Method, device and system for identifying bird-related faults of an electrical grid

Abstract

The method, processing device, equipment, and system for identifying bird-related faults in power grids provided in this application acquire target data through multi-source sensors. The target data includes preprocessed temporal motion data, acoustic spectrum data, and spatial morphological data. Based on a machine learning model, the target data is weighted and fused to construct a three-dimensional feature matrix. The target three-dimensional feature matrix is ​​matched with a preset fault feature library to determine the fault type, thus achieving accurate identification of bird-related faults in power grids. Simultaneously, based on the fault type and a preset risk threshold, a graded expulsion command is generated. Based on the graded expulsion command, a combined expulsion operation is executed, achieving effective bird expulsion and improving the accuracy of bird-related fault identification and the efficiency of bird expulsion.

Description

Technical Field

[0001] This application relates to the field of power system safety protection technology, and in particular to a method, processing device, equipment and system for identifying bird-related faults in power grids. Background Technology

[0002] In recent years, bird-related faults have become one of the main hidden dangers affecting the safe operation of the power grid. According to power grid operation data, line tripping faults caused by bird activity account for more than 30% in South China, with short circuits caused by birds colliding with conductors, line grounding caused by bird nests, and insulator flashovers caused by bird droppings being the main types of faults.

[0003] In existing technologies, a single sensor detection device or a simple multi-sensor triggering device is mainly used to monitor the movement trajectory of birds, or a fixed frequency and fixed intensity sound and light deterrence method is used to drive away birds, thereby realizing the monitoring and deterrence of birds.

[0004] However, the above-mentioned technical solutions for power grid bird-related fault prevention and control have the problem of low multimodal monitoring fusion, which makes it impossible to deeply integrate sensor data, accurately distinguish fault types, and result in insufficient identification accuracy. Summary of the Invention

[0005] This application provides a method, processing device, equipment, and system for identifying bird-related faults in power grids, in order to improve the accuracy of bird-related fault identification and the efficiency of bird removal.

[0006] In a first aspect, this application provides a method for identifying bird-related faults in power grids, including:

[0007] Based on multi-source sensors, target data is acquired, including preprocessed temporal motion data, acoustic spectrum data, and spatial morphology data.

[0008] Based on the machine learning model, the target data is weighted and fused to construct a target three-dimensional feature matrix;

[0009] The target three-dimensional feature matrix is ​​matched with a preset fault feature library to determine the fault type and complete the identification and processing of bird-related faults in the power grid.

[0010] Furthermore, after performing feature matching between the target three-dimensional feature matrix and a preset fault feature library to determine the fault type and complete the identification and processing of bird-related faults in the power grid, the method further includes:

[0011] The fault identification confidence level is determined based on the fault type and the preset risk threshold.

[0012] Based on the fault identification confidence level, the fault level is determined, including low level, medium level and high level;

[0013] Based on the fault level, a graded expulsion command is generated, which includes a low-level expulsion command, a medium-level expulsion command, and a high-level expulsion command.

[0014] According to the tiered bird removal instructions, execute the combined bird removal operation to complete the handling of bird-related faults in the power grid.

[0015] Furthermore, the graded expulsion commands include low-level expulsion commands, medium-level expulsion commands, and high-level expulsion commands;

[0016] The low-level expulsion command is a low-intensity expulsion operation that activates the ultrasonic loudspeaker.

[0017] The medium-level expulsion command is a combined expulsion operation that activates an ultrasonic horn, a multi-color strobe light, and a low-frequency warning sound.

[0018] The high-level expulsion command is a combined expulsion operation that activates ultrasonic horns, multi-color strobe lights, low-frequency warning sounds, and laser spotting.

[0019] Furthermore, after executing a combined bird removal operation according to the tiered bird removal command to complete the handling of bird-related faults in the power grid, the method further includes:

[0020] Based on the multi-source sensors, collect target update data;

[0021] Based on the target update data, the fault identification result is determined;

[0022] If the fault identification result indicates that the current fault is still at a medium or high level, the graded removal command is executed repeatedly until the fault identification result indicates that there is no fault.

[0023] Furthermore, based on multi-source sensors, target data is acquired, including:

[0024] Identify the radar sensor, acoustic signature collector, and laser detector among the multi-source sensors;

[0025] Based on the radar sensor, the target's distance, velocity, and azimuth time-series data are collected to obtain time-series motion data;

[0026] Based on the aforementioned voiceprint collector, the voiceprint data of the target is collected to obtain voiceprint spectrum data;

[0027] Based on the laser detector, the target's contour dimensions and three-dimensional coordinate data are collected to obtain spatial morphology data;

[0028] The temporal motion data, the acoustic signature spectrum data, and the spatial morphology data are preprocessed respectively to obtain the target data.

[0029] Secondly, this application provides a device for handling bird-related faults in power grids, comprising:

[0030] The main unit is used for installation on transmission towers;

[0031] A multimodal monitoring module, integrated on the main unit, is used to synchronously collect multi-source sensor data within a predetermined area of ​​the transmission tower. The multi-source sensor data includes at least temporal motion data, acoustic spectrum data, and spatial morphology data.

[0032] The main control module, integrated within the main unit, is communicatively connected to the multimodal monitoring module. It is used to perform fusion analysis and identification based on the multi-source sensor data, determine the fault type and risk level, and generate graded expulsion commands according to the risk level.

[0033] A graded expulsion module, integrated on the main unit and communicatively connected to the main control module, is used to receive the graded expulsion command and execute a combined expulsion operation that matches the risk level.

[0034] A strong electromagnetic interference resistant power supply component is connected to the main unit and is used to provide power to the multimodal monitoring module, intelligent main control module and graded drive-off module. The strong electromagnetic interference resistant power supply component has an electromagnetic shielding structure.

[0035] Furthermore, the multimodal monitoring module includes a radar sensor, an acoustic signature collector, and a laser detector;

[0036] The main control module includes a fault identification unit, a drive-off strategy unit, and a communication unit.

[0037] Furthermore, the fault identification unit is equipped with a machine learning model for constructing a three-dimensional feature matrix;

[0038] The graded expulsion module is equipped with an ultrasonic speaker group and an audio-visual warning component.

[0039] The ultrasonic speaker assembly is used to emit high-frequency ultrasonic waves.

[0040] The sound and light warning component is used to emit audible warning sounds and visible warning lights.

[0041] Thirdly, this application provides a device for handling bird-related faults in power grids, including: a memory and a processor;

[0042] The memory stores computer-executed instructions;

[0043] The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any of the first aspects.

[0044] Fourthly, this application provides a system for handling bird-related faults in power grids, including a computer program and a device for handling bird-related faults in power grids;

[0045] Wherein, when the computer program is executed by the processor, it implements the method described in any one of the first aspects;

[0046] The device for handling bird-related faults in the power grid is the device described in any of the second aspects.

[0047] The method, processing device, equipment, and system for identifying bird-related faults in power grids provided in this application acquire target data through multi-source sensors. The target data includes preprocessed temporal motion data, acoustic spectrum data, and spatial morphological data. Based on a machine learning model, the target data is weighted and fused to construct a three-dimensional feature matrix. The target three-dimensional feature matrix is ​​matched with a preset fault feature library to determine the fault type, thus achieving accurate identification of bird-related faults in power grids. Simultaneously, based on the fault type and a preset risk threshold, a graded expulsion command is generated. Based on the graded expulsion command, a combined expulsion operation is executed, achieving effective bird expulsion and improving the accuracy of bird-related fault identification and the efficiency of bird expulsion. Attached Figure Description

[0048] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0049] Figure 1 A schematic diagram of the structure of the power grid bird-related fault handling device provided in this application;

[0050] Figure 2 A flowchart illustrating an embodiment of the method for identifying and repelling bird-related faults in power grids provided in this application;

[0051] Figure 3 A flowchart illustrating Embodiment 2 of the method for identifying and repelling bird-related faults in power grids provided in this application;

[0052] Figure 4 A flowchart illustrating Embodiment 3 of the method for identifying and repelling bird-related faults in power grids provided in this application;

[0053] Figure 5 A schematic diagram of the structure of the power grid bird-related fault handling equipment provided in this application.

[0054] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0055] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0056] This application applies to bird activity monitoring and control scenarios on 110kV-500kV high-voltage transmission lines, especially targeting the complex environment of South China characterized by high temperature (40℃+), high humidity (relative humidity ≥80%), strong electromagnetic interference (100kV / m+), and abundant salt spray. The device, installed in a main unit below the crossarm of the tower, integrates a multi-modal monitoring module (radar, microphone, laser detector), a graded decoy module (ultrasonic, strobe light, high-volume speaker), and a strong electromagnetic interference-resistant power supply component. It achieves bidirectional communication with the power grid's backend system via a 4G / LoRa network, forming a distributed intelligent control network of "front-end perception - mid-end analysis - back-end linkage."

[0057] The existing technologies have the following problems in the prevention and control of bird-related faults in power grids: (1) Single monitoring dimension: radar or infrared sensors can only capture movement trajectories and cannot distinguish different fault types such as bird nests, bird invasion, and bird droppings accumulation; (2) Fixed expulsion strategy: fixed frequency sound and light expulsion is easily adapted by birds and lacks a dynamic adjustment mechanism for risk level, resulting in a decrease in expulsion effect; (3) Insufficient anti-interference capability: power supply components and communication modules lack full-link electromagnetic shielding design, and data interruption or sensor failure is likely to occur in the strong electromagnetic environment of 500kV high-voltage lines; (4) Poor regional adaptability: sensor lenses are prone to fogging due to high humidity environment, and the identification model lacks local bird characteristic data support, resulting in low identification accuracy in South China; (5) Lack of closed-loop feedback: no secondary verification and closed-loop feedback mechanism is set after expulsion, and the fault hazards may not be completely eliminated, and operation and maintenance rely on manual inspection.

[0058] To address the aforementioned technical challenges, this application constructs a comprehensive intelligent prevention and control system for bird-related faults in power grids through multimodal sensor fusion and deep learning model collaboration. Specifically, it centers on "synchronous multi-source data acquisition - three-dimensional feature matrix fusion - dynamic risk level determination - tiered expulsion strategy linkage - closed-loop feedback optimization." This system utilizes multimodal data acquisition from K-band radar, omnidirectional microphones, and laser detectors, combined with spatial-temporal-acoustic feature fusion of a CNN-LSTM hybrid network model, to achieve accurate classification and risk assessment of bird activity. Simultaneously, a tiered expulsion strategy is designed based on risk levels, and the expulsion effect and device stability are ensured through laser-based conductor damage prevention linkage mechanisms and low-power standby modes. This concept overcomes the limitations of existing technologies, such as single-monitoring, fixed expulsion methods, and weak anti-interference capabilities, achieving intelligent closed-loop prevention and control of bird-related faults in power grids.

[0059] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0060] Figure 1 A schematic diagram of the structure of the power grid bird-related fault handling device provided in this application. Figure 1 As shown, the power grid bird-related fault handling device 100 includes a main unit 101 installed on the power grid, and a multi-modal monitoring module 102, a main control module 103, a graded expulsion module 104, and a strong electromagnetic interference protection power supply component 105 integrated on the main unit 101. Each module is detachably connected via shielded aviation connectors, facilitating on-site maintenance and replacement, and improving the modularity and maintainability of the device. The main unit 101 is made of die-cast aluminum alloy, offering advantages such as high structural strength and light weight. Its surface undergoes anodizing for corrosion protection, achieving an overall protection level of no less than IP67. This effectively solves the problems of existing devices being prone to corrosion and poor sealing leading to damage to internal components in harsh outdoor environments such as high temperature and humidity, ensuring the long-term structural integrity and reliability of the device under complex climatic conditions.

[0061] Specifically, the main unit 101 is installed on the transmission tower of the power grid, providing an integrated installation platform and physical protection for each functional module. In one possible implementation of this application, the main unit is made of die-cast aluminum alloy, with dimensions of 350mm×250mm×180mm. The surface is treated with hard anodizing for corrosion protection, and the overall protection level is IP67, which can adapt to the outdoor environment of high temperature, high humidity, rain, and salt spray in Guangdong. The main unit is fixed to the underside of the crossarm of the 500kV transmission tower by a customized mounting bracket. The mounting bracket is compatible with crossarm angle steel with a diameter of 100mm, making installation convenient and requiring no modification to the tower.

[0062] The multimodal monitoring module 102, integrated into the main unit, is used to synchronously collect multi-source sensor data within a predetermined area of ​​the transmission tower, including visible light, infrared thermal imaging, and radar, to overcome the problem of insufficient recognition accuracy caused by the limited dimensions and incomplete information of a single sensor. The multimodal monitoring module 102 includes a radar sensor, an acoustic signature collector (or omnidirectional pickup array), and a laser detector. The multi-source sensor data includes at least temporal motion data, acoustic signature spectrum data, and spatial morphological data. The synchronous acquisition of multi-source heterogeneous data provides a complete data foundation for subsequent high-precision fusion recognition.

[0063] In this application, the radar sensor consists of two sets of K-band radar sensors symmetrically mounted on both sides of the top of the main unit. The operating frequency is 24.125 GHz, the detection range is 0.5-50 meters, the angular resolution is ≤1° (e.g., 0.8°), the beamwidth is 60° horizontally and 30° vertically, the data sampling rate is 10 Hz, and the built-in Kalman filter algorithm can accurately capture the flight speed, distance, and azimuth data of birds within a 50-meter range. This is used for high-precision, all-weather acquisition of time-series motion data such as distance, speed, and azimuth of targets, solving the problem of traditional video monitoring failing under low visibility conditions such as nighttime and foggy weather.

[0064] The omnidirectional microphone array consists of four omnidirectional microphones evenly distributed 90° around the main unit. It has a frequency response range of 200Hz-15kHz, a sensitivity of -38dB, and a pickup distance of 30 meters. It has a built-in windproof and noise reduction filter circuit, which can effectively filter wind noise and rain noise during typhoons and rainy weather in Guangdong. It can collect bird voiceprint signals with a signal-to-noise ratio of ≥45dB, which can be used to collect bird voiceprint signals without blind spots, making up for the shortcomings of pure visual or radar monitoring in being unable to identify stationary or hidden birds (such as in their nests).

[0065] The laser detector consists of a single 532nm wavelength laser detector, installed in the middle of the main unit. It achieves 360° circumferential rotation and 0°-180° pitch adjustment via a rotation mechanism driven by a two-phase stepper motor, with a step angle of 0.1°, detection accuracy of ±2cm, and a scanning speed of 30° / s. It features high spatial resolution, and the lens is equipped with a hydrophobic coating to ensure detection performance in rainy and foggy weather. It prevents water accumulation and fogging in rainy weather. The default scanning area is high-incidence areas of bird-related faults, such as insulator strings on towers, tension clamps of conductors, and ends of crossarms. It is used to accurately collect the outline size and three-dimensional coordinate data of the target, providing key spatial morphological information for distinguishing different fault types such as bird bodies, bird nests, and bird droppings accumulation.

[0066] The main control module 103, integrated within the main unit 101, communicates with the multimodal monitoring module 102. It is used for fusion analysis and intelligent identification based on multi-source sensor data to determine the fault type and risk level, and generate graded expulsion commands according to the risk level. This realizes automation from data acquisition to intelligent decision-making, solving the problems of low efficiency and slow response of manual inspection.

[0067] Specifically, the main control module 103 includes a core processor, a fault identification unit, a bird deterrence strategy unit, and a communication unit. In this application, the core processor uses a high-performance STM32H743VIT6 microcontroller with a main frequency of 480MHz, 1MB RAM, and 2MB Flash memory, meeting real-time data processing and control requirements and ensuring the efficient operation of complex algorithms. The fault identification unit incorporates a multi-source data fusion algorithm and a pre-trained CNN-LSTM hybrid network model to construct a unified three-dimensional feature matrix from radar time-series data, acoustic signature spectral features, and laser profile data. This enables accurate classification and identification of various bird-related faults, such as nest construction, bird intrusion into conductor danger zones, and bird droppings accumulation on insulators. Its identification accuracy is significantly higher than that of identification methods based on a single data source. The bird deterrence strategy unit presets three levels of bird-related fault risk thresholds, outputting corresponding graded deterrence commands based on the identification results and quantified risk levels. This achieves precise allocation and intelligent management of bird control resources, avoiding the problem of birds developing adaptive behavior due to single-intensity deterrence. The communication unit integrates a 4GCat.1 all-network compatible module and a LoRa gateway for bidirectional data transmission with the power grid backend. The 4GCat.1 module supports China Mobile, China Unicom, and China Telecom networks, ensuring reliable remote communication in most areas. The LoRa gateway has a communication distance of ≤3km and supports a customizable data upload cycle of 1-60 minutes, providing flexibility for local networking and low-power data transmission. It also supports remote updates of drive-off strategies and model parameters from the power grid backend, enabling the device to perform remote operation and maintenance and algorithm iteration upgrades.

[0068] In one implementation of this application, the main control module 103 also features a low-power standby mode. When the device has no fault identification results for a period of time, it automatically extends the radar detection cycle to 10 seconds / time and shuts down the microphone and laser detector, ensuring standby power consumption ≤50mW. This design significantly reduces the overall energy consumption of the device, extends the battery's power supply time during continuous rainy weather, and improves the system's sustainable operating capability. Furthermore, the main control module 103 has a built-in local storage unit with a capacity of no less than 100,000 records, used to record the entire process data, including fault occurrence time, duration, identification results, expulsion process, and early warning information. It supports local data retrieval and remote export, solving the problem of data loss during communication interruptions and providing complete data traceability for fault analysis and strategy optimization.

[0069] The graded expulsion module 104 is integrated into the main unit and communicates with the main control module 103. It is used to receive graded expulsion instructions and execute combined and tiered expulsion operations that match the risk level. Through the combination and intensity adjustment of various physical stimuli, such as lasers (light curtains, motion effects), ultrasound, bionic voice and explosion-proof lights, it effectively reduces the adaptability of birds and improves the success rate of long-term expulsion.

[0070] Specifically, in this application, the graded deterrence module includes a tweeter group, an ultrasonic speaker group, and a multi-color strobe light group. The tweeter group consists of two 15W tweeters installed on the front and rear sides of the main unit, with a frequency response range of 1kHz-20kHz and a maximum sound pressure level of 115dB. It pre-stores the calls of predators of birds commonly found in South China that can cause line faults, as well as high-frequency warning sounds. For example, it includes 12 calls of predators of birds (magpies, crows, mynas, herons, etc.) (hawks, peregrine falcons, snakes), and 8 high-frequency warning sounds. It supports stepless volume adjustment from 60-110dB, using the birds' auditory fear response for deterrence, making the sound content highly targeted. In one possible implementation of this application, the maximum sound pressure level of the tweeter group is ≥110dB, supporting graded volume adjustment from 60-110dB, allowing for flexible adjustment of the deterrence intensity based on the risk level and ambient noise.

[0071] The ultrasonic speaker assembly consists of four ultrasonic speakers symmetrically distributed on the left and right sides of the main unit. Operating at a frequency of 20-40kHz, it supports intermittent pulse output and utilizes the ultrasonic frequency band sensitive to birds for deterrence, with minimal interference to the human ear. In one possible implementation of this application, the ultrasonic speaker assembly has an output power of 8W, an effective deterrence distance of ≤20 meters, supports intermittent pulse output, and can dynamically adjust the output frequency to prevent birds from developing adaptation, providing an effective means of deterring birds at medium to close ranges.

[0072] The multi-color strobe light assembly consists of six red, blue, and yellow strobe lights arranged around the bottom of the main unit. It supports dynamic adjustment of the flashing frequency and color combination, using the visual fright of birds to specific flashing lights to deter them, with particularly significant effects at night. In one possible implementation of this application, the multi-color strobe light assembly has a rated voltage of 12V, a single lamp power of 3W, a beam angle of 120°, and a flashing frequency that supports dynamic adjustment from 5-20Hz. The color combination and flashing frequency can be switched according to the risk level to create diverse visual deterrence modes.

[0073] The strong electromagnetic interference protection power supply component 105 is connected to the main unit 101 and is used to provide stable and clean power to the multi-mode monitoring module, the main control module and the graded drive-off module. The strong electromagnetic interference protection power supply component has a multi-level electromagnetic shielding structure, which fundamentally solves the core problem of existing devices being unstable, malfunctioning or even damaged due to power interference and high signal noise in the strong electromagnetic environment of 110kV-500kV high voltage transmission lines.

[0074] Specifically, in one possible implementation of this application, the strong electromagnetic interference protection power supply component includes two 50W arc-shaped monocrystalline silicon solar panels with a conversion efficiency of 23.5%. These panels feature an arc-shaped structure with a curvature radius of 50cm, adaptable to the installation angle of the tower crossarm, and are covered with 3.2mm thick anti-glare tempered glass, providing strong impact resistance. The lithium iron phosphate battery pack is a 12V / 20Ah lithium iron phosphate battery with a cycle life ≥2500 cycles and an operating temperature of -40℃ to 65℃, suitable for the high temperatures of summer in Guangdong. It is equipped with an intelligent BMS management system, providing overcharge, over-discharge, over-temperature, short circuit, and overcurrent protection. The power supply components with a full-link electromagnetic shielding structure use a 1.5mm thick galvanized steel plate for the shielding shell, with a 0.3mm thick high-conductivity copper foil inside. The shell is connected to the tower grounding body via a dedicated grounding wire, with a grounding resistance ≤3Ω. All power supply and signal lines use shielded cables with 95% shielding coverage, and the shielding layer is grounded at both ends. The power supply lines are connected in series with an electromagnetic filter module, operating at a frequency of 10kHz-1GHz, with an insertion loss ≥45dB. The main control module PCB board adopts a 4-layer design, with the middle two layers being the power layer and ground layer, achieving full isolation between power and signals, and can operate stably in the strong electromagnetic environment of a 500kV high-voltage line. The DC-DC power conversion module uses an industrial-grade DC-DC module with a wide input voltage of 9-36V, outputting a stable 5V / 12V DC voltage with a ripple ≤30mV, providing stable power to all modules. The end-to-end electromagnetic shielding structure includes a 1.5mm thick galvanized steel shielding shell encasing the power supply components, a 0.3mm thick copper foil layer inside the shell, shielded cables with a shielding coverage of ≥90%, and an electromagnetic filter module connected in series with the power supply line. The grounding resistance of the shielding shell is ≤4Ω. This composite shielding design achieves dual suppression of strong electromagnetic field radiation and conducted interference, providing a "clean" working environment for the internal precision electronic circuits. A DC-DC power conversion module provides a stable 5V / 12V power supply to each module, ensuring that voltage fluctuations do not affect sensor accuracy and processor operation.

[0075] In one implementation of this application, the arc-shaped monocrystalline silicon solar panel consists of two 50W monocrystalline silicon panels with a conversion efficiency of ≥23%. The arc-shaped structure with a curvature radius of 50cm better conforms to the tower installation angle, maximizing solar radiation reception and improving energy collection efficiency. The lithium iron phosphate battery pack is a 12V / 20Ah battery with a cycle life of ≥2000 cycles. It is equipped with a BMS battery management system, supporting overcharge, over-discharge, and over-temperature protection. Its high-temperature stability is excellent, making it particularly suitable for the high-temperature environment of transmission line corridors, ensuring the safety and long lifespan of the energy storage system.

[0076] In one implementation of this application, the fault identification unit is pre-configured with a neural network model, namely a CNN-LSTM hybrid network model. This model includes a CNN network with 3 convolutional layers and 2 pooling layers for extracting laser contour features and acoustic signature spectral features, a 2-layer bidirectional LSTM network for extracting radar temporal motion features, and a feature fusion attention layer. The feature fusion attention layer adaptively weights and fuses the spatial features output by the CNN network and the temporal features output by the LSTM network to construct a target three-dimensional feature matrix containing spatial morphological features, temporal motion features, and acoustic signature spectral features. The fault type and identification confidence are then output through a fully connected layer. This model structure fully utilizes the characteristics of different data, achieving deep feature fusion and extraction, significantly improving the classification accuracy and robustness of bird-related faults in complex scenarios. Furthermore, the CNN-LSTM hybrid network model includes a pre-configured fault feature library trained on over 30,000 bird samples collected in South China, supporting effective fault identification with a confidence level ≥85%. This localized training makes the model more closely aligned with the actual bird situation in the region, reducing false alarms and false negatives.

[0077] In one implementation of this application, the three-level risk threshold preset in the deterrence strategy unit are as follows: (1) Low risk threshold, i.e., when the bird is ≥30 meters away from the wire, or the area of ​​bird droppings accumulation is <10cm², there is no short circuit risk; (2) Medium risk threshold, i.e., when the bird is 10-30 meters away from the wire, or the area of ​​bird droppings accumulation is 10-20cm², there is a potential short circuit risk; (3) High risk threshold, i.e., when the bird is ≤10 meters away from the wire, or the area of ​​bird droppings accumulation is ≥20cm², or the bird nest outline size increases by ≥3cm in three consecutive collections, there is an immediate short circuit risk. This quantitative threshold system makes risk judgment based on evidence, accurately matches response strategies, and avoids insufficient prevention and control or overreaction.

[0078] On the other hand, the decoy strategy unit also features a laser-based conductor damage prevention linkage mechanism. Specifically, when the laser detector detects a target within ≤5 meters of the conductor, it automatically reduces the laser output power to a safe range below 50mW. Simultaneously, it controls the rotating mechanism to ensure the laser beam forms an angle ≥30° with the conductor, preventing direct laser damage to the conductor's insulation layer. This mechanism, while ensuring monitoring accuracy, proactively eliminates potential safety hazards to the power grid from the laser equipment, reflecting the safety considerations in the device design.

[0079] In one implementation of this application, the data sampling rate of the radar sensor can be set to 10Hz, and a built-in Kalman filter algorithm effectively smooths the data trajectory and improves the accuracy of motion parameter estimation; the omnidirectional microphone group has built-in high-pass and low-pass filter circuits to filter environmental noise below 200Hz and above 15kHz (such as wind noise and power grid hum), highlighting the bird's voiceprint characteristics and improving the signal-to-noise ratio of acoustic recognition; the laser detector has a scanning step size of 0.5° and a built-in median filter algorithm to optimize contour data, effectively suppressing point cloud noise caused by tiny particles such as raindrops and flying insects, and ensuring the clarity of the 3D reconstruction.

[0080] The power grid bird-related fault handling device provided in this application solves the problem of insufficient information dimensions of single monitoring methods by synchronously collecting data from heterogeneous sensors in a multimodal monitoring module. Through multi-source fusion identification and three-level risk quantification judgment based on a CNN-LSTM hybrid network within the main control module, it achieves high-precision (≥95%) classification and intelligent early warning of bird-related faults, overcoming the shortcomings of traditional methods such as low recognition rate and high false alarm rate. The tiered repulsion module, using a combination of sound, light, and ultrasonic waves for step-by-step repulsion, effectively reduces the adaptability of birds and improves the long-term repulsion success rate (≥90%). The device's robust electromagnetic shielding structure and IP67 high-protection main body design ensure extreme stability and reliability under complex environments of high voltage, strong electromagnetic fields, high temperature, and high humidity. This device forms a complete intelligent closed loop of "monitoring-identification-decision-repulsion-feedback," which can significantly reduce the line tripping rate caused by bird damage in the power grid and improve the intelligent protection level of power grid safety operation.

[0081] Figure 2 This is a flowchart illustrating an embodiment of the method for identifying and repelling bird-related faults in power grids provided in this application. Figure 2 As shown, the method includes:

[0082] S201. Acquire target data based on multi-source sensors.

[0083] Among them, multi-source sensors refer to those integrated into Figure 1The sensor set on the power grid bird-related fault handling device shown is capable of collecting information from different physical dimensions. In this embodiment, it specifically refers to a K-band radar sensor, an omnidirectional microphone group (acoustic fingerprint collector), and a laser detector.

[0084] Target data refers to the set of data that, after preprocessing, can be used for subsequent intelligent identification and characterizes the features of targets (mainly birds and their activity products) within the monitoring area. It includes preprocessed temporal motion data, acoustic spectrum data, and spatial morphological data.

[0085] Specifically, this step is the foundation of environmental perception and includes the following steps: (1) Identify the radar sensor, acoustic signature collector, and laser detector among the multi-source sensors and start their collaborative operation. (2) Based on the radar sensor, collect the target's distance, speed, and azimuth time-series data to obtain time-series motion data. The radar sensor, with its all-weather and strong anti-interference characteristics, continuously outputs the target's motion trajectory information, solving the problem of traditional video surveillance failing at night and in rainy or foggy weather, and providing key kinematic basis for judging whether birds are active in dangerous areas. (3) Based on the acoustic signature collector, collect the target's acoustic signature data to obtain acoustic signature spectrum data. Through a circumferentially distributed microphone array, capture the sound features of birds such as calls and flapping wings without blind spots and convert them into frequency domain features. This makes up for the deficiency of pure vision or radar monitoring in identifying stationary and concealed (such as in nests) birds, and realizes the acoustic verification of the existence of birds. (4) Based on the laser detector, collect the target's outline size and three-dimensional coordinate data to obtain spatial morphology data. High-precision laser scanning is used to obtain the precise geometric dimensions and spatial location of the target. This provides irreplaceable spatial information for distinguishing fault types with different physical forms, such as "bird body", "bird nest" and "bird droppings", and is a prerequisite for accurate classification. (5) The temporal motion data, acoustic spectrum data and spatial morphology data are preprocessed to obtain the target data. Preprocessing is a key step in improving data quality.

[0086] For example, radar data is smoothed using a Kalman filter algorithm (an optimal estimation algorithm), with a 5-frame filtering window to remove random noise and transient interference, outputting stable and accurate target motion parameters. The voiceprint signal is sequentially passed through a high-pass filter (cutoff frequency 200Hz) and a low-pass filter (cutoff frequency 15kHz) circuit to effectively filter out low-frequency wind noise, power grid interference, and high-frequency electronic noise. Then, a Fourier transform is used to convert the time-domain sound signal into a frequency-domain spectrum that better characterizes bird features, significantly improving the signal-to-noise ratio of voiceprint recognition. Laser contour data is processed using a 3×3 window median filter algorithm. This algorithm effectively removes discrete noise generated by raindrops, flying insects, dust, etc., while preserving the true edges of the target, optimizing the quality of the 3D point cloud data and providing a clear and reliable data foundation for subsequent contour recognition.

[0087] This step, through synchronous acquisition and targeted preprocessing of heterogeneous sensors, constructs a comprehensive and clean multi-source data pool, fundamentally solving the problems of single monitoring dimensions and data quality being greatly affected by environmental noise in existing technologies, thus laying a solid data foundation for subsequent high-precision fusion identification.

[0088] S202. Based on the machine learning model, the target data is weighted and fused to construct a three-dimensional feature matrix of the target.

[0089] In this embodiment, the machine learning model specifically refers to the CNN-LSTM hybrid network model, which has been trained using a large number of labeled samples and has the ability to extract effective features from the data.

[0090] Weighted fusion processing refers to the adaptive assignment of different importance weights to features from different sensors through the attention mechanism in the model.

[0091] The target three-dimensional feature matrix refers to a data structure that integrates features from three dimensions: spatial morphology, temporal motion, and acoustic spectrum.

[0092] This step inputs the three types of target data obtained from S201 into the pre-trained CNN-LSTM hybrid network model. The specific workflow is as follows: (1) CNN network branch: using its 3-layer convolution and 2-layer pooling structure, extracts spatial morphological features (such as shape and size) from laser contour data and acoustic spectrum features (such as frequency distribution and harmonic structure) from acoustic spectrum data. (2) Bidirectional LSTM network branch: using its 2-layer bidirectional structure, extracts temporal motion features (such as motion trajectory and speed change pattern) from radar time series data. (3) Feature fusion attention layer: this layer receives the features output by the CNN and LSTM branches. Instead of simply splicing them together, it adaptively calculates the importance weights of each feature dimension through the attention mechanism and performs weighted fusion. For example, when identifying a stationary bird's nest, the weight of spatial morphological features will be higher; when identifying flying birds, the weight of temporal motion features will be more prominent. Finally, a unified and complementary three-dimensional feature matrix of the target is constructed.

[0093] This step achieves deep and adaptive multi-source information fusion through an advanced deep learning model, overcoming the problems of insufficient feature utilization and poor recognition robustness caused by traditional simple data splicing or rule judgment, and significantly improving the model's ability to understand and represent complex and variable bird-related scenes.

[0094] S203. Perform feature matching between the target's three-dimensional feature matrix and the preset fault feature library to determine the fault type.

[0095] The preset fault feature library refers to a feature set stored locally on the device, trained based on a large number of bird samples (voiceprints, morphology) collected in a specific region (such as South China), which is used as a "template" for identification and comparison.

[0096] Specifically, this step involves performing similarity matching between the target 3D feature matrix constructed in step S202 and the feature templates in a pre-defined fault feature library. This feature library is localized and trained on over 30,000 samples from 16 common wading birds in South China. During training, data augmentation techniques such as noise addition, time shifting, and spectral stretching were employed to improve the model's generalization ability and robustness. After matching, the model outputs a fault type (e.g., no fault, bird nest, bird intrusion, bird droppings accumulation).

[0097] This step, by introducing a localized and highly robust feature library and a confidence judgment mechanism, ensures that the recognition results not only closely match the actual bird situation in the region but also have high reliability, thus solving the problem of insufficient recognition accuracy and high false alarm rate of general models in specific scenarios.

[0098] S204. Generate graded expulsion instructions based on the fault type and preset risk threshold.

[0099] Among them, the preset risk threshold refers to the pre-set numerical standard used to quantify and assess the degree of fault risk. In this embodiment, it is specifically defined as three levels of thresholds: low, medium, and high.

[0100] Graded expulsion commands refer to a set of commands that control different expulsion devices to operate in specific modes, corresponding to different risk levels.

[0101] This step achieves an intelligent leap from "recognition" to "decision-making." Specifically, it includes: (1)

[0102] Based on the fault type and preset risk threshold, the fault identification confidence level is determined, that is, the identification result (fault type) and precise data such as laser ranging are combined and judged against the preset quantitative threshold. In this embodiment, the low risk threshold is set as birds are ≥30 meters away from the conductor, or bird droppings accumulation area <10cm²; the medium risk threshold is set as birds are 10-30 meters away from the conductor, or bird droppings accumulation area 10-20cm²; the high risk threshold is set as birds are ≤10 meters away from the conductor, or bird droppings accumulation area ≥20cm², or bird nest size continues to increase. (2) Based on the fault identification confidence level, the fault level is determined, that is, low level, medium level or high level. (3) Based on the fault level, specific drive-away instructions for the low, medium and high levels are generated. In particular, when the laser ranging finds that the target is ≤5 meters away from the conductor, the drive-away strategy unit will simultaneously trigger the laser anti-conductor damage linkage mechanism, automatically reduce the laser power and adjust the beam angle, while ensuring the drive-away effect, completely avoiding potential damage to the insulation layer of the transmission conductor by the laser, reflecting the design concept of active safety.

[0103] This step transforms vague risk perception into precise, tiered quantitative decisions, making prevention and control responses more scientific and accurate. It solves the problems of traditional methods having a single response strategy, resulting in either insufficient prevention and control or excessive interference, thereby maximizing prevention and control effectiveness and minimizing negative impacts.

[0104] S205. According to the graded expulsion instructions, execute the expulsion combination operation to complete the handling of bird-related faults in the power grid.

[0105] The drive-away combination operation refers to the process of coordinating and controlling one or more drive-away devices, such as high-volume loudspeakers, ultrasonic loudspeakers, and multi-color strobe lights, to work together in a specific working mode (such as frequency, volume, and flashing pattern) according to the drive-away command.

[0106] Specifically, after receiving a tiered deterrence command, the device precisely controls each deterrence component: when the command is low-level, it executes a low-intensity, intermittent deterrence by activating an ultrasonic horn (working for 3 seconds, pausing for 10 seconds, frequency 30kHz, power 5W), resulting in low energy consumption and minimal interference; when the command is medium-level, it executes a combined deterrence by activating an ultrasonic horn (continuous operation), a multi-color strobe light flashing alternately in red and yellow (frequency 10Hz), and outputting an 80dB low-frequency warning sound, forming a combined audio-visual deterrent; when the command is high-level, it executes the most intense and comprehensive combined deterrence: sequentially activating the maximum-power ultrasonic wave, the rapid red / blue flashing strobe light, and the 100dB predator call horn, supplemented by safe laser spot firing when necessary. In particular, for persistent high-risk scenarios, a step-by-step escalation strategy (such as increasing laser power and volume every 30 seconds) is adopted to dynamically adjust the deterrence intensity, effectively breaking the birds' adaptability and significantly improving the success rate of deterring persistent bird damage.

[0107] This step achieves the dual goals of "precision strike" and "prevention of adaptation" by implementing a variety of dynamic and adjustable deterrence combinations that are precisely matched to the risk level. It solves the industry problem that existing deterrence methods are too simplistic and have fixed intensity, causing birds to quickly become accustomed to them and thus lose their effectiveness.

[0108] The power grid bird-related fault identification and removal method provided in this application constructs a complete "perception-cognition-decision-execution" intelligent closed loop through a series of collaborative technical means, including synchronous acquisition and preprocessing of multi-modal sensors, adaptive weighted fusion identification based on a CNN-LSTM hybrid network, intelligent decision-making combining a localized feature library and quantified risk thresholds, and dynamic hierarchical removal execution precisely matched with risk levels. This method achieves a fundamental shift in bird-related fault control from extensive to intelligent, precise, and adaptive control, effectively solving problems such as low monitoring fusion, insufficient identification accuracy, easy adaptation in removal, and unscientific response strategies in existing technologies. Ultimately, it achieves high identification accuracy (≥95%), high removal success rate (≥90%), and stable operation in complex electromagnetic environments, significantly reducing the power grid bird-related tripping rate.

[0109] Figure 3 This is a flowchart illustrating Embodiment Two of the method for identifying and repelling bird-related faults in power grids provided in this application. Figure 3 As shown, in Figure 2 Based on the embodiments, after executing a combined bird removal operation according to the tiered bird removal command to complete the handling of bird-related faults in the power grid, the method further includes:

[0110] S301. Collect target update data based on multi-source sensors.

[0111] Among them, target update data refers to the data collected immediately after a new round of monitoring is started after a round of expulsion operations are completed, which is used to evaluate the expulsion effect.

[0112] Specifically, after the bird removal operation in Example 1 is completed, the system containing the power grid bird-related fault handling device does not immediately enter sleep mode. Instead, it initiates a continuous secondary data acquisition for 30 seconds at a high frequency of 1 second per acquisition. The multimodal monitoring modules (radar, microphone, laser) work synchronously again to acquire real-time status data of the monitored area after the bird removal. This is equivalent to installing an "effect evaluator" for the system, ensuring that the prevention and control action is carried out from beginning to end, with execution and verification.

[0113] S302. Update the data according to the target and determine the fault identification result.

[0114] This step involves sending the target update data collected in S301 into the same fusion identification process as steps S202-S203 for rapid analysis to obtain a new fault identification result that reflects the current state. This result is used to directly determine whether the expulsion has taken effect.

[0115] S303. If the fault identification result indicates that the current fault is still at a medium or high level, the graded removal command is repeatedly executed until the fault identification result indicates that there is no fault.

[0116] This step implements closed-loop feedback and iterative processing. Specifically, if the secondary identification result is "no fault," it indicates that the decoy was successful, the target has left the danger zone, or the threat has been eliminated. At this time, the device automatically enters a low-power standby mode (such as extending the radar detection cycle to 10 seconds / time and shutting down other sensors), reducing system power consumption to ≤50mW, greatly extending the standby time of the device in the absence of sunlight, reflecting energy-saving design.

[0117] If the secondary identification result is still "medium risk" or "high risk", it indicates that the initial expulsion was not completely effective. The system will automatically repeat the expulsion process (starting from S204) and may automatically increase the expulsion intensity as needed (such as the high-risk continuous processing strategy in Example 1). The system is set to repeat a maximum of 3 times to avoid ineffective infinite loops.

[0118] If the fault persists after three attempts to remove the bird, it indicates an extremely persistent bird infestation that may exceed the device's automatic handling capabilities. In this case, the main control module will send a warning signal containing fault details, location coordinates, and real-time data to the power grid backend via the 4G network, and store the entire process data completely in the local storage unit. This provides maintenance personnel with accurate remote alarms and handling information, while also preserving a complete field data chain for in-depth post-event analysis to optimize bird removal strategies or model parameters.

[0119] This step, through a closed-loop mechanism of "repelling-verification-feedback-re-action," endows the system with the ability to self-verify, self-adjust, and report anomalies. It solves the problem of traditional devices "only repelling, not verifying," and the inability to guarantee effectiveness. It ensures the effectiveness of every prevention and control action and promptly reports anomalies that cannot be handled automatically, forming a highly efficient collaborative operation and maintenance model of "primarily device autonomous handling, supplemented by remote manual intervention," further improving the reliability and intelligence level of the entire power grid bird pest control system.

[0120] The power grid bird-related fault identification and removal method provided in this application, based on the intelligent processing flow of Embodiment 1, innovatively adds a closed-loop feedback mechanism based on secondary verification. This method performs high-frequency status verification immediately after removal and intelligently decides whether to enter low-power standby, iteratively strengthen removal, or report an early warning based on the verification results, thus achieving closed-loop and autonomous processing. This mechanism effectively ensures the effectiveness of a single removal action, significantly reduces ineffective energy consumption, and can promptly report complex situations outside the device's capability boundaries, thereby further improving the overall reliability, economy, and intelligent operation and maintenance level of power grid bird-related fault prevention and control.

[0121] Figure 4 This is a flowchart illustrating Embodiment 3 of the method for identifying and repelling bird-related faults in power grids provided in this application. Figure 4 As shown, in Figure 2 and Figure 3 Based on the previous embodiments, this embodiment proposes a method for preventing bird-related faults on the 500kV Linghai line of the Guangdong power grid. The specific process is as follows:

[0122] S401, Device initialization and parameter configuration.

[0123] Specifically, after the power grid bird-related fault handling device is powered on, the main control module automatically completes sensor calibration: the radar completes zero-point calibration through a preset targetless open area, the laser completes ranging calibration with the fixed bolts of the tower crossarm as the reference point, and the microphone automatically adjusts the gain according to the ambient noise. The total calibration time is 25 seconds. Through the Guangdong Power Grid operation and maintenance backend, the preset radar detection cycle is 1 second / time, the laser scanning cycle is 5 seconds / round, the acoustic signature collection time is 2 seconds / time, and the risk threshold adopts the preset level 3 standard.

[0124] S402, Synchronous acquisition and preprocessing of multimodal data.

[0125] Specifically, after the power grid bird-related fault handling device enters normal operation, the radar sensor detects moving objects within a 50-meter range in real time, collecting data on the target's distance, speed, and azimuth. The Kalman filter algorithm is used to remove noise from environmental vibrations and swaying leaves. The omnidirectional microphone simultaneously collects environmental acoustic signals, removes wind noise and electromagnetic interference through high-pass and low-pass filtering, and converts them into frequency domain spectral characteristics. The laser detector scans insulator strings, crossarms, and other parts at a cycle of 5 seconds per revolution, collecting the target's three-dimensional coordinates and contour data. Median filtering is used to remove interference from raindrops and dust.

[0126] S403, Multi-source data fusion identification and risk level determination.

[0127] Specifically, the fault identification unit inputs the preprocessed multi-source data into the CNN-LSTM hybrid network model, matches it with the feature database of birds in Guangdong, and outputs the identification results and confidence scores:

[0128] (1) When a magpie is identified to be standing on a crossarm 25 meters away from the conductor, the identification confidence level is 92%, which is considered a valid identification. Combined with the distance measurement data, it is considered a medium risk.

[0129] (2) When it is detected that crows are building a nest on the crossbeam, the nest outline size increases by 4cm and the distance from the wire is 8 meters after 3 consecutive data collections. The identification confidence is 95%, which is considered a valid identification. Combined with the distance measurement data, it is considered a high risk.

[0130] (3) When bird droppings are found to be piled up on the insulator string with an area of ​​15cm², the identification confidence level is 90%, and it is determined to be a valid identification. Combined with the area data, it is determined to be a medium risk.

[0131] S404, tiered expulsion execution and dynamic adjustment.

[0132] In this step, corresponding deportation strategies are implemented based on different risk levels:

[0133] (1) Medium-risk scenario (magpie is 25 meters away from the wire): Start the continuously working ultrasonic speaker (output power 8W, frequency 35kHz) + red / yellow alternating flashing light (frequency 10Hz), and at the same time, the high-frequency speaker outputs a high-frequency warning sound of 80dB. After 15 seconds of driving away, the magpie flies away.

[0134] (2) High-risk scenario (crows building nests): Activate the following sequence: 8W maximum power ultrasonic speaker → red / blue fast flashing light (frequency 20Hz) → 100dB eagle call → pulsed laser spotting; after 30 seconds, if the crows do not fly away, the laser power is increased to 75mW and the loudspeaker volume is increased to 105dB; after 1 minute, the crows fly away, and the drive-away is successful; during the process, when the laser irradiation area is less than 5 meters away from the wire, the laser power is automatically reduced to 50mW, and the laser beam angle is adjusted to ensure that the angle with the wire is greater than 30° to avoid damaging the wire.

[0135] S405, Secondary verification and closed-loop status feedback.

[0136] After the expulsion process is completed, the multimodal monitoring module will perform a secondary data collection and verification at a frequency of 1 time per second for 30 seconds.

[0137] (1) In the above two scenarios, the second verification is identified as fault-free, the expulsion is determined to be successful, the device enters low power standby mode, the radar detection cycle is extended to 10 seconds / time, the microphone and laser detector are turned off, and the standby power consumption is 45mW.

[0138] (2) If the fault is not resolved after 3 attempts to remove the fault, the main control module will send an early warning message to the Guangdong Power Grid Operation and Maintenance Backend via the 4G module. The message will include the fault type, tower number, latitude and longitude, real-time pictures and data. At the same time, the entire process data will be stored locally so that the operation and maintenance personnel can arrange on-site handling in a timely manner.

[0139] The method of this embodiment was piloted on the 500kV Linghai line of the Guangdong power grid for 6 months. The line's bird-related fault tripping rate decreased by 92% year-on-year, the bird removal success rate was ≥91%, the device online rate was 100%, and no faults caused by strong electromagnetic interference or high humidity environment occurred. The operation effect was excellent.

[0140] Figure 5 A schematic diagram of the structure of the power grid bird-related fault handling equipment provided in this application. Figure 5 As shown, the power grid bird-related fault handling device 50 provided in this embodiment includes at least one processor 501 and a memory 502. Optionally, the device 50 further includes a communication component 503. The processor 501, memory 502, and communication component 503 are connected via a bus 504.

[0141] In a specific implementation, at least one processor 501 executes computer execution instructions stored in memory 502, causing at least one processor 501 to perform the above-described method.

[0142] The specific implementation process of processor 501 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0143] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0144] The memory may include high-speed memory (Random Access Memory, RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0145] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0146] This application also provides a system for handling bird-related faults in power grids, including a computer program and a device for handling bird-related faults in power grids. When the computer program is executed by a processor, it implements the above-described method, and the device for handling bird-related faults in power grids is as described above.

[0147] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0148] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0149] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in application-specific integrated circuits (ASICs). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0150] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0151] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0152] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0153] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0154] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0155] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A method for identifying bird-related faults in power grids, characterized in that, include: Based on multi-source sensors, target data is acquired, including preprocessed temporal motion data, acoustic spectrum data, and spatial morphology data. Based on the machine learning model, the target data is weighted and fused to construct a target three-dimensional feature matrix; The target three-dimensional feature matrix is ​​matched with a preset fault feature library to determine the fault type and complete the identification and processing of bird-related faults in the power grid.

2. The identification method according to claim 1, characterized in that, After performing feature matching between the target three-dimensional feature matrix and a preset fault feature library to determine the fault type and complete the identification and processing of bird-related faults in the power grid, the method further includes: The fault identification confidence level is determined based on the fault type and the preset risk threshold. Based on the fault identification confidence level, the fault level is determined, including low level, medium level and high level; Based on the fault level, a graded expulsion command is generated, which includes a low-level expulsion command, a medium-level expulsion command, and a high-level expulsion command. According to the tiered bird removal instructions, execute the combined bird removal operation to complete the handling of bird-related faults in the power grid.

3. The identification method according to claim 2, characterized in that, The tiered expulsion commands include low-level expulsion commands, medium-level expulsion commands, and high-level expulsion commands; The low-level expulsion command is a low-intensity expulsion operation that activates the ultrasonic loudspeaker. The medium-level expulsion command is a combined expulsion operation that activates an ultrasonic horn, a multi-color strobe light, and a low-frequency warning sound. The high-level expulsion command is a combined expulsion operation that activates ultrasonic horns, multi-color strobe lights, low-frequency warning sounds, and laser spotting.

4. The identification method according to claim 2, characterized in that, After executing a combined bird removal operation according to the tiered bird removal command to complete the handling of bird-related faults in the power grid, the method further includes: Based on the multi-source sensors, collect target update data; Based on the target update data, the fault identification result is determined; If the fault identification result indicates that the current fault is still at a medium or high level, the graded removal command is executed repeatedly until the fault identification result indicates that there is no fault.

5. The identification method according to any one of claims 1 to 4, characterized in that, Based on multi-source sensors, target data is acquired, including: Identify the radar sensor, acoustic signature collector, and laser detector among the multi-source sensors; Based on the radar sensor, the target's distance, velocity, and azimuth time-series data are collected to obtain time-series motion data; Based on the aforementioned voiceprint collector, the voiceprint data of the target is collected to obtain voiceprint spectrum data; Based on the laser detector, the target's contour dimensions and three-dimensional coordinate data are collected to obtain spatial morphology data; The temporal motion data, the acoustic signature spectrum data, and the spatial morphology data are preprocessed respectively to obtain the target data.

6. A device for handling bird-related faults in power grids, characterized in that, include: The main unit is used for installation on transmission towers; A multimodal monitoring module, integrated on the main unit, is used to synchronously collect multi-source sensor data within a predetermined area of ​​the transmission tower. The multi-source sensor data includes at least temporal motion data, acoustic spectrum data, and spatial morphology data. The main control module, integrated within the main unit, is communicatively connected to the multimodal monitoring module. It is used to perform fusion analysis and identification based on the multi-source sensor data, determine the fault type and risk level, and generate graded expulsion commands according to the risk level. A graded expulsion module, integrated on the main unit and communicatively connected to the main control module, is used to receive the graded expulsion command and execute a combined expulsion operation that matches the risk level. A strong electromagnetic interference resistant power supply component is connected to the main unit and is used to provide power to the multimodal monitoring module, intelligent main control module and graded drive-off module. The strong electromagnetic interference resistant power supply component has an electromagnetic shielding structure.

7. The processing apparatus according to claim 6, characterized in that, The multimodal monitoring module includes a radar sensor, an acoustic signature collector, and a laser detector; The main control module includes a fault identification unit, a drive-off strategy unit, and a communication unit.

8. The processing apparatus according to claim 6, characterized in that, The fault identification unit is equipped with a machine learning model for constructing a three-dimensional feature matrix; The graded expulsion module is equipped with an ultrasonic speaker group and an audio-visual warning component. The ultrasonic speaker assembly is used to emit high-frequency ultrasonic waves. The sound and light warning component is used to emit audible warning sounds and visible warning lights.

9. A device for handling bird-related faults in power grids, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-5.

10. A system for handling bird-related faults in power grids, characterized in that, This includes computer programs and devices for handling bird-related faults in power grids; When the computer program is executed by the processor, it implements the method described in any one of claims 1-5; The device for handling bird-related faults in the power grid is the device as described in any one of claims 6 to 8.

Images