearance features and local action features from a single frame image. Subsequently, the convolutional features of L consecutive frames are input sequentially into the LSTM temporal modeling layer. This layer learns the dynamic patterns of bird movements over time, such as repeated landings, twig carrying, and nest material accumulation. The temporal feature vectors output from the LSTM layer are further input into the fully connected classification output layer, which outputs probability vectors corresponding to various nest-building behaviors. During training, visual sample sequences from the visual transmission line nest-building behavior sample set and their corresponding nest-building behavior labels are input into the model for forward propagation. Behavior classification loss functions, such as cross-entropy loss, are calculated. The parameters of the convolutional, LSTM, and fully connected layers are updated using a backpropagation algorithm. An adaptive optimizer, such as Adam, is then used for parameter updates. The recognition accuracy and loss changes are monitored on the validation set. Training ends when the validation performance converges or reaches a preset threshold, resulting in a visual nest-building behavior recognition branch. This branch can output nest-building behavior features, nest-building behavior category probabilities, and corresponding confidence scores for the input visual data.

[0029] For the non-visual brood-building behavior sample set of transmission lines, the system selects a behavior recognition model based on a multi-channel temporal convolutional network as the non-visual brood-building behavior recognition branch to jointly model multi-source temporal signals such as acoustics, vibration, and infrared. Before training, the original acoustic data, vibration data, and infrared temperature data are divided into uniform time windows, such as 1 second or 2 seconds, and the corresponding time-frequency feature representations are extracted, such as acoustic Mel-frequency spectrum, vibration time-frequency spectrum, and infrared temperature difference change sequence. Subsequently, the different modal features are normalized and concatenated along the feature dimension to form a multi-channel temporal feature input. In terms of model structure, the non-visual nesting behavior recognition branch includes a multi-channel temporal convolutional layer, a feature fusion layer, and a fully connected classification output layer. The multi-channel temporal convolutional layer uses one-dimensional or two-dimensional convolutional kernels for different modal features to extract local variation patterns in various temporal signals, such as bird wing flapping frequency, continuous call features, or infrared heat source variation trends. The feature fusion layer concatenates and compresses the outputs of each channel's convolution to form a unified high-dimensional behavioral feature representation. The fully connected classification output layer outputs probability vectors corresponding to various nesting behaviors based on the obtained feature vectors. During training, the system inputs non-visual temporal feature samples and corresponding behavior labels into the model, performs forward propagation to calculate the classification loss, and updates the convolutional kernel weights and fully connected layer parameters through backpropagation. The recognition accuracy of each behavior category is monitored synchronously during training, and the model is optimized by adjusting the convolutional kernel size, number of channels, or learning rate. Once the validation set performance is stable, training is complete, resulting in the non-visual nesting behavior recognition branch, whose output also includes nesting behavior features, nesting behavior category probabilities, and corresponding confidence scores.

[0030] After obtaining the visual nesting behavior recognition branch and the non-visual nesting behavior recognition branch, the system performs weighted fusion and validation optimization on the two types of branch models. Specifically, initial weight coefficients are assigned to the output results of the visual nesting behavior recognition branch and the non-visual nesting behavior recognition branch, respectively. The output results of the two branches are then weighted and summed within the same time window to generate a fusion result. Subsequently, the system uses validation samples to evaluate the performance of the fusion result, iteratively adjusting the weight coefficients based on changes in recognition accuracy, false alarm rate, and false negative rate, and adaptively calibrating the behavior judgment threshold. When the fusion model reaches the preset performance indicators on the validation set, the optimized visual nesting behavior recognition branch, the non-visual nesting behavior recognition branch, and their fusion parameters are encapsulated into a unified inference structure, ultimately generating a nesting behavior recognition branch channel. This branch channel can output stable and consistent nesting behavior recognition results for the input multimodal data stream, providing standardized input for subsequent threat level assessment branches. This reduces the probability of false alarms and false negatives caused by relying solely on visual or single non-visual information, improving the decision-making accuracy and operational reliability of the overall transmission line nesting prevention adaptive protection.

[0031] Furthermore, a threat level assessment branch channel is constructed, including: Based on the safety application standards for power transmission lines, a bird nesting threat assessment system is established. This system includes bird species, nesting location, nest material type, size, and nesting behavior dimensions. The threat level of the dynamic fault dataset of nesting behavior output by the nesting behavior identification branch channel is assessed and labeled according to the bird nesting threat assessment system, resulting in a power transmission line nesting fault threat sample set. Threat assessment training, verification, and optimization are then performed based on this sample set to construct a threat level assessment branch channel.

[0032] Optionally, the system first constructs a bird nesting threat assessment system based on the power transmission line safety application standards. This system defines nesting threat input parameters in the form of multi-dimensional feature vectors. These multi-dimensional feature vectors include dimensions such as bird species, nesting location, nest material type, size, and nesting behavior. Specifically, the bird species dimension is quantified by the identified category; the nesting location dimension is quantified by the minimum spatial distance and relative height difference between the nest and conductors, insulators, and fittings; the nest material type dimension is quantified by the conductivity and hygroscopicity levels of the nest material; the size dimension is quantified by the nest volume and coverage area; and the nesting behavior dimension is quantified by the behavior type, frequency, and duration. After completing the threat assessment system, the system performs threat level assessment and labeling on the dynamic fault dataset of nesting behavior output from the nesting behavior identification branch channel according to the bird nesting threat assessment system. Specifically, the system reads dynamic fault data in time windows and automatically maps the bird behavior identification results, spatial location information, nest material feature indicators, and behavioral statistical parameters included in the nesting behavior features to the corresponding threat feature vector fields. Subsequently, the features of each dimension are weighted and summed according to preset weights to obtain the corresponding raw threat score. This score is then labeled with different threat levels based on preset threshold ranges. The calculated threat scores, threat level labels, and dynamic fault data related to nesting behavior are then associated and stored, thus constructing a transmission line nesting fault threat sample set containing "dynamic fault data - threat score - threat level label". After obtaining the transmission line nesting fault threat sample set, the system performs a threat assessment training and validation optimization process based on this sample set to construct a threat level assessment branch channel. During the training phase, the multi-dimensional threat feature vectors in the threat sample set are used as model input, and the corresponding threat level or risk score is used as supervised output. A supervised learning method is used to train a threat assessment model based on a regression model, enabling the model to learn the mapping relationship between different combinations of nesting features and transmission line safety risks. During model training, the system evaluates the model's discrimination accuracy at different threat levels through cross-validation and iteratively optimizes the feature weights and model parameters based on the validation results. After multiple rounds of training, verification, and optimization, a stable and convergent threat level assessment branch channel was finally obtained. This branch channel can receive nesting behavior recognition results during the operation phase and output nesting threat level parameters that match the current nesting status in real time, thereby providing accurate and quantitative risk assessment basis for protection strategy analysis and anti-nesting linkage control.

[0033] Furthermore, the method also includes: A dataset of the operating environment of the transmission line towers is collected, which includes meteorological conditions, geographical environment data, and equipment status data. The threat level of the operating environment dataset is assessed to determine the operating environment threat gain coefficient, and the nesting threat level parameter is adjusted by gain addition based on the operating environment threat gain coefficient.

[0034] Optionally, during the operation phase, the system periodically collects operational environment datasets for transmission line towers and categorizes them into three types based on their source: meteorological data, geographical environment data, and equipment status data. Meteorological data includes wind speed, instantaneous maximum wind speed, rainfall, relative humidity, and ambient temperature; geographical environment data includes tower elevation, terrain type code, surrounding vegetation coverage index, and water body distance parameters; and equipment status data includes insulator pollution level, conductor sag deviation, hardware aging coefficient, cumulative number of historical defects, and current operating load rate. The system performs timestamp unification, missing value imputation, and normalization on these data, mapping data from different dimensions to fixed-length operational environment features. After obtaining standardized operational environment features, the system performs environmental threat assessment calculations on the operational environment dataset. Specifically, it compares each operational environment feature in the dataset with its corresponding benchmark interval, determines the direction of deviation, and calculates the degree of deviation from the maximum or minimum value using the absolute difference. For features without deviation, the deviation degree is set to 0. Subsequently, the deviation of each quantified operating environment feature is weighted and integrated according to the preset influence weight of each operating environment feature to obtain a comprehensive environmental risk score. Based on the preset normalization function, the score is converted into an operating environment threat gain coefficient. The lower the environmental risk, the smaller the gain coefficient; the higher the environmental risk, the larger the gain coefficient. After determining the environmental threat gain coefficient, the system applies a gain-added correction to the nesting threat level parameter based on this coefficient. Specifically, the original nesting threat level parameter output from the dual-channel nesting threat assessment is matched with the environmental threat gain coefficient within the corresponding time window. The threat score corresponding to the original nesting threat level parameter is multiplied by the gain coefficient and then added to the original nesting threat level parameter's threat score value to obtain the environmentally corrected threat score value. This corrected threat score value is then mapped to the nesting threat level parameter using the same method described above. The corrected nesting threat level parameter can simultaneously reflect the danger level of bird nesting behavior itself and the dynamic impact of the external operating environment on this danger level. This allows the system to automatically improve the risk assessment results under severe weather, complex geographical conditions, or equipment degradation, thereby providing a more accurate decision-making basis for subsequent protection strategy analysis and nesting blocking and repulsion linkage control.

[0035] A bird nesting protection strategy database is established, and the nesting threat level parameters are analyzed using the database to determine the target nesting protection strategy parameters.

[0036] In one embodiment, a bird nesting protection strategy library is first built in the background, and the protection strategies are modeled and stored in a structured manner. This bird nesting protection strategy library uses "threat level - protection target - execution method - parameter constraints" as the core index dimension, and includes at least multiple types of protection strategy entries, such as passive physical barrier strategies, gentle removal strategies, combined linkage strategies, and monitoring enhancement strategies. Each protection strategy is pre-configured with corresponding applicable threat level, applicable pole / tower type, applicable environmental condition constraints, and an adjustable set of execution parameters. The adjustable set of execution parameters includes removal intensity, action frequency, duration, trigger threshold, and linkage priority. During the strategy parsing phase, the system receives the nesting threat level parameter after dual-channel output of the nesting threat assessment and environmental gain correction, and uses this parameter as the input condition for strategy parsing, filtering out matching nesting protection strategy combinations that meet the current threat level from the strategy library. Subsequently, the matching nest-building protection strategy combination is adjusted and optimized according to the nest-building protection constraints. The final protection strategy type and its corresponding execution parameter combination are parsed to obtain the target nest-building protection strategy parameters. These target nest-building protection strategy parameters are then output to the subsequent nest-building blocking and removal linkage device to guide the execution of specific nest-building protection actions. This achieves the adaptability and pertinence of nest-building protection measures, avoids excessive intervention or insufficient protection, and improves the safety and stability of transmission line operation.

[0037] Furthermore, determining the target nesting defense strategy parameters includes: The nesting threat level parameter is matched with the bird nesting protection strategy library to obtain a matching nesting protection strategy combination; the matching nesting protection strategy combination is then adjusted and optimized according to the nesting protection constraints to determine the target nesting protection strategy parameters.

[0038] Preferably, the nesting threat level parameter, after threat assessment and environmental gain correction, is first used as the core input for strategy matching. Protection strategy matching is performed according to a preset strategy index structure. First, numerical matching is performed between the nesting threat level parameter and the threat level in the strategy entries to filter out all protection strategies that meet the current threat level. Then, considering the type of key tower component where nesting has occurred, a secondary conditional matching is performed on the filtered results to eliminate strategy entries that are inapplicable or unexecutable under the current operating conditions. This results in a matched nesting protection strategy combination, represented as a parameter set containing multiple candidate protection actions and their corresponding initial execution parameters. After obtaining the matched nesting protection strategy combination, the system applies the current nesting protection constraints to the combination for constraint adjustment and optimization. Specifically, the system substitutes the initial parameters of each candidate protection strategy into the nesting protection constraints, verifying one by one whether they meet the preset constraints. For example, it verifies whether the action amplitude, speed, and range of the protection device are within the allowable range of the current transmission line equipment through protection action safety constraints; it limits the sound intensity, light intensity, or action frequency of repulsion strategies to not exceed the safety threshold of the current bird species through ecological friendliness constraints; and it verifies whether the current remaining power, mechanical travel, and execution count of the protection device support the implementation of the strategy parameters through equipment capacity constraints. For strategy parameters that do not meet the constraints, the system corrects them by parameter trimming, range compression, or execution order adjustment; if the constraints are still not met after correction, the corresponding strategy is removed from the strategy combination. After completing the constraint adjustment and optimization, the system determines the protection strategies that meet the constraints and their corresponding execution parameters as the target nesting protection strategy parameters. Through the above specific matching, constraint, and optimization process, the selection of protection strategies and parameter configuration can dynamically change with the degree of nesting threat and on-site conditions, realizing the transformation of nesting protection from fixed strategies to adaptive and refined control, thereby improving the safety, effectiveness, and intelligence level of transmission line nesting protection.

[0039] The nest-building blocking and repelling linkage device is obtained, and the nest-building blocking and repelling linkage device is controlled to perform anti-nesting feedback protection on the transmission line tower based on the target nest-building protection strategy parameters.

[0040] In one embodiment, the system first acquires and registers a nesting prevention and expulsion linkage device installed on the transmission line tower. This device includes a physical blocking component, an expulsion component, or a combination of both. The system establishes a control connection with the device via wired or wireless communication and performs device status detection to ensure the device is in an executable state. After acquiring the device and confirming its status, the system receives the determined target nesting protection strategy parameters and performs instruction parsing and execution mapping on these parameters. That is, it maps parameters such as the protection method type, action intensity, execution frequency, duration, and linkage sequence in the target nesting protection strategy parameters into a set of control instructions that the nesting prevention and expulsion linkage device can recognize, and generates an anti-nesting control instruction sequence according to the execution logic defined in the strategy. Subsequently, the system issues this anti-nesting control instruction sequence to the nesting prevention and expulsion linkage device, and the control device performs anti-nesting actions at the corresponding key parts of the tower, including but not limited to activating or adjusting the physical blocking state, triggering the expulsion action, adjusting the expulsion intensity, or performing intermittent execution at a preset rhythm. During the execution of anti-nesting actions, the system simultaneously acquires operational feedback information from the nesting blocking and deterrence linkage device and dynamically evaluates the effectiveness of the protection, generating corresponding nesting protection effect parameters. If the protection effect is not as expected or new nesting risks are detected, the system can adjust the target nesting protection strategy parameters in real time based on the feedback information and reissue updated control commands, achieving closed-loop feedback control of the anti-nesting actions. Through the above execution and feedback process, the nesting blocking and deterrence linkage device can adaptively execute anti-nesting protection actions according to different nesting threat levels and protection strategy parameters, and dynamically adjust according to the actual effect during operation, thereby improving the timeliness, effectiveness, and reliability of anti-nesting protection, avoiding risks such as tripping and short circuits, and ensuring the safe and stable operation of transmission lines.

[0041] Furthermore, controlling the nest-building blocking and dispersing linkage device to perform anti-nesting feedback protection on the transmission line tower based on the target nest-building protection strategy parameters includes: According to the target nesting protection strategy parameters, the nesting blocking and driving linkage device is controlled to perform nesting prevention monitoring and feedback on the transmission line tower to obtain nesting protection effect parameters; based on the nesting protection effect parameters, the target nesting protection strategy parameters are optimized and updated, and nesting prevention optimization protection is performed through the updated target nesting protection strategy parameters.

[0042] Preferably, the nest-building blocking and deterrence linkage device is first controlled in real time according to the target nest-building protection strategy parameters, so that it performs anti-nesting protection actions at key parts of the corresponding transmission line towers. Simultaneously, a multimodal sensing network monitors the protection process, acquiring data such as changes in bird activity intensity before and after the protection actions, whether nest-building behavior is interrupted or disappears, the addition or reduction of nesting materials, and changes in bird dwell time. This monitoring data is aligned according to a unified time window, and the monitoring results before and after protection are compared. The difference is used to calculate nest-building protection effectiveness parameters to quantify the protection's effectiveness. After obtaining the nest-building protection effectiveness parameters, the system compares them with a preset protection effectiveness evaluation threshold. When the protection effectiveness parameters are lower than the target threshold or nest-building behavior occurs repeatedly, it is determined that the current protection strategy has room for optimization. At this time, the target nest-building protection strategy parameters are adaptively updated based on the changing trend of the protection effectiveness parameters, for example, adjusting the intensity, frequency, or duration of the deterrence actions, or changing the opening status of physical barriers. When the protection effect parameters meet or exceed the target threshold, the system determines that the current protection strategy parameters are valid and can maintain the existing configuration. After completing the strategy parameter update, the system redistributes the updated target nesting protection strategy parameters to the nesting blocking and expulsion linkage device, controlling it to continue executing anti-nesting protection actions according to the optimized strategy parameters. Through the above closed-loop feedback process, the anti-nesting protection strategy can be continuously optimized based on the actual protection effect, gradually approaching the optimal protection configuration. This ensures the effectiveness of protection while avoiding excessive expulsion or unnecessary frequent device actions, improving the long-term stability, adaptability, and operational reliability of transmission line anti-nesting protection.

[0043] In summary, the embodiments of this application have at least the following technical effects: First, key components of the transmission line towers are identified and multiple sensors are deployed to construct a multimodal sensing network. This network is used to detect real-time multimodal bird activity data streams. Next, an edge computing unit invokes a dual-channel nesting threat assessment mechanism. This dual-channel consists of a nesting behavior recognition branch and a threat level assessment branch, connected in series. Based on this dual-channel assessment, the multimodal bird activity data stream is identified and evaluated, outputting nesting threat level parameters. Then, a bird nesting protection strategy library is built. This library is used to analyze the nesting threat level parameters and determine the target nesting protection strategy parameters. Finally, a nesting blocking and repelling linkage device is acquired. Based on the target nesting protection strategy parameters, this device is controlled to perform anti-nesting feedback protection on the transmission line towers. This solution addresses the technical challenges posed by bird nesting to power transmission lines, including short circuits and tripping. It achieves real-time and accurate detection of bird activity through a multimodal sensing network, combined with edge computing dual-channel assessment to identify nesting behavior and quantify the threat level, thereby reducing the risk of short circuits and tripping, and improving the reliability and intelligent operation and maintenance of power transmission lines.

[0044] Example 2, based on the same inventive concept as the transmission line anti-nesting adaptive protection method in the aforementioned examples, such as... Figure 2 As shown, this application provides an adaptive protection system for transmission lines against burying, wherein the system includes: Real-time detection module 11: Identifies key parts of transmission line towers and deploys multiple sensors to construct a multimodal perception network. This network is used to detect multimodal bird activity data streams in real time. Threat identification module 12: Utilizes an edge computing unit to call a dual-channel nesting threat assessment mechanism. This mechanism consists of a nesting behavior identification branch channel and a threat level assessment branch channel connected in series. Based on this dual-channel assessment, it identifies and assesses the multimodal bird activity data streams and outputs nesting threat level parameters. Strategy analysis module 13: Builds a bird nesting protection strategy library. This library is used to analyze the nesting threat level parameters and determine the target nesting protection strategy parameters. Anti-nesting module 14: Acquires nesting blocking and removal linkage devices. Based on the target nesting protection strategy parameters, it controls these devices to perform anti-nesting feedback protection on the transmission line towers.

[0045] Furthermore, the real-time detection module 11 is used to perform the following method: The process involves scanning to acquire a spatial point cloud dataset of the transmission line towers, performing denoising preprocessing and 3D reconstruction based on the dataset to generate a 3D point cloud model of the towers. Key components of the 3D point cloud model are identified based on bird nesting behavior characteristics, yielding information on N key tower components. Monitoring requirements are analyzed for these N key components, and sensor selection and deployment coverage analysis are performed based on the monitoring requirements to determine multi-sensor model specifications and deployment strategy parameters. Finally, multi-sensor deployment is implemented on the transmission line towers based on these parameters to construct a multimodal sensing network.

[0046] Furthermore, the real-time detection module 11 is used to perform the following method: DBSCAN was used to perform point cloud clustering analysis on the three-dimensional point cloud model of the tower to obtain the three-dimensional point cloud clustering analysis results. Based on the three-dimensional point cloud clustering analysis results, structural features were extracted from the three-dimensional point cloud model of the tower to obtain the structural feature set of the tower area. High-frequency areas of the structural feature set of the tower area were identified according to the fault frequency of the transmission line to obtain information on multiple high-frequency parts of the tower. Based on the nesting behavior characteristics of birds, key parts of the multiple high-frequency parts of the tower were screened to obtain information on multiple key parts of the tower.

[0047] Furthermore, the threat identification module 12 is used to perform the following methods: Historical datasets of nesting faults on power transmission lines are collected and standardized to obtain a dataset of available nesting faults on power transmission lines. The available datasets are then subjected to structured classification and behavior recognition training to generate a nesting behavior recognition branch channel. Threat assessment training is performed on the dynamic fault dataset of nesting behavior output by the nesting behavior recognition branch channel to construct a threat level assessment branch channel. The nesting behavior recognition branch channel and the nesting behavior recognition branch channel are then connected in series to form a dual-channel nesting threat assessment system, which is then stored in the edge computing unit.

[0048] Furthermore, the threat identification module 12 is used to perform the following methods: The available power transmission line nesting fault dataset is structured and classified to obtain a visual power transmission line nesting fault dataset and a non-visual power transmission line nesting fault dataset. Bird activity behavior is labeled on both datasets to obtain a visual power transmission line nesting behavior sample set and a non-visual power transmission line nesting behavior sample set. Behavior recognition training is performed based on these samples to obtain a visual nesting behavior recognition branch and a non-visual nesting behavior recognition branch. The visual and non-visual nesting behavior recognition branches are then weighted, fused, validated, and optimized to generate a nesting behavior recognition branch channel.

[0049] Furthermore, the threat identification module 12 is used to perform the following methods: Based on the safety application standards for power transmission lines, a bird nesting threat assessment system is established. This system includes bird species, nesting location, nest material type, size, and nesting behavior dimensions. The threat level of the dynamic fault dataset of nesting behavior output by the nesting behavior identification branch channel is assessed and labeled according to the bird nesting threat assessment system, resulting in a power transmission line nesting fault threat sample set. Threat assessment training, verification, and optimization are then performed based on this sample set to construct a threat level assessment branch channel.

[0050] Furthermore, the threat identification module 12 is used to perform the following methods: A dataset of the operating environment of the transmission line towers is collected, which includes meteorological conditions, geographical environment data, and equipment status data. The threat level of the operating environment dataset is assessed to determine the operating environment threat gain coefficient, and the nesting threat level parameter is adjusted by gain addition based on the operating environment threat gain coefficient.

[0051] Furthermore, the policy parsing module 13 is used to perform the following method: The nesting threat level parameter is matched with the bird nesting protection strategy library to obtain a matching nesting protection strategy combination; the matching nesting protection strategy combination is then adjusted and optimized according to the nesting protection constraints to determine the target nesting protection strategy parameters.

[0052] Furthermore, the anti-nesting module 14 is used to perform the following methods: According to the target nesting protection strategy parameters, the nesting blocking and driving linkage device is controlled to perform nesting prevention monitoring and feedback on the transmission line tower to obtain nesting protection effect parameters; based on the nesting protection effect parameters, the target nesting protection strategy parameters are optimized and updated, and nesting prevention optimization protection is performed through the updated target nesting protection strategy parameters.

[0053] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An adaptive protection method for preventing nesting on power transmission lines, characterized in that, The method includes: Key parts of transmission line towers are identified and multiple sensors are deployed to construct a multimodal sensing network. The multimodal sensing network is used to detect multimodal bird activity data streams in real time. The nesting threat assessment dual channel is invoked through the edge computing unit. The nesting threat assessment dual channel is composed of a nesting behavior recognition branch channel and a threat level assessment branch channel connected in series. Based on the nesting threat assessment dual channel, the multi-modal activity data stream of birds is identified and assessed, and the nesting threat level parameter is output. A bird nesting protection strategy database is established, and the nesting threat level parameters are analyzed using the bird nesting protection strategy database to determine the target nesting protection strategy parameters. The nest-building blocking and repelling linkage device is obtained, and the nest-building blocking and repelling linkage device is controlled to perform anti-nesting feedback protection on the transmission line tower based on the target nest-building protection strategy parameters.

2. The adaptive protection method for preventing nesting on transmission lines as described in claim 1, characterized in that, Constructing a multimodal sensing network includes: The spatial point cloud dataset of the transmission line tower is obtained by scanning. Based on the spatial point cloud dataset, noise reduction preprocessing and three-dimensional reconstruction are performed to generate a three-dimensional point cloud model of the tower. Based on the nesting behavior characteristics of birds, the key parts of the three-dimensional point cloud model of the tower are identified to obtain information on N key parts of the tower. The monitoring requirements of the key parts of the N towers are analyzed, and the sensor selection and deployment coverage are analyzed according to the tower monitoring requirements to determine the multi-sensor model specifications and multi-sensor deployment strategy parameters. Based on the multi-sensor model specifications and multi-sensor deployment strategy parameters, multi-sensors are deployed on the transmission line towers to construct a multi-modal sensing network.

3. The adaptive protection method for preventing nesting on transmission lines as described in claim 2, characterized in that, Obtain information on key components of N towers, including: DBSCAN was used to perform point cloud clustering analysis on the three-dimensional point cloud model of the tower to obtain the three-dimensional point cloud clustering analysis results; Based on the clustering analysis results of the three-dimensional point cloud, structural features are extracted from the three-dimensional point cloud model of the tower to obtain the structural feature set of the tower area. Based on the fault frequency of the transmission line, the high-frequency region identification is performed on the structural feature set of the tower area to obtain information on the high-frequency parts of multiple towers. Based on the nesting behavior characteristics of birds, the high-frequency information of the multiple poles was filtered to obtain the key information of multiple poles.

4. The adaptive protection method for preventing nesting on transmission lines as described in claim 1, characterized in that, The nesting threat assessment is accessed via a dual-channel system through edge computing units, including: Collect historical datasets of nesting faults on transmission lines, and standardize the historical datasets of nesting faults on transmission lines to obtain a usable dataset of nesting faults on transmission lines. The available power transmission line nesting fault dataset is subjected to structured classification and behavior recognition training to generate nesting behavior recognition branch channels; Threat assessment training is performed based on the dynamic fault dataset of nest building behavior output by the nest building behavior recognition branch channel to construct a threat level assessment branch channel. The nest-building behavior recognition branch channel and the nest-building behavior recognition branch channel are connected in series to form a nest-building threat assessment dual channel, and the nest-building threat assessment dual channel is stored in the edge computing unit.

5. The adaptive protection method for preventing nesting on transmission lines as described in claim 4, characterized in that, Generate nest-building behavior recognition branch channels, including: The available transmission line nesting fault dataset is structured and classified to obtain a visual transmission line nesting fault dataset and a non-visual transmission line nesting fault dataset. Bird activity behavior was labeled on the visual transmission line nesting fault dataset and the non-visual transmission line nesting fault dataset to obtain a visual transmission line nesting behavior sample set and a non-visual transmission line nesting behavior sample set. Based on the visual and non-visual transmission line nest building behavior sample sets, behavior recognition training is performed to obtain visual nest building behavior recognition branches and non-visual nest building behavior recognition branches. The visual nest-building behavior recognition branch and the non-visual nest-building behavior recognition branch are weighted, fused, verified, and optimized to generate a nest-building behavior recognition branch channel.

6. The adaptive protection method for preventing nesting on transmission lines as described in claim 4, characterized in that, Construct a threat level assessment branch channel, including: Based on the safety application standards for power transmission lines, a bird nesting threat assessment system is established, which includes bird species, nesting location, nesting material type, size, and nesting behavior dimensions. According to the bird nesting threat assessment system, the threat level of the dynamic fault dataset of nesting behavior output by the nesting behavior identification branch channel is assessed and labeled to obtain the power transmission line nesting fault threat sample set. Threat assessment training, verification, and optimization are conducted based on the aforementioned power transmission line nesting fault threat sample set, and a threat level assessment branch channel is constructed.

7. The adaptive protection method for preventing nesting on transmission lines as described in claim 1, characterized in that, The method further includes: Collect the operating environment dataset of the transmission line towers, which includes meteorological condition data, geographical environment data, and equipment status data; The threat level of the operating environment dataset is assessed to determine the operating environment threat gain coefficient, and the nesting threat level parameter is adjusted by gain addition based on the operating environment threat gain coefficient.

8. The adaptive protection method for preventing nesting on transmission lines as described in claim 1, characterized in that, Determine the target nesting defense strategy parameters, including: The nesting threat level parameter is matched with the bird nesting protection strategy library to obtain a matching nesting protection strategy combination. The matching nesting protection strategy combination is constrained and optimized according to the nesting protection constraints to determine the target nesting protection strategy parameters.

9. The adaptive protection method for preventing nesting on transmission lines as described in claim 1, characterized in that, Based on the target nesting protection strategy parameters, the nesting blocking and dispersing linkage device is controlled to perform anti-nesting feedback protection on the transmission line tower, including: According to the target nesting protection strategy parameters, the nesting blocking and driving linkage device is controlled to perform anti-nesting monitoring and feedback on the transmission line tower to obtain nesting protection effect parameters. Based on the nesting protection effect parameters, the target nesting protection strategy parameters are optimized and updated, and nesting protection is optimized and protected using the updated target nesting protection strategy parameters.

10. A self-adaptive protection system for preventing nesting on power transmission lines, characterized in that: For implementing the adaptive protection method for preventing nesting on transmission lines according to any one of claims 1-9, the system comprises: Real-time detection module: Identifies key parts of transmission line towers and deploys multiple sensors to construct a multimodal sensing network, and uses the multimodal sensing network to detect multimodal bird activity data streams in real time; Threat identification module: The edge computing unit calls the nesting threat assessment dual channel, which is composed of a nesting behavior identification branch channel and a threat level assessment branch channel connected in series. Based on the nesting threat assessment dual channel, the multi-modal activity data stream of birds is identified and assessed, and nesting threat level parameters are output. Strategy Analysis Module: Builds a bird nesting protection strategy library, uses the bird nesting protection strategy library to analyze the nesting threat level parameters, and determines the target nesting protection strategy parameters; Anti-nesting module: acquires nesting blocking and driving-off linkage device, and controls the nesting blocking and driving-off linkage device to perform anti-nesting feedback protection on the transmission line tower based on the target nesting protection strategy parameters.

Images