A kind of unmanned aerial vehicle-based power transmission line insulator zero value detection system and method

Abstract

The application relates to the technical field of power equipment detection, in particular to a power transmission line insulator zero value detection system and method based on a UAV, which comprises a pre-judgment and multi-source detection module, an edge node control and device storage module, a central control and remote control module and a system disaster recovery and auxiliary module; the application breaks through the traditional single inspection mode, carries out risk early warning through a fixed monitoring unit, guides the UAV to realize "on-demand precision inspection", and innovatively adopts an "edge-cloud" collaborative computing architecture, carries out real-time preprocessing and feature purification on multi-source data at the UAV end, effectively compresses the data volume, improves the response speed, finally fuses multi-source heterogeneous data in the cloud for deep intelligent diagnosis, and the whole-chain optimization and collaboration are realized from task generation, data acquisition to analysis and diagnosis, so that the overall efficiency, accuracy and real-time performance of the inspection operation are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of power equipment testing technology, specifically to a system and method for detecting zero values ​​of power transmission line insulators based on unmanned aerial vehicles (UAVs). Background Technology

[0002] To ensure power grid safety, regular and efficient insulator inspection is crucial. Inspection technology has evolved from traditional contact methods that require power outages and tower climbing, are inefficient, and susceptible to environmental influences, to optical non-contact methods (such as infrared and ultraviolet imaging) that, while allowing for long-distance operation with live power, have limited sensitivity and are easily affected by interference. In recent years, while drone inspections equipped with multiple sensors have improved efficiency and safety, their application remains limited by weather conditions, flight endurance, and strong electromagnetic interference. Furthermore, vision-based automatic identification algorithms still lack accuracy and real-time analysis capabilities in complex scenarios.

[0003] For example, a patent application with application number CN202511550322.8 and publication date of 20260127 based on RGDS... A method and system for detecting defects in power transmission line insulators using YOLO. This invention first obtains an initial dataset; then, it constructs RGDS based on the YOLOv11 framework. The YOLO model was then introduced; the DCSPConv module was then proposed; the C3kRFGSC module, based on the C3k2 module, was introduced again in the Backbone and Neck; and an improved DCSP module was introduced once more in the Neck. SDI module; then optimize the YOLOv11 model to obtain RGDS. The YOLO model is trained on the training set and its performance is evaluated on the validation set; finally, insulator defect images or videos from the test set are input into RGDS. The YOLO model outputs images or videos with defect bounding boxes showing the category and confidence level. This invention offers significant advantages in detection accuracy, real-time performance, lightweight model design, and adaptability.

[0004] Traditional zero-value detection of insulators in transmission lines focuses on optimizing a single aspect of the UAV platform itself or the back-end identification algorithm. It is usually based on fixed cycles or manual settings, lacks linkage with fixed online monitoring data, and cannot achieve risk warning and accurate task guidance based on real-time status. Secondly, the reliability of data acquisition and communication stability of UAVs in strong electromagnetic environments are insufficient, and their endurance and adaptability to complex weather conditions are limited, affecting the continuity of operations. Secondly, the real-time performance of data processing is poor, and analysis can only be performed after all data has been transmitted back, resulting in response delays. Furthermore, there is a lack of effective edge data purification and compression mechanisms. In addition, the fusion and utilization of multi-source heterogeneous data (such as infrared, ultraviolet and electric field) is insufficient, and the accuracy and generalization ability of the diagnostic model need to be improved. Finally, the systems generally lack a closed-loop optimization mechanism from intelligent diagnosis to operation and maintenance feedback, as well as a comprehensive disaster recovery and security management strategy to deal with communication interruptions, equipment failures, and severe weather. The overall robustness and intelligence level of the systems still need to be strengthened.

[0005] In view of this, there is an urgent need to design a zero-value detection system and method for power transmission line insulators based on UAVs to solve the above problems. Summary of the Invention

[0006] The purpose of this invention is to provide a zero-value detection system and method for power transmission line insulators based on unmanned aerial vehicles (UAVs) to address the aforementioned shortcomings in the prior art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A zero-value detection system for power transmission line insulators based on unmanned aerial vehicles (UAVs) includes a pre-judgment and multi-source detection module, an edge node control and equipment storage module, a central control and remote control module, and a system disaster recovery and auxiliary module. The pre-judgment and multi-source detection module includes a fixed monitoring and early warning unit, a UAV multi-source collaborative detection unit, a meteorological adaptation unit, and an electromagnetic interference suppression unit, wherein: The fixed monitoring and early warning unit is installed on the transmission line towers / conductors to collect leakage current, meteorological and appearance data, and generate early warning coordinates containing location and risk level through local analysis; The fixed monitoring and early warning unit integrates a leakage current sensor, a micro weather station, and a low-power visible light camera, wherein: The leakage current sensor has a detection range of 0-10mA and an accuracy of ±0.1mA. The micro-weather station is used to collect temperature, humidity, wind speed and air pressure data, with accuracies of ±0.5℃, ±2%RH and ±0.1m / s, respectively. The low-power visible light camera has a resolution of 1080P and a frame rate of 25fps. The fixed monitoring and early warning unit uploads the collected data via LoRa or 4G / 5G network and performs local analysis using a lightweight random forest algorithm to achieve early warning of abnormal leakage current of insulators. The early warning information includes tower number, insulator string location and risk level. The UAV multi-source collaborative detection unit is mounted on the UAV and integrates multiple sensors to synchronously collect data on the appearance, temperature, discharge and electric field of the insulator based on the warning coordinates or flight path. The UAV multi-source collaborative detection unit integrates multiple types of sensors that are triggered synchronously, specifically including: High-definition visible light cameras, with a resolution of no less than 4K and a frame rate of no less than 30fps, are used to capture surface defects of insulators. Infrared thermal imagers, with a temperature measurement range of -20℃ to 150℃ and an accuracy of ±0.5℃, are used to detect the Joule thermal temperature difference generated by leakage current in zero-value insulators. The ultraviolet imager, with a detection band of 240nm-280nm and a sensitivity of ≤5pC, is used to identify early corona discharge in insulators. The electric field detection device has a range of 0-50kV / m and an accuracy of ±1%. The probe of the electric field detection device is physically isolated from the drone body to ensure a vertical distance of not less than 2.5 meters and a horizontal distance of not less than 0.2 meters, so as to avoid interference from the drone carrier to the electric field measurement. The meteorological adaptation unit is used to monitor wind speed, humidity and visibility, and automatically adjusts or suspends the detection task when the limits are exceeded. The electromagnetic interference suppression unit employs metal shielding and preliminary filtering to suppress power frequency electric fields and harmonic interference. The edge node control and device storage module includes an edge computing and anti-interference unit, a UAV swarm collaborative control unit, and a local storage and status management unit, wherein: The edge computing and anti-interference unit is deployed on the UAV and fixed monitoring terminal to perform real-time preprocessing and feature extraction of raw data, and to filter out electromagnetic interference and compensate for sensor deviations. The edge computing and anti-interference unit is mounted on the edge node and is used to perform real-time preprocessing and anti-interference processing on the raw detection data, wherein: The preprocessing includes: A neighborhood wavelet coefficient algorithm with an improved threshold is used to denoise infrared thermal image data, and an improved differential evolution algorithm with the Otsu function as the evaluation function is used to segment the insulator target region, with a segmentation accuracy of >95%. The anti-interference processing includes an adaptive filtering algorithm and a reinforcement learning algorithm based on Win-or-Fast Learning Strategy (WoLF-PHC) to filter out power frequency electromagnetic interference and dynamically optimize the communication channel. The edge computing and anti-interference unit is also used to extract multi-dimensional features, including insulator temperature variance, maximum temperature difference and electric field distortion rate, from the preprocessed data, and to perform zero-value probability screening using a lightweight BP neural network with less than 100k parameters, with a screening accuracy of >90%, thereby compressing redundant data. The drone swarm collaborative control unit receives instructions to control the drones to fly autonomously and work collaboratively, and supports docking with mobile platforms or automated airports to achieve endurance management. The local storage and status management unit adopts a fault-tolerant storage architecture, caches critical data and logs, supports resume transmission after network outage, and monitors device status to trigger alarms. The central control and remote control module includes an intelligent task scheduling unit, a multi-source data fusion intelligent diagnostic unit, a visualization and operation and maintenance decision-making unit, and a global data management unit, wherein: The intelligent task scheduling unit is used to plan UAV inspection routes, schedule multi-UAV collaborative or relay inspections, and dynamically allocate task priorities. The multi-source data fusion intelligent diagnostic unit is used to fuse multi-source features and historical data, and to achieve accurate identification and defect level classification of zero-value insulators through intelligent analysis models. The multi-source data fusion intelligent diagnostic unit is used to fuse and deeply analyze multi-source feature data from edge nodes and fixed monitoring units; It employs a fusion analysis of the XGBoost model and an improved BP neural network model. The improved BP neural network has an input layer of 16 nodes and a hidden layer of 8 nodes. The multi-source data fusion intelligent diagnostic unit can output the judgment result of zero-value insulators, defect level and deterioration trend analysis, and the system's comprehensive identification accuracy is not less than 95%. The visualization and operation and maintenance decision-making unit is used to display the lines, drones and defect information in a map, and automatically generate inspection reports and maintenance work orders; The global data management unit is used to centrally store all system data and model parameters, and supports data retrieval, archiving and model updates; The system disaster recovery and support module includes a communication link disaster recovery unit, a power and mission disaster recovery unit, a meteorological and airspace safety decision-making unit, and a data security and operation and maintenance closed-loop unit, wherein: The communication link disaster recovery unit uses multiple communication methods to form redundant links and applies intelligent anti-interference technology to optimize transmission, ensuring the stability and reliability of data transmission in a strong electromagnetic environment. The communication link disaster recovery unit uses at least two of the following to form a primary and backup redundant link: 5G network, fiber optic network, and satellite communication. 5G links are used for high-bandwidth, low-latency data interaction between drones and edge nodes, while fiber optic links are used for massive data backhaul between edge nodes and the cloud. When the primary link is interrupted, the system can automatically switch to the backup link within 30 seconds; Meanwhile, a reinforcement learning algorithm based on WoLF-PHC is applied to dynamically select and optimize the transmission channel to ensure communication reliability in a strong electromagnetic environment. The power and mission disaster recovery unit provides a hybrid power supply and endurance guarantee mechanism for monitoring equipment and drones, supporting continuous and relay operations of drones; The power and mission disaster recovery unit includes: A hybrid power supply system combining solar panels and batteries is configured for fixed monitoring and early warning units; The drone is equipped with a high-energy-density lithium battery that can be quickly replaced, as well as a vehicle-mounted mobile charging pod or ground-based automated airport that supports drone relay. With the support of the system, a single drone can perform effective inspections for no less than 60 minutes. The meteorological and airspace safety decision-making unit integrates meteorological and airspace information to achieve automatic task avoidance and safety threshold judgment; The meteorological and airspace safety decision-making unit integrates real-time meteorological and airspace data and presets safety thresholds, wherein: When any of the following conditions are detected: wind speed ≥8m / s, humidity ≥85%, or visibility <500m, or when the drone enters a no-fly zone, the unit automatically triggers a command to control the drone to suspend its mission, return to base, or execute an emergency landing protection procedure, and pushes a warning message to the operation and maintenance terminal. The data security and operation and maintenance closed-loop unit manages the security of critical system data and optimizes system algorithms and strategies through a feedback mechanism. The data security and operation and maintenance closed-loop unit uses encrypted transmission and off-site disaster recovery backup mechanisms to ensure data security. Simultaneously, it can transmit the maintenance results data fed back by on-site operation and maintenance personnel back to the central control and remote control module, which is used to jointly optimize and iteratively update the prediction algorithm of the fixed monitoring and early warning unit, the identification model of the multi-source data fusion intelligent diagnosis unit, and the compensation parameters of the edge computing and anti-interference unit.

[0008] A method for detecting zero values ​​in power transmission line insulators based on unmanned aerial vehicles (UAVs) includes the following steps: Step 1. Risk Prediction and Intelligent Task Planning: Real-time data of the route is collected through a fixed monitoring network, and local analysis is performed using intelligent algorithms to generate early warnings. The cloud platform integrates early warnings, historical data and meteorological information, identifies high-risk points through predictive models, plans precise flight paths and detection strategies for UAVs, and forms detection tasks. Step 2. UAV Adaptive Anti-interference Data Acquisition: The UAV, equipped with multiple sensors, flies to the target and adaptively selects sensor combinations for synchronous data acquisition based on the environment. Through physical shielding and anti-interference communication technology, it ensures stable and accurate data acquisition in strong electromagnetic environments. Step 3. Real-time edge processing and feature purification: The raw data is preprocessed in real time on the UAV, including image denoising, target segmentation and key feature extraction, and a lightweight model is used for initial screening of zero-value probabilities. The data is then compressed and sent back. Step 4. Multi-source data fusion and intelligent diagnosis: The cloud platform integrates edge data and historical information, filters key features, and uses intelligent models to perform fusion analysis to achieve accurate identification, grade classification, and diagnostic report generation for zero-value insulators; Step 5. Visualized Positioning and Closed-Loop Operation and Maintenance Management: The diagnostic results are combined with precise positioning and displayed on an electronic map for visualization. Repair work orders are automatically generated and pushed to the operation and maintenance system. On-site handling results are fed back to the platform, forming a management closed loop. Step 6. Model Iteration and System Disaster Recovery Optimization: The diagnostic and prediction models are continuously optimized using on-site feedback data. The system also has communication redundancy and emergency mechanisms to ensure the safety and data integrity of the UAV in adverse conditions or failures.

[0009] In the above technical solution, the present invention provides a zero-value detection system and method for transmission line insulators based on unmanned aerial vehicles (UAVs), which has the following advantages: (1) This invention constructs a full-process intelligent inspection system of "precise early warning - edge purification - intelligent diagnosis", which breaks through the traditional single inspection mode. It conducts risk early warning through fixed monitoring units, guides drones to achieve "precise inspection on demand", and innovatively adopts the "edge-cloud" collaborative computing architecture to perform real-time preprocessing and feature purification of multi-source data on the drone end, effectively compressing the amount of data and improving the response speed. Finally, it integrates multi-source heterogeneous data in the cloud for deep intelligent diagnosis. This system realizes the optimization and collaboration of the entire chain from task generation, data collection to analysis and diagnosis, which significantly improves the overall efficiency, accuracy and real-time performance of inspection operations.

[0010] (2) This invention realizes the deep fusion of multi-dimensional information, which greatly improves the accuracy and reliability of defect identification. In view of the complexity of zero-value insulator detection, the system comprehensively uses multi-source heterogeneous data such as leakage current, infrared thermography, ultraviolet discharge, electric field distribution and visible light images, and adopts an advanced fusion diagnostic model for comprehensive judgment. This multi-dimensional information complementarity mechanism overcomes the limitations of a single sensor, greatly improves the accuracy of zero-value insulator identification and the generalization ability of judgment of its defect level, and makes the diagnostic results more comprehensive and reliable.

[0011] (3) The present invention has high environmental adaptability, system robustness and continuous self-optimization capability. The system is designed with multiple fault tolerance and security guarantee mechanisms from hardware to software. It ensures reliable data acquisition in strong electromagnetic environment through physical isolation, electromagnetic shielding and intelligent anti-interference algorithm. It ensures stable information transmission through multi-mode redundant communication and fast self-healing link. It copes with complex outdoor working conditions through hybrid power supply, continuous relay and meteorological airspace safety decision. At the same time, the system establishes operation and maintenance feedback closed loop and continuously optimizes early warning and diagnostic models by using on-site maintenance results. These designs together ensure the reliable operation of the system in complex and harsh environment and the self-improvement of long-term performance, and the overall robustness is significantly enhanced. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0013] Figure 1 This is a schematic diagram of the system frame structure provided in an embodiment of the zero-value detection system and method for power transmission line insulators based on unmanned aerial vehicles (UAVs) of the present invention. Figure 2 This is a flowchart illustrating an embodiment of a method for detecting zero values ​​of power transmission line insulators based on unmanned aerial vehicles (UAVs). Detailed Implementation

[0014] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0015] like Figure 1 As shown in the figure, the embodiment of the present invention provides a zero-value detection system for power transmission line insulators based on UAVs, including a pre-judgment and multi-source detection module, an edge node control and device storage module, a central control and remote control module, and a system disaster recovery and auxiliary module; The pre-judgment and multi-source detection module includes a fixed monitoring and early warning unit, a UAV multi-source collaborative detection unit, a meteorological adaptation unit, and an electromagnetic interference suppression unit, wherein: The fixed monitoring and early warning unit is fixedly deployed on the transmission line towers or conductors. It integrates a leakage current sensor, a micro weather station, and a low-power visible light camera to collect leakage current, environmental weather, and appearance image data related to the insulator's operating status. The unit uploads the collected data through LoRa or 4G / 5G networks and performs real-time analysis locally using a lightweight random forest algorithm to generate early warning coordinate information that includes the specific tower number, insulator string location, and risk level, providing target guidance for the precise inspection of drones. The UAV multi-source collaborative detection unit is the core mobile detection platform of the system. It integrates a high-definition visible light camera, an infrared thermal imager, an ultraviolet imager, and an electric field detection device that can be triggered synchronously. These sensors synchronously collect multi-dimensional data such as the appearance, temperature, discharge, and electric field distribution of the insulator based on the warning coordinates or preset flight path. It needs to be clarified that: The UAV multi-source collaborative detection unit is equipped with a high-definition visible light camera with a resolution of no less than 4K and a frame rate of no less than 30fps, which can clearly capture appearance defects such as cracks, damage, and flashover marks on the surface of insulators. The infrared thermal imager has a temperature measurement range of -20℃ to 150℃ and an accuracy of ±0.5℃, which can accurately detect local abnormal temperature rise (Joule heating effect) caused by the decrease in zero-value insulator resistance and the increase in leakage current. The ultraviolet imager has a detection band of 240nm-280nm and a sensitivity of ≤5pC, which is used to identify early and weak corona discharge phenomena caused by electric field distortion on the surface of insulators. The electric field detection device has a range of 0-50kV / m and an accuracy of ±1%. Its probe is kept at a vertical distance of no less than 2.5 meters and a horizontal distance of no less than 0.2 meters from the UAV body through methods such as extending the pod. This physical isolation structure can minimize the interference of the UAV itself on the spatial electric field measurement and ensure the accuracy of the electric field distortion rate data. The meteorological adaptation unit monitors key meteorological parameters such as wind speed, humidity and visibility in the work area in real time. When the parameters exceed the preset safety threshold, the unit automatically sends instructions to the system to adjust the UAV's flight attitude, detection strategy or suspend the mission to ensure flight safety and data validity. The electromagnetic interference suppression unit addresses the issue from a hardware perspective by designing a metal shielding shell for the sensitive sensors (especially electric field detection devices) carried by the UAV and performing preliminary filtering to effectively suppress the interference of the strong power frequency electric field and its harmonics generated by the power transmission line on the measurement data.

[0016] The edge node control and device storage module includes an edge computing and anti-interference unit, a UAV swarm collaborative control unit, and a local storage and status management unit, wherein: The edge computing and anti-interference unit is deployed on the UAV's onboard computer and fixed monitoring equipment. Its core function is to perform real-time preprocessing and feature extraction on the massive amounts of raw data collected by sensors. For example, it uses a neighborhood wavelet coefficient algorithm with improved thresholds to denoise infrared thermal image data, and uses an improved differential evolution algorithm with the Otsu function as the evaluation function to accurately segment the insulator target area with a segmentation accuracy of >95%. At the same time, the unit also integrates an adaptive filtering algorithm and a reinforcement learning algorithm based on Win-or-Fast Learning Strategy (WoLF-PHC) to dynamically filter out electromagnetic interference and optimize the communication channel. After preprocessing, the unit extracts multi-dimensional features such as insulator temperature variance, maximum temperature difference, and electric field distortion rate, and uses a lightweight BP neural network model with fewer than 100k parameters to quickly screen for zero-value probabilities (accuracy >90%), thereby achieving great data compression. Only key features and screening results are sent back, significantly reducing communication bandwidth requirements and cloud processing pressure. It should be noted that: The edge computing and anti-interference unit uses an improved differential evolution algorithm for image segmentation. It uses the maximization of inter-class variance (Otsu function) as the evaluation function to guide population evolution, thereby quickly and accurately segmenting the insulator target region in infrared or visible light images, laying the foundation for subsequent feature extraction. The WoLF-PHC reinforcement learning algorithm used enables edge nodes to learn the optimal channel selection and power control strategy in complex electromagnetic spectrum environments through continuous interaction with the environment, dynamically avoiding interference and improving the quality of communication links. The drone swarm collaborative control unit receives instructions from the central control module and is responsible for controlling the autonomous take-off, flight path, hovering detection, collaborative operation and automatic return of one or more drones. The unit also supports precise docking of drones with vehicle-mounted mobile charging cabins or ground-based automatic airports to achieve rapid battery replacement or charging, so as to complete the long-distance relay inspection. The local storage and status management unit adopts a storage architecture with fault tolerance mechanism to cache the raw data collected by the drone, the feature data after edge processing, the flight logs and the device status information. This unit supports caching data when the network is temporarily interrupted and resumes the transmission after the network is restored. At the same time, it continuously monitors the drone's battery level, sensor status, etc., and triggers local alarms when abnormalities occur.

[0017] The central control and remote control module includes an intelligent task scheduling unit, a multi-source data fusion intelligent diagnostic unit, a visualization and operation and maintenance decision-making unit, and a global data management unit, wherein: The intelligent task scheduling unit is the "command center" of the system. It integrates the early warning information reported by the fixed monitoring and early warning unit, historical defect data, real-time meteorological information and operation and maintenance plan, uses the prediction model to identify high-risk sections, and then plans the optimal drone inspection route. It dynamically schedules multiple drones to carry out collaborative or relay inspections and can dynamically allocate priorities according to the urgency of the task. The multi-source data fusion intelligent diagnostic unit is the "intelligent brain" of the system. It receives purified multi-source feature data uploaded from various edge nodes and merges it with the historical database. This unit uses the XGBoost model and an improved BP neural network (16 nodes in the input layer and 8 nodes in the hidden layer) for fusion analysis, realizing accurate identification of zero-value insulators, classification of defect levels (such as slight, moderate, and severe), and analysis of deterioration trends. The overall identification accuracy of the system is no less than 95%. It should be noted that: The improved BP neural network used in the multi-source data fusion intelligent diagnostic unit has a simplified and optimized structure (16-8-output nodes), which reduces computational complexity while ensuring diagnostic accuracy. It is suitable for parallel or cascade fusion with the XGBoost model. The XGBoost model is good at processing structured features and capturing nonlinear relationships, while the BP neural network has advantages in feature abstraction. The fusion of the two can give full play to their respective characteristics and improve the comprehensive judgment ability of complex and fuzzy defect patterns. The visualization and operation and maintenance decision-making unit integrates diagnostic results, real-time drone location, flight path, and tower information onto an electronic map for unified display, providing operation and maintenance personnel with an intuitive global situational awareness. This unit can also automatically generate structured inspection reports and maintenance work orders, and push them to the corresponding maintenance teams through the operation and maintenance system. The global data management unit serves as the system's data center, centrally storing all system operation data, diagnostic model parameters, and configuration information. It supports rapid data retrieval, long-term archiving, and online updates and version management of diagnostic models.

[0018] The system disaster recovery and support module includes a communication link disaster recovery unit, a power and mission disaster recovery unit, a meteorological and airspace safety decision-making unit, and a data security and operation and maintenance closed-loop unit, wherein: The communication link disaster recovery unit uses at least two of the following methods to construct a primary and backup redundant communication link: 5G, fiber optic and satellite communication. Typically, the 5G link is used for high-bandwidth, low-latency interaction between the drone and the edge node, and the fiber optic is used for massive data backhaul from the edge node to the cloud. When the primary link is interrupted due to strong electromagnetic interference or physical blockage, the system can automatically and seamlessly switch to the backup link within 30 seconds to ensure the continuous transmission of commands and data. At the same time, the reinforcement learning algorithm based on WoLF-PHC is applied to intelligently select and optimize the transmission channel to actively adapt to the complex electromagnetic environment. It should be noted that: The reinforcement learning algorithm based on WoLF-PHC applied in the communication link disaster recovery unit has a "win or fast learning" mechanism that enables the algorithm to adopt a conservative strategy to consolidate its achievements when it is in an advantageous position, and to adopt a fast learning strategy to explore new solutions when it is in a disadvantageous position. This characteristic is very suitable for communication anti-interference decision-making in dynamic and adversarial strong electromagnetic environments, and can effectively improve the survivability and transmission efficiency of the link under harsh conditions. The power and mission disaster recovery unit provides continuous energy security for the system: it equips the fixed monitoring and early warning unit with a hybrid power supply system consisting of solar panels and batteries; it equips the drones with high-energy-density lithium batteries that can be quickly replaced, and supports drones to carry out relay flights by relying on vehicle-mounted mobile charging cabins or automatic airport networks, ensuring that the flight time for a single effective inspection is not less than 60 minutes. It should be noted that: The vehicle-mounted mobile charging cabin or ground-based automated airport in the power and mission disaster recovery unit is not only a charging point, but also an intelligent relay station and warehouse. They are usually deployed at key nodes along the power transmission line and can be autonomously dispatched to designated locations according to mission requirements, or receive drones with insufficient power after completing the mission, and perform automatic charging / battery swapping, data export and simple maintenance. They are the infrastructure for realizing a large-scale, long-distance, automated drone inspection network. The meteorological and airspace safety decision-making unit integrates real-time updated meteorological and airspace monitoring information and presets safety thresholds (such as wind speed ≥8m / s, humidity ≥85%, visibility <500m or entering a no-fly zone). Once the monitored value exceeds the limit, the unit will immediately and automatically trigger an instruction to control the drone to perform protection procedures such as pausing, returning to home or forced landing, and push an early warning to the background. It should be noted that: The meteorological and airspace safety decision-making unit accesses real-time data from authoritative meteorological service departments and airspace management departments through API interfaces, and combines it with UAV onboard sensor data to perform multi-source information verification. Its safety threshold judgment logic adopts multi-layer condition judgment. For example, when encountering sudden thunderstorm weather, the system will not only consider wind speed and humidity, but also combine lightning warning information to execute the highest priority emergency avoidance and return command. The data security and operation and maintenance closed-loop unit uses encrypted transmission and off-site disaster recovery backup mechanisms to ensure data security. More importantly, this unit transmits the actual inspection results (such as whether it is a real defect, the type of defect, etc.) fed back by the on-site operation and maintenance personnel to the central control module. Using this feedback data, the predictive algorithm of the fixed monitoring and early warning unit, the identification model of the multi-source data fusion intelligent diagnosis unit, and the sensor compensation parameters of the edge computing unit are continuously optimized and iteratively updated, thereby forming a complete operation and maintenance closed loop of "detection-diagnosis-repair-optimization" and driving the continuous improvement of the system's intelligence level. It should be noted that the model iteration update mechanism of the data security and operation and maintenance closed-loop unit can adopt incremental learning or periodic full training. After receiving a batch of valid on-site feedback data each time, the system will start a low-priority background training task to fine-tune or retrain the existing model in an isolated environment. After verification, it will be gradually updated to the production environment in a gray-scale manner to ensure that the system's diagnostic capabilities can continue to evolve with the accumulation of operation and maintenance experience. The leakage current sensor integrated into the fixed monitoring and early warning unit has a preferred detection range of 0-10mA and an accuracy of ±0.1mA. It can sensitively capture minute changes in leakage current caused by factors such as dirt and moisture on the insulator surface. The temperature, humidity, wind speed, and air pressure measurement accuracy of the micro-weather station reach ±0.5℃, ±2%RH, and ±0.1m / s, respectively, providing accurate environmental parameters for the local early warning algorithm. The low-power visible light camera has a resolution of 1080P and a frame rate of 25fps, which is sufficient to meet the daily appearance monitoring needs. The lightweight random forest algorithm for local analysis can run efficiently on resource-limited embedded devices, enabling rapid preliminary judgment of risks such as abnormal leakage current.

[0019] like Figure 2 The method shown includes the following steps for detecting zero values ​​in power transmission line insulators based on unmanned aerial vehicles (UAVs): Step 1. Risk Prediction and Intelligent Task Planning: The fixed monitoring and early warning network deployed on the poles continuously collects leakage current, meteorological, and appearance data of the lines, and performs real-time analysis locally using a lightweight random forest algorithm to generate preliminary early warnings. The intelligent task scheduling unit of the central control platform gathers these early warning information and combines them with historical defect databases, weather forecasts, and power grid operation plans. It uses predictive models (such as time series forecasts and risk probability models) for comprehensive analysis to identify high-risk sections and specific locations of insulator zero-value defects in the near future. Based on this, the unit automatically plans the optimal precision inspection route for the UAV, determines the priority of the inspection task, and generates a detailed inspection task instruction package containing information such as target coordinates, sensor operating mode, and flight path. Step 2. UAV Adaptive Anti-interference Data Acquisition: After receiving the mission command, the UAV swarm collaborative control unit controls the UAVs to autonomously fly to the target area. During flight and hovering detection, the UAV multi-source collaborative detection unit adaptively selects the best sensor combination (e.g., focusing on visible light and infrared under strong light, and focusing on ultraviolet and electric field in cloudy and humid weather) based on mission requirements and current environmental information such as light intensity and electromagnetic intensity for synchronous triggering and acquisition. Throughout the process, the electromagnetic interference suppression unit provides protection for sensitive sensors through hardware shielding, while the communication system uses anti-interference algorithms to dynamically maintain link stability, ensuring that various sensor data can be stably, accurately, and synchronously acquired in the strong electromagnetic field environment of power transmission lines. Step 3. Real-time edge processing and feature purification: While the UAV is collecting data, the onboard edge computing and anti-interference unit immediately starts the real-time processing flow. First, the original image and video data are preprocessed by denoising and enhancement. In particular, the infrared thermal image data is denoised using an improved wavelet algorithm, and the insulator region is accurately segmented using an improved differential evolution algorithm. Then, key multi-dimensional features such as temperature statistical features (such as mean, variance, and maximum temperature difference), electric field distortion features, and ultraviolet photon counts are quickly extracted from the processed data. Next, a lightweight BP neural network model is used to quickly calculate these features and give a preliminary zero-value probability judgment. Finally, the purified and condensed feature set, the initial screening results, and the necessary metadata (time, location) are packaged and transmitted back to the cloud through an anti-interference communication link. The original massive data is cached locally or discarded as needed, which greatly reduces the transmission burden. Step 4. Multi-source data fusion and intelligent diagnosis: After receiving purified feature data from different drones and fixed monitoring points, the cloud-based multi-source data fusion intelligent diagnosis unit first performs spatiotemporal alignment and data fusion to form a multi-dimensional feature vector of the target insulator. Subsequently, this feature vector is compared with historical data records of insulators of the same tower and type. The fused data is sent to the fusion diagnosis model composed of XGBoost and improved BP neural network for in-depth analysis. The model comprehensively evaluates various feature indicators and finally outputs a definitive judgment on whether the insulator has a zero value, the specific defect level (such as I, II, III), and a deterioration trend analysis report based on historical data. After the diagnosis report is automatically generated, it is pushed to the visualization unit. Step 5. Visualized Positioning and Closed-Loop Operation and Maintenance Management: The visualization and operation and maintenance decision-making unit integrates the diagnostic results generated in Step 4 with information such as high-precision maps, tower coordinates, and real-time drone locations. On the electronic map, different colored icons (such as green for normal, yellow for warning, and red for defect) are used to visually display the health status of the line and the precise location of defect points. At the same time, the system automatically generates structured inspection reports and maintenance work orders containing defect locations, levels, and suggested handling measures. These are directly dispatched to relevant operation and maintenance teams through the Production Management System (PMS) or mobile APP. After completing the maintenance according to the work order, the on-site operation and maintenance personnel need to feed back the actual maintenance situation (confirmation of defects, handling methods, replacement of parts, etc.) to the system through the mobile terminal. Step Six. Model Iteration and System Disaster Recovery Optimization: The data security and operation and maintenance closed-loop unit receives feedback data from the field and stores it as valuable labeled samples in the global database. The system periodically or triggeredly uses these new samples to jointly train and optimize the local early warning model of the fixed monitoring and early warning unit, the lightweight initial screening model at the edge, and the multi-source fusion diagnostic model in the cloud, continuously improving the accuracy and adaptability of the algorithms at each stage. Throughout the entire method execution process, the system disaster recovery and auxiliary modules provide full protection: the communication link disaster recovery unit ensures uninterrupted data transmission; the power and mission disaster recovery unit ensures continuous operation of the UAV; and the meteorological and airspace safety decision unit automatically intervenes when adverse conditions occur, directing the UAV to safely evacuate or take protective measures, thereby ensuring that the entire detection process can operate safely, reliably, and efficiently or exit in an orderly manner under any circumstances.

[0020] Working Principle: The system establishes an integrated terrestrial and ground-based sensing network through "fixed monitoring + UAV mobile detection." Fixed monitoring points enable wide-area, continuous preliminary screening and risk warning, providing UAVs with precise target acquisition. UAVs, as flexible execution units, carry multiple sensors to acquire high-quality, multi-dimensional detailed data. The system innovatively adopts an "edge-cloud" collaborative computing paradigm, delegating the preprocessing, feature extraction, and initial screening of large amounts of raw data to the edge of the UAV, significantly reducing data backhaul volume and cloud processing latency, and improving system real-time performance. The cloud focuses on complex data fusion and deep intelligent diagnosis, utilizing more powerful computing power and more comprehensive historical data to make accurate judgments. The entire system achieves automated operation through intelligent task scheduling and ensures long-term stable, reliable, and efficient operation in the complex and harsh power grid inspection environment through multiple disaster recovery designs in communication, power, and security, as well as a closed-loop optimization mechanism from operation and maintenance feedback to model iteration, and possesses the ability to continuously improve itself.

[0021] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A UAV-based zero-value detection system for transmission line insulators, comprising a pre-judgment and multi-source detection module, an edge node control and equipment storage module, a central control and remote control module, and a system disaster recovery and auxiliary module, characterized in that: The pre-judgment and multi-source detection module includes a fixed monitoring and early warning unit, a UAV multi-source collaborative detection unit, a meteorological adaptation unit, and an electromagnetic interference suppression unit, wherein: The fixed monitoring and early warning unit is installed on the transmission line towers / conductors to collect leakage current, meteorological and appearance data, and generate early warning coordinates containing location and risk level through local analysis; The UAV multi-source collaborative detection unit is mounted on the UAV and integrates multiple sensors to synchronously collect data on the appearance, temperature, discharge and electric field of the insulator based on the warning coordinates or flight path. The meteorological adaptation unit is used to monitor wind speed, humidity and visibility, and automatically adjusts or suspends the detection task when the limits are exceeded. The electromagnetic interference suppression unit employs metal shielding and preliminary filtering to suppress power frequency electric fields and harmonic interference. The edge node control and device storage module includes an edge computing and anti-interference unit, a UAV swarm collaborative control unit, and a local storage and status management unit, wherein: The edge computing and anti-interference unit is deployed on the UAV and fixed monitoring terminal to perform real-time preprocessing and feature extraction of raw data, and to filter out electromagnetic interference and compensate for sensor deviations. The drone swarm collaborative control unit receives instructions to control the drones to fly autonomously and work collaboratively, and supports docking with mobile platforms or automated airports to achieve endurance management. The local storage and status management unit adopts a fault-tolerant storage architecture, caches critical data and logs, supports resume transmission after network outage, and monitors device status to trigger alarms. The central control and remote control module includes an intelligent task scheduling unit, a multi-source data fusion intelligent diagnostic unit, a visualization and operation and maintenance decision-making unit, and a global data management unit, wherein: The intelligent task scheduling unit is used to plan UAV inspection routes, schedule multi-UAV collaborative or relay inspections, and dynamically allocate task priorities. The multi-source data fusion intelligent diagnostic unit is used to fuse multi-source features and historical data, and to achieve accurate identification and defect level classification of zero-value insulators through intelligent analysis models. The visualization and operation and maintenance decision-making unit is used to display the lines, drones and defect information in a map, and automatically generate inspection reports and maintenance work orders; The global data management unit is used to centrally store all system data and model parameters, and supports data retrieval, archiving and model updates; The system disaster recovery and support module includes a communication link disaster recovery unit, a power and mission disaster recovery unit, a meteorological and airspace safety decision-making unit, and a data security and operation and maintenance closed-loop unit, wherein: The communication link disaster recovery unit uses multiple communication methods to form redundant links and applies intelligent anti-interference technology to optimize transmission, ensuring the stability and reliability of data transmission in a strong electromagnetic environment. The power and mission disaster recovery unit provides a hybrid power supply and endurance guarantee mechanism for monitoring equipment and drones, supporting continuous and relay operations of drones; The meteorological and airspace safety decision-making unit integrates meteorological and airspace information to achieve automatic task avoidance and safety threshold judgment; The data security and operation and maintenance closed-loop unit performs security management of critical system data and optimizes system algorithms and strategies through a feedback mechanism.

2. The zero-value detection system for transmission line insulators based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The fixed monitoring and early warning unit integrates a leakage current sensor, a micro weather station, and a low-power visible light camera, wherein: The leakage current sensor has a detection range of 0-10mA and an accuracy of ±0.1mA. The micro-weather station is used to collect temperature, humidity, wind speed and air pressure data, with accuracies of ±0.5℃, ±2%RH and ±0.1m / s, respectively. The low-power visible light camera has a resolution of 1080P and a frame rate of 25fps. The fixed monitoring and early warning unit uploads the collected data via LoRa or 4G / 5G networks and performs local analysis using a lightweight random forest algorithm to achieve early warning of abnormal insulator leakage current. The early warning information includes the tower number, insulator string location, and risk level.

3. The zero-value detection system for transmission line insulators based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The UAV multi-source collaborative detection unit integrates multiple types of sensors that are triggered synchronously, specifically including: High-definition visible light cameras, with a resolution of no less than 4K and a frame rate of no less than 30fps, are used to capture surface defects of insulators. Infrared thermal imagers, with a temperature measurement range of -20℃ to 150℃ and an accuracy of ±0.5℃, are used to detect the Joule thermal temperature difference generated by leakage current in zero-value insulators. The ultraviolet imager, with a detection band of 240nm-280nm and a sensitivity of ≤5pC, is used to identify early corona discharge in insulators. The electric field detection device has a range of 0-50kV / m and an accuracy of ±1%. The probe of the electric field detection device is physically isolated from the drone body to ensure a vertical distance of not less than 2.5 meters and a horizontal distance of not less than 0.2 meters, so as to avoid interference from the drone carrier on the electric field measurement.

4. The UAV-based zero-value detection system for transmission line insulators according to claim 1, characterized in that, The edge computing and anti-interference unit is mounted on the edge node and is used to perform real-time preprocessing and anti-interference processing on the raw detection data, wherein: The preprocessing includes: A neighborhood wavelet coefficient algorithm with an improved threshold is used to denoise infrared thermal image data, and an improved differential evolution algorithm with the Otsu function as the evaluation function is used to segment the insulator target region, with a segmentation accuracy of >95%. The anti-interference processing includes an adaptive filtering algorithm and a reinforcement learning algorithm based on Win-or-Fast Learning Strategy (WoLF-PHC) to filter out power frequency electromagnetic interference and dynamically optimize the communication channel. The edge computing and anti-interference unit is also used to extract multi-dimensional features, including insulator temperature variance, maximum temperature difference and electric field distortion rate, from the preprocessed data, and to perform zero-value probability screening using a lightweight BP neural network with less than 100k parameters, achieving a screening accuracy of >90% and compressing redundant data.

5. The zero-value detection system for transmission line insulators based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The multi-source data fusion intelligent diagnostic unit is used to fuse and deeply analyze multi-source feature data from edge nodes and fixed monitoring units; It employs a fusion analysis of the XGBoost model and an improved BP neural network model. The improved BP neural network has an input layer of 16 nodes and a hidden layer of 8 nodes. The multi-source data fusion intelligent diagnostic unit can output the judgment result of zero-value insulators, defect level and deterioration trend analysis, and the overall system identification accuracy is not less than 95%.

6. The zero-value detection system for transmission line insulators based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The communication link disaster recovery unit uses at least two of the following to form a primary and backup redundant link: 5G network, fiber optic network, and satellite communication. 5G links are used for high-bandwidth, low-latency data interaction between drones and edge nodes, while fiber optic links are used for massive data backhaul between edge nodes and the cloud. When the primary link is interrupted, the system can automatically switch to the backup link within 30 seconds; Meanwhile, a reinforcement learning algorithm based on WoLF-PHC is applied to dynamically select and optimize the transmission channel to ensure communication reliability in strong electromagnetic environments.

7. The zero-value detection system for transmission line insulators based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The power and mission disaster recovery unit includes: A hybrid power supply system combining solar panels and batteries is configured for fixed monitoring and early warning units; The drone is equipped with a high-energy-density lithium battery that can be quickly replaced, as well as a vehicle-mounted mobile charging pod or ground-based automated airport that supports drone relay. With the support of the system, a single drone can perform effective inspections for at least 60 minutes.

8. A zero-value detection system for transmission line insulators based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The meteorological and airspace safety decision-making unit integrates real-time meteorological and airspace data and presets safety thresholds, wherein: When any of the following conditions are detected: wind speed ≥8m / s, humidity ≥85%, or visibility <500m, or when the drone enters a no-fly zone, the unit automatically triggers a command to control the drone to suspend its mission, return to base, or execute an emergency landing protection procedure, and pushes a warning message to the operation and maintenance terminal.

9. A zero-value detection system for transmission line insulators based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The data security and operation and maintenance closed-loop unit uses encrypted transmission and off-site disaster recovery backup mechanisms to ensure data security. Simultaneously, it can transmit the maintenance results data fed back by on-site operation and maintenance personnel back to the central control and remote control module, which is used to jointly optimize and iteratively update the prediction algorithm of the fixed monitoring and early warning unit, the identification model of the multi-source data fusion intelligent diagnosis unit, and the compensation parameters of the edge computing and anti-interference unit.

10. A method for detecting zero values ​​in transmission line insulators based on unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: Step 1. Risk Prediction and Intelligent Task Planning: Real-time data of the route is collected through a fixed monitoring network, and local analysis is performed using intelligent algorithms to generate early warnings. The cloud platform integrates early warnings, historical data and meteorological information, identifies high-risk points through predictive models, plans precise flight paths and detection strategies for UAVs, and forms detection tasks. Step 2. UAV Adaptive Anti-interference Data Acquisition: The UAV, equipped with multiple sensors, flies to the target and adaptively selects sensor combinations for synchronous data acquisition based on the environment. Through physical shielding and anti-interference communication technology, it ensures stable and accurate data acquisition in strong electromagnetic environments. Step 3. Real-time edge processing and feature purification: The raw data is preprocessed in real time on the UAV, including image denoising, target segmentation and key feature extraction, and a lightweight model is used for initial screening of zero-value probabilities. The data is then compressed and sent back. Step 4. Multi-source data fusion and intelligent diagnosis: The cloud platform integrates edge data and historical information, filters key features, and uses intelligent models to perform fusion analysis to achieve accurate identification, grade classification, and diagnostic report generation for zero-value insulators; Step 5. Visualized Positioning and Closed-Loop Operation and Maintenance Management: The diagnostic results are combined with precise positioning and displayed on an electronic map for visualization. Repair work orders are automatically generated and pushed to the operation and maintenance system. On-site handling results are fed back to the platform, forming a management closed loop. Step 6. Model Iteration and System Disaster Recovery Optimization: The diagnostic and prediction models are continuously optimized using on-site feedback data. The system also has communication redundancy and emergency mechanisms to ensure the safety and data integrity of the UAV in adverse conditions or failures.

Images