Unmanned aerial vehicle electric power inspection monitoring method and system

Abstract

The invention relates to an unmanned aerial vehicle electric power inspection monitoring method and system, and relates to the technical field of electric power inspection, and the method comprises the steps: obtaining multi-source data of a monitoring area, determining potential defect risk information based on the multi-source data, a power grid digital twin model and a fault propagation knowledge graph, and determining a potential defect priority based on the defect risk information, an inspection task is generated in combination with available unmanned aerial vehicle resources and an optimization algorithm, the inspection task is sent to an unmanned aerial vehicle cluster to collect data, so that the unmanned aerial vehicle cluster performs data collection on each to-be-verified risk area according to the unmanned aerial vehicle inspection task, and defects are identified based on the collected monitoring data and a defect identification AI model. And finally generating a maintenance decision suggestion based on the identification result and the risk information. The technical effects of accurately determining the potential defect risk of the power system, reasonably distributing the unmanned aerial vehicle resources for routing inspection, accurately identifying the defects and generating effective maintenance decision suggestions are achieved, and the efficiency and accuracy of power routing inspection monitoring are improved.

Description

Technical Field

[0001] This invention relates to the field of power line inspection technology, and in particular to a method and system for monitoring and inspecting power lines using unmanned aerial vehicles (UAVs). Background Technology

[0002] With the rapid development of the power industry, the power grid is constantly expanding, and power facilities are increasingly distributed, placing higher demands on the safe and stable operation of the power system. Power grid inspection, as a crucial means of ensuring power grid safety, directly impacts the reliability of power supply through its efficiency and accuracy. Effective power grid inspection can promptly identify potential safety hazards, prevent faults, and reduce power outages and economic losses, playing a vital role in ensuring the normal operation of society and the stable development of the economy.

[0003] In traditional power grid inspections, two methods are typically used to detect potential defects and faults: manual inspection and fixed sensor monitoring. Manual inspection relies on inspectors periodically visiting the site to check power equipment, directly observing its appearance and operating status to determine if any abnormalities are present. This method allows for a relatively detailed inspection and obtains intuitive information. Fixed sensor monitoring involves installing various sensors at key locations in the power grid to collect real-time operating data such as temperature, current, and voltage. Analysis of this data helps identify potential problems. This method enables real-time monitoring of equipment operating status and timely detection of abnormal data changes.

[0004] However, traditional power grid inspection methods have certain drawbacks. Manual inspections are limited by manpower and time costs, making it difficult to achieve real-time and comprehensive monitoring of large-area power grids, and some hidden defects may not be detected in time. While fixed sensor monitoring can collect data in real time, its monitoring range is limited, and it cannot comprehensively assess complex environmental factors and potential risks, making it difficult to accurately predict the future development and impact of potential defects. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and system for unmanned aerial vehicle (UAV) power line inspection and monitoring, aiming to solve at least one of the above-mentioned technical problems.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: Firstly, this application provides a method for unmanned aerial vehicle (UAV) power line inspection and monitoring, employing the following technical solution: A method for unmanned aerial vehicle (UAV) power line inspection and monitoring includes: Acquire multi-source data for the monitoring area, including real-time power grid operation data, meteorological forecast data, geographic information data, public opinion data streams, and disaster early warning data for the monitoring area; Based on multi-source data of the monitoring area, a preset digital twin model of the power grid, and a fault propagation knowledge graph, risk information of potential defects in the monitoring area is determined. The risk information includes the path, probability, and scope of impact of potential defects developing into faults within a preset time window in the future. Based on the risk information of each potential defect in the monitoring area, the priority of each potential defect is determined; Based on the priority of each potential defect, the status of available drone resources, and a preset multi-objective optimization algorithm, a drone inspection task is generated. The drone inspection task includes allocating drone information, inspection path, and execution time to each risk area to be verified. The drone inspection task is sent to the corresponding drone cluster so that the drone cluster can collect data on each of the risk areas to be verified according to the drone inspection task, and obtain the monitoring data of each of the risk areas to be verified, wherein the risk area to be verified is the location area where the potential defect is located. Based on the preset defect identification AI model, defects are identified in the monitoring data of each of the risk areas to be verified, and the defect identification results of each of the risk areas to be verified are obtained. Based on the defect identification results and risk information, maintenance decision recommendations for the monitored area are generated.

[0007] The beneficial effects of this invention are as follows: By integrating multi-source heterogeneous information such as real-time power grid data, meteorological, geographical, public opinion, and disaster early warning, and dynamically extrapolating based on a high-fidelity power grid digital twin model and fault propagation knowledge graph, it can accurately predict the development path, probability, and impact range of potential defects, thereby identifying high-risk points before a fault occurs. This transforms inspection work from indiscriminate periodic coverage into a targeted verification action. Based on the quantitative prioritization of defect risks and combined with the resource status of UAVs, UAV scheduling is performed through spatiotemporal clustering and multi-objective optimization algorithms, solving the problems of low resource utilization and untimely response in traditional manual planning or simple path planning. The UAV swarm collects monitoring data and performs defect identification, accurately identifying defects in risk areas to be verified. Combining defect identification results and risk information to generate maintenance decision suggestions provides a scientific basis for power grid maintenance.

[0008] Based on the above technical solution, the present invention can be further improved as follows.

[0009] Furthermore, the step of determining risk information regarding potential defects in the monitoring area based on multi-source data of the monitoring area, a pre-set digital twin model of the power grid, and a fault propagation knowledge graph includes: Semantic analysis of the public opinion data stream in the monitored area is performed based on natural language processing to obtain public opinion events in the monitored area. The public opinion events represent the event type, location, and timestamp of power failure and damage events in the monitored area. By spatiotemporally aligning real-time power grid operation data, weather forecast data, geographic information data, disaster early warning data, and public opinion events, we obtain spatiotemporally aligned real-time power grid operation data, weather forecast data, geographic information data, disaster early warning data, and public opinion events. The real-time operation data of the power grid after spatiotemporal alignment is mapped to the corresponding electrical nodes in the power grid digital twin model, and the electrical state of the power grid digital twin model is updated. The power grid digital twin model is a virtual simulation model established based on the three-dimensional geometric model of the power grid equipment in the monitoring area, the electrical topology connection relationship, the equipment physical parameters and historical operation data. Based on geographic information technology and a preset spatial interpolation algorithm, the spatiotemporally aligned meteorological forecast data, geographic information data, disaster early warning data, and public opinion events are loaded into the spatial locations corresponding to each physical device in the power grid digital twin model to obtain the power grid digital twin of the monitoring area. Driven by the meteorological forecast data, disaster early warning data, and public opinion events, the digital twin of the power grid is simulated based on the preset equipment physical degradation model and the preset fault propagation knowledge graph to determine the risk information of potential defects in the monitoring area. The preset fault propagation knowledge graph defines the chain reaction rules between defects and faults.

[0010] The beneficial effects of adopting the above-mentioned further solutions are as follows: semantic analysis of public opinion data streams can identify public opinion events, thereby providing a more comprehensive understanding of the damage caused by power outages in the monitored area; spatiotemporal alignment of multi-source data enables data to be coordinated and unified in time and space, mapping real-time power grid operation data to the power grid digital twin model to update the electrical state, and using the fused dynamic data to drive equipment degradation models and fault propagation knowledge graphs to perform simulations in the twin model. This allows for quantitative and visual prediction of the path, probability, and scope of impact of defects developing into faults, providing a clear understanding of how potential defects develop into faults. This effectively improves the early warning capability and proactive defense level of the power grid's safe operation.

[0011] Furthermore, the semantic analysis of the public opinion data stream in the monitored area based on natural language processing to obtain public opinion events in the monitored area includes: Based on the power industry dictionary and preset event rule templates, keywords related to the risk of external damage to the power grid are identified from the text of public opinion data streams to obtain the first event information. The keywords represent words describing construction machinery, fire, fallen trees, theft, and geographical location descriptions. Target detection is performed on the images in the public opinion data stream, and the identified targets are matched with the coordinates of the electronic map to obtain the second event information. The targets are cranes, excavators, wildfires, and thick smoke. Based on the first event information and the second event information, the public opinion events in the monitored area are determined.

[0012] The beneficial effects of adopting the above-mentioned further solutions are as follows: by using a power industry dictionary and preset event rule templates to identify keywords related to external damage risks to the power grid from the text of public opinion data streams, and by performing target detection on public opinion data stream images and matching them with electronic map coordinates, it is possible to comprehensively and accurately obtain public opinion events in the monitoring area, providing more detailed and reliable data for subsequent identification of potential defect risk information, improving the accuracy and comprehensiveness of risk information, and ultimately enhancing the effectiveness and quality of UAV power inspection and monitoring.

[0013] Furthermore, determining the priority of each potential defect based on the risk information of each potential defect in the monitoring area includes: Based on the first historical statistical data of the defect types of each potential defect, a first type risk coefficient is determined for each potential defect. The type risk coefficient characterizes the frequency and speed at which such a defect develops into a failure in the historical record. Based on the second historical statistical data of the defect types of each potential defect within their corresponding impact range, the second type risk coefficient corresponding to each potential defect is determined; Based on the safety importance of power grid assets within the scope of each potential defect, a severity coefficient corresponding to each potential defect is determined. The severity coefficient represents the degree of impact on power grid supply after the defect causes a fault. Based on the probability and path of each potential defect developing into a failure within a future preset time window, an urgency coefficient corresponding to each potential defect is determined. The urgency coefficient represents the time urgency at which the defect needs to be dealt with. Based on the public opinion events reflected in the digital twin model of the power grid, which reflect the attention and risk situation in the area where the potential defects are located, the public opinion impact coefficient corresponding to each potential defect is determined. The public opinion impact coefficient characterizes the degree of public attention and pressure to deal with the defect. For any of the aforementioned potential defects, the priority of the potential defect is determined based on the first type risk coefficient, the second risk coefficient, the consequence severity coefficient, the urgency coefficient, the public opinion impact coefficient, and a preset weighted fusion algorithm.

[0014] The beneficial effects of adopting the above-mentioned further scheme are as follows: by comprehensively considering factors such as historical statistical data on defect types, the importance of power grid assets within the affected area, the probability and path of development into a fault, and the level of attention and risk situation reflected in public opinion events, the risk level of each potential defect can be measured more accurately. Through a preset weighted fusion algorithm, the coefficients of these five dimensions are dynamically integrated into a comprehensive priority index, thereby realizing the objective ranking of potential hazards, enabling more reasonable allocation of UAV resources, improving inspection efficiency and targeting, and optimizing resource allocation efficiency.

[0015] Furthermore, the process of generating a drone inspection task based on the priority of each potential defect, the available drone resource status, and a preset multi-objective optimization algorithm includes: S41, based on the preset time window, location and priority of each potential defect, a spatiotemporal clustering algorithm is used to aggregate each defect point into multiple candidate inspection clusters, each candidate inspection cluster including the dominant defect type, the highest risk level and the suggested time window; S42, For each candidate inspection cluster, generate initial inspection task requirements based on the dominant defect type, highest risk level, and preset inspection strategy knowledge base of the candidate inspection cluster. S43. Based on the initial inspection task requirements of each candidate inspection cluster, the status of available UAV resources, airspace restrictions and weather window constraints, with the optimization objectives of maximizing task completion rate and minimizing total energy consumption and task completion time, an improved multi-objective evolutionary algorithm is used to optimize the UAV scheduling scheme to obtain the initial UAV scheduling scheme. S44, Perform Monte Carlo simulation on the initial UAV scheduling scheme, evaluate the initial UAV scheduling scheme by injecting random disturbance events, and generate evaluation results. The simulation environment integrates UAV dynamics model, sensor performance model and dynamic environment interference model. S45, for each candidate inspection cluster, adjust the parameters of the spatiotemporal clustering algorithm, and / or the weights of the multi-objective optimization algorithm and / or the initial inspection task requirements based on the evaluation results, and execute steps S43 and S44 until the evaluation results meet the set requirements or reach the maximum number of iterations, and determine the current UAV scheduling scheme as the final UAV inspection task.

[0016] The beneficial effects of adopting the above-mentioned further scheme are as follows: considering the preset time window, location and priority of potential defects, the spatiotemporal clustering algorithm is used to aggregate defect points into candidate inspection clusters. The initial inspection task requirements are generated by combining the dominant defect type, the highest risk level and the inspection strategy knowledge base. The UAV scheduling scheme is optimized with the goal of maximizing the task completion rate and minimizing the total energy consumption and task completion time. The scheme is evaluated by Monte Carlo simulation and the parameters and weights are adjusted. More reasonable, efficient and reliable UAV inspection tasks can be generated, improving inspection efficiency and quality and reducing energy consumption and time costs.

[0017] Furthermore, based on the preset time window, location, and priority of each potential defect, a spatiotemporal clustering algorithm is used to aggregate each defect point into multiple candidate inspection clusters, including: Based on the preset time window, location, and priority of each potential defect, a spatiotemporal feature vector is constructed for each defect point; Based on the spatiotemporal feature vector, the weighted spatiotemporal distance between any two defect points is calculated, wherein the weighted spatiotemporal distance is the weighted sum of spatial Euclidean distance, the reciprocal of temporal window overlap, and priority difference; Clustering analysis is performed based on the weighted spatiotemporal distance as a metric and a density-based clustering algorithm, making high-priority defect points the core points, generating multiple clusters, and defining each cluster as a candidate inspection cluster.

[0018] The beneficial effects of adopting the above-mentioned further scheme are as follows: constructing spatiotemporal feature vectors based on the preset time window, location, and priority of potential defects can quantify and fuse various information of defect points; calculating the weighted spatiotemporal distance between any two defect points, and comprehensively considering spatial distance, time window overlap, and priority differences, can more accurately measure the correlation between defect points; using the weighted spatiotemporal distance as a metric for cluster analysis, high-priority defect points become core points to generate candidate inspection clusters, which can reasonably aggregate defect points and provide a foundation for subsequent generation of UAV inspection tasks.

[0019] Furthermore, the monitoring data includes one or more of high-resolution visible light images, infrared thermal imaging data, and ultraviolet imaging data; the defect identification based on the preset defect identification AI model for each of the monitoring data of the risk areas to be verified, to obtain the defect identification result for each of the risk areas to be verified, includes: For each of the risk areas to be verified, a visible light image is input into a visual defect detection model based on a convolutional neural network to obtain a first recognition result. The visual defect detection model is used to identify insulator damage, missing bolts, corrosion, and foreign objects hanging. For each of the risk areas to be verified, the infrared thermal imaging data is input into the overheating defect detection model based on temperature field analysis to obtain a second identification result. The overheating defect detection model is used to identify abnormal heating points and calculate the relative temperature difference. For each of the aforementioned risk regions to be verified, the ultraviolet imaging data is input into a discharge defect detection model based on photon statistics to obtain a third identification result. The discharge defect detection model is used to identify corona discharge and surface arc. For each of the risk areas to be verified, a defect identification result is determined based on the first identification result, the second identification result, and the third identification result.

[0020] The beneficial effects of adopting the above-mentioned further scheme are as follows: using a visual defect detection model based on convolutional neural networks to identify insulator damage, missing bolts, corrosion, and foreign object suspension; using an overheating defect detection model based on temperature field analysis to identify abnormal heating points and calculate relative temperature differences; and using a discharge defect detection model based on photon statistics to identify corona discharge and surface arcs. By combining the three identification results, the defect identification result can be determined, which can accurately detect multiple defects in the risk area to be verified and improve the accuracy of defect identification.

[0021] Furthermore, the step of generating maintenance decision recommendations for the monitoring area based on the defect identification results and risk information includes: For any potential defect, the defect identification result is compared with the risk information to obtain a matching value; For any potential defect, if the matching value between the defect identification result and the risk information is greater than a set matching threshold, then the defect feature similarity is obtained by comparing the defect identification result with the corresponding features of each historical case. For any potential defect, the load rate, years of operation, and recent maintenance records of the equipment corresponding to the defect identification result are compared with the equipment status in the same period of each historical case to calculate the status similarity. For any potential defect, the environmental similarity is calculated by comparing the environmental information corresponding to the defect identification result with the environmental conditions of each historical case. For any potential defect, the risk similarity is calculated by comparing it with the risks corresponding to various historical cases based on the failure probability, scope of impact, and development speed in the risk information. For any potential defect, based on the similarity of defect features, state, environment, and risk, a reference historical maintenance decision recommendation is determined; For any potential defect, a maintenance decision recommendation is determined based on historical maintenance decision recommendations and the priority of the defects; For any potential defect, if the matching value between the defect identification result and the risk information is not greater than a set matching threshold, then a maintenance decision recommendation is determined based on expert review.

[0022] The beneficial effects of adopting the above-mentioned further scheme are as follows: the defect identification results are compared with risk information to obtain a matching value. If the matching value is greater than the set matching threshold, the similarity of defect features, state, environment, and risk is calculated by comparing with historical cases to determine the reference historical maintenance decision recommendations. Then, the maintenance decision recommendations are determined in combination with the defect priority. If the matching value is not greater than the set matching threshold, the maintenance decision recommendations are determined based on expert review. This approach can comprehensively consider multiple factors to generate more reasonable and accurate maintenance decision recommendations for the monitoring area.

[0023] Furthermore, it also includes: Based on the defect identification results, risk information, and actual power grid events, the parameters of the preset power grid digital twin model and defect identification AI model are optimized.

[0024] The beneficial effects of adopting the above-mentioned further solutions are: by comparing and verifying the AI ​​identification results with the early risk predictions, and combining them with actual power grid events, the parameters of the digital twin prediction model and the defect identification AI model can be continuously calibrated and incrementally trained.

[0025] Secondly, this application provides an unmanned aerial vehicle (UAV) power line inspection and monitoring system, which adopts the following technical solution: A drone-based power line inspection and monitoring system includes: The acquisition module is used to acquire multi-source data of the monitoring area, including real-time power grid operation data, meteorological forecast data, geographic information data, public opinion data stream and disaster early warning data of the monitoring area; The potential defect determination module is used to determine the risk information of potential defects in the monitoring area based on multi-source data of the monitoring area, a preset power grid digital twin model and a fault propagation knowledge graph. The risk information includes the path, probability and scope of impact of the potential defect developing into a fault within a preset time window in the future. The priority determination module is used to determine the priority of each potential defect based on the risk information of each potential defect in the monitoring area; The inspection task generation module is used to generate drone inspection tasks based on the priority of each potential defect, the status of available drone resources, and a preset multi-objective optimization algorithm. The drone inspection task includes allocating drone information, inspection path, and execution time for each risk area to be verified. The drone inspection module is used to send the drone inspection task to the corresponding drone cluster, so that the drone cluster can collect data on each of the risk areas to be verified according to the drone inspection task, and obtain the monitoring data of each of the risk areas to be verified, wherein the risk area to be verified is the location area where the potential defect is located. The defect identification module is used to identify defects in the monitoring data of each of the risk areas to be verified based on a preset defect identification AI model, and to obtain the defect identification result for each of the risk areas to be verified. The generation module is used to generate maintenance decision suggestions for the monitored area based on the defect identification results and risk information.

[0026] Additional aspects and advantages of this application will be set forth in part in the description which follows, and will become apparent from the description or may be learned by practice of this application. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating a method for monitoring and inspecting power lines using a drone, as provided in one embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of a UAV power line inspection and monitoring system provided in one embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0029] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0030] This application provides a method for monitoring and inspecting power lines using unmanned aerial vehicles (UAVs). This method can be executed by an electronic device, which can be a server or a mobile terminal device. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services. The mobile terminal device can be a laptop computer, a desktop computer, etc., but is not limited to these.

[0031] like Figure 1 As shown, a method for unmanned aerial vehicle (UAV) power line inspection and monitoring mainly includes: S1, acquire multi-source data of the monitoring area, including real-time power grid operation data, meteorological forecast data, geographic information data, public opinion data stream and disaster early warning data of the monitoring area; In this embodiment, real-time power grid operation data can be acquired through various sensors installed on power grid equipment, such as current sensors and voltage sensors. These sensors can collect equipment operating parameters in real time, reflecting the actual operating status of the power grid. Weather forecast data can be obtained from meteorological departments. This data includes weather conditions, temperature, wind speed, and other information for the monitored area, as weather conditions may affect the operation of power grid equipment. Geographic information data can be acquired through Geographic Information Systems (GIS). This data provides information on the topography, landforms, and geographical location of the monitored area, helping to understand the distribution of power grid equipment. Public opinion data streams can be collected from social media, news websites, and other channels. These streams reflect public attention to the power system and reports on related events. Disaster early warning data can be obtained from relevant disaster early warning departments, such as earthquake and flood warnings. This early warning information can help prepare for potential responses.

[0032] S2, based on multi-source data of the monitoring area, a preset digital twin model of the power grid, and a fault propagation knowledge graph, determine the risk information of potential defects in the monitoring area. The risk information includes the path, probability, and scope of impact of potential defects developing into faults within a preset time window in the future. In this embodiment of the application, determining the risk information of potential defects in the monitoring area based on multi-source data of the monitoring area, a preset power grid digital twin model, and a fault propagation knowledge graph includes: Semantic analysis of the public opinion data stream in the monitored area is performed based on natural language processing to obtain public opinion events in the monitored area. The public opinion events represent the event type, location, and timestamp of power failure and damage events in the monitored area. By spatiotemporally aligning real-time power grid operation data, weather forecast data, geographic information data, disaster early warning data, and public opinion events, we obtain spatiotemporally aligned real-time power grid operation data, weather forecast data, geographic information data, disaster early warning data, and public opinion events. The real-time operation data of the power grid after spatiotemporal alignment is mapped to the corresponding electrical nodes in the power grid digital twin model, and the electrical state of the power grid digital twin model is updated. The power grid digital twin model is a virtual simulation model established based on the three-dimensional geometric model of the power grid equipment in the monitoring area, the electrical topology connection relationship, the equipment physical parameters and historical operation data. Based on geographic information technology and a preset spatial interpolation algorithm, the spatiotemporally aligned meteorological forecast data, geographic information data, disaster early warning data, and public opinion events are loaded into the spatial locations corresponding to each physical device in the power grid digital twin model to obtain the power grid digital twin of the monitoring area. Driven by the meteorological forecast data, disaster early warning data, and public opinion events, the digital twin of the power grid is simulated based on the preset equipment physical degradation model and the preset fault propagation knowledge graph to determine the risk information of potential defects in the monitoring area. The preset fault propagation knowledge graph defines the chain reaction rules between defects and faults.

[0033] In the above embodiments, semantic analysis of the public opinion data stream of the monitored area is performed based on natural language processing to obtain public opinion events in the monitored area, including: Based on the power industry dictionary and preset event rule templates, keywords related to the risk of external damage to the power grid are identified from the text of public opinion data streams to obtain the first event information. The keywords represent words describing construction machinery, fire, fallen trees, theft, and geographical location descriptions. Target detection is performed on the images in the public opinion data stream, and the identified targets are matched with the coordinates of the electronic map to obtain the second event information. The targets are cranes, excavators, wildfires, and thick smoke. Based on the first event information and the second event information, the public opinion events in the monitored area are determined.

[0034] Using a pre-set power industry dictionary and event rule templates, natural language processing (NLP) is performed on text from social media and news platforms to identify keywords such as "crane construction," "wildfire," and "fallen tree," as well as geographical location descriptions. At the same time, deep learning-based target detection is performed on associated images to identify visual targets such as construction machinery and smoke. These targets are then matched with electronic maps through georegistration to obtain structured public opinion events, including event type, precise coordinates, and occurrence time.

[0035] Then, real-time power grid operation data, weather forecast data, geographic information data, disaster early warning data, and public opinion events are spatiotemporally aligned to ensure consistency in time and space. All data are synchronized through a unified time server and mapped to a unified spatiotemporal grid coordinate system using spatial interpolation algorithms based on their geographic location information, thus completing spatiotemporal alignment and forming a multi-dimensional, spatiotemporally consistent "data field".

[0036] The spatiotemporally aligned real-time power grid operation data is mapped to the corresponding electrical nodes in the model, updating their electrical status. Simultaneously, using geographic information technology, aligned meteorological data, such as local wind speed, rainfall, disaster warning areas, and the location of public opinion events, are used as dynamic loads and disturbance sources. These data are then precisely loaded into the spatial locations of affected physical devices in the model, such as towers, conductors, and insulators, through spatial interpolation algorithms.

[0037] Finally, on a twin loaded with dynamic data, driven by injected disturbances such as weather, disasters, and public opinion, the instantaneous changes in the state of each device are first calculated using a physical degradation model. Once the state of a device reaches the defect threshold defined in the knowledge graph, subsequent chain reaction simulations are triggered according to the graph rules. Through numerous simulations, the probability of different potential defects developing into failures, the expected development time window, the specific failure propagation path, and the scope of impact on power grid operation can be statistically determined.

[0038] S3, based on the risk information of each potential defect in the monitoring area, determine the priority of each potential defect; In this embodiment of the application, S3 includes the following sub-steps: S31, based on the first historical statistical data of the defect types of each potential defect, determine the first type risk coefficient corresponding to each potential defect, wherein the type risk coefficient characterizes the frequency and speed at which such defect develops into a failure in the historical record. S32, based on the second historical statistical data of the defect type of each potential defect within the corresponding influence range, determine the second type risk coefficient corresponding to each potential defect; S33, based on the safety importance of power grid assets within the scope of each potential defect, determine the severity coefficient of each potential defect, wherein the severity coefficient of each potential defect characterizes the degree of impact on power grid supply after the defect causes a fault. S34. Based on the probability and path of each potential defect developing into a failure within a future preset time window, determine the urgency coefficient corresponding to each potential defect. The urgency coefficient represents the time urgency at which the defect needs to be dealt with. S35, based on the attention and risk situation of the potential defect area reflected by the public opinion events loaded into the power grid digital twin model, determine the public opinion impact coefficient corresponding to each potential defect, wherein the public opinion impact coefficient characterizes the degree of public attention and pressure to deal with the defect. S36. For any of the potential defects, the priority of the potential defect is determined based on the first type risk coefficient, the second risk coefficient, the consequence severity coefficient, the urgency coefficient, the public opinion impact coefficient, and a preset weighted fusion algorithm.

[0039] In the above implementation, firstly, a historical defect database is accessed to calculate inherent risk coefficients in two dimensions. The first type of risk coefficient is derived by statistically analyzing the average failure conversion rate and average development time of similar defects across the entire network's historical data, reflecting its general risk level. The second type of risk coefficient is further refined by statistically analyzing historical data from specific regions or similar environments (such as coastal areas or heavy industrial zones) where the defect is located, reflecting regional specificity. Secondly, based on a pre-set power grid asset importance map, the sum of importance scores for all equipment within the defect's impact range is calculated and normalized to obtain the consequence severity coefficient. Next, the urgency coefficient is directly driven by the prediction data from the digital twin model, typically calculated using a function model that is directly proportional to the "predicted failure probability" and inversely proportional to the "predicted development time window" to quantify the time urgency.

[0040] Next, sentiment analysis, dissemination intensity assessment, and key entity identification are performed on related public opinion events. These are then transformed into a quantitative value reflecting social pressure through pre-defined mapping rules. Finally, a pre-defined weighted fusion algorithm is used, for example, the comprehensive risk value = W1 × first-type risk coefficient + W2 × second-type risk coefficient + W3 × consequence severity coefficient + W4 × urgency coefficient + W5 × public opinion impact coefficient. The weights W1 to W5 can be used as strategy parameters and dynamically adjusted according to different operational phases. The comprehensive risk value of all defects is calculated and sorted in descending order, generating a final priority list to guide precise resource allocation.

[0041] S4. Based on the priority of each potential defect, the status of available drone resources, and a preset multi-objective optimization algorithm, a drone inspection task is generated. The drone inspection task includes allocating drone information, inspection path, and execution time to each risk area to be verified. In this embodiment of the application, the step of generating a drone inspection task based on the priority of each potential defect, the available drone resource status, and a preset multi-objective optimization algorithm includes: S41, based on the preset time window, location and priority of each potential defect, a spatiotemporal clustering algorithm is used to aggregate each defect point into multiple candidate inspection clusters, each candidate inspection cluster including the dominant defect type, the highest risk level and the suggested time window; S42, For each candidate inspection cluster, generate initial inspection task requirements based on the dominant defect type, highest risk level, and preset inspection strategy knowledge base of the candidate inspection cluster. S43. Based on the initial inspection task requirements of each candidate inspection cluster, the status of available UAV resources, airspace restrictions and weather window constraints, with the optimization objectives of maximizing task completion rate and minimizing total energy consumption and task completion time, an improved multi-objective evolutionary algorithm is used to optimize the UAV scheduling scheme to obtain the initial UAV scheduling scheme. S44, Perform Monte Carlo simulation on the initial UAV scheduling scheme, evaluate the initial UAV scheduling scheme by injecting random disturbance events, and generate evaluation results. The simulation environment integrates UAV dynamics model, sensor performance model and dynamic environment interference model. S45, for each candidate inspection cluster, adjust the parameters of the spatiotemporal clustering algorithm, and / or the weights of the multi-objective optimization algorithm and / or the initial inspection task requirements based on the evaluation results, and execute steps S43 and S44 until the evaluation results meet the set requirements or reach the maximum number of iterations, and determine the current UAV scheduling scheme as the final UAV inspection task.

[0042] In the above implementation, a spatiotemporal feature vector is first constructed for each potential defect based on its preset time window, location, and priority. Based on these spatiotemporal feature vectors, a weighted spatiotemporal distance is calculated between any two defect points. This weighted spatiotemporal distance is the weighted sum of spatial Euclidean distance, the reciprocal of the time window overlap, and the priority difference. Clustering analysis is then performed using the weighted spatiotemporal distance as the metric and a density-based clustering algorithm, making high-priority defect points the core points and generating multiple clusters. Each cluster is defined as a candidate inspection cluster. Each candidate inspection cluster includes the dominant defect type, the highest risk level, and a suggested time window.

[0043] For each candidate inspection cluster, initial inspection task requirements are generated based on the cluster's dominant defect type, highest risk level, and a pre-defined inspection strategy knowledge base. Then, based on these initial task requirements, available UAV resource status, airspace restrictions, and weather window constraints, an improved multi-objective evolutionary algorithm is used to optimize the UAV scheduling scheme, aiming to maximize task completion rate and minimize total energy consumption and task completion time. This yields an initial UAV scheduling scheme. Monte Carlo simulations are then performed on the initial UAV scheduling scheme, injecting random disturbance events to evaluate it and generate evaluation results. The simulation environment integrates UAV dynamics models, sensor performance models, and dynamic environmental interference models. For each candidate inspection cluster, the parameters of the spatiotemporal clustering algorithm and / or the weights of the multi-objective optimization algorithm and / or the initial inspection task requirements are adjusted based on the evaluation results. The optimization and evaluation steps are repeated until the evaluation results meet the set requirements or the maximum number of iterations is reached. The current UAV scheduling scheme is then determined as the final UAV inspection task. This UAV inspection task includes assigning UAV information, inspection paths, and execution times to each risk area to be verified.

[0044] Wherein, the objective function F is in the form of a weighted summation: F = α*Σ(Mission Risk Coverage) + β*(1 / Total Flight Time) + γ*(1 / Maximum Single-Unit Energy Consumption) - δ*Σ(Constraint Violation Penalty) Where α, β, γ, and δ are adjustable weight coefficients, Σ (task risk coverage value) is the sum of the comprehensive risk values ​​of all defect points of the assigned task, and the constraints include UAV endurance, payload, airspace, time window and sensor matching constraints. The improved multi-objective evolutionary algorithm adopts a cluster center-based greedy strategy to generate high-quality initial solutions during population initialization, and introduces a simulated annealing mechanism in the mutation operation to maintain population diversity and avoid premature convergence.

[0045] In this embodiment of the application, the termination condition for collaborative iterative optimization is: The comprehensive performance score in the evaluation results improves by less than a preset threshold Δ for K consecutive generations, or the total number of iterations reaches a preset maximum value M. If the process terminates due to reaching the maximum number of iterations, the scheduling scheme with the highest overall performance score is selected from the iteration history as the final scheme.

[0046] S5, the drone inspection task is sent to the corresponding drone cluster so that the drone cluster can collect data on each of the risk areas to be verified according to the drone inspection task and obtain the monitoring data of each of the risk areas to be verified. The risk areas to be verified are the locations of potential defects. S6, Based on the preset defect identification AI model, perform defect identification on the monitoring data of each of the risk areas to be verified, and obtain the defect identification result of each of the risk areas to be verified; In this embodiment of the application, the monitoring data includes one or more of the following: high-resolution visible light images, infrared thermal imaging data, and ultraviolet imaging data; The AI ​​model for defect identification, based on a preset model, performs defect identification on the monitoring data of each of the risk areas to be verified, and obtains the defect identification result for each risk area to be verified, including: For each of the risk areas to be verified, a visible light image is input into a visual defect detection model based on a convolutional neural network to obtain a first recognition result. The visual defect detection model is used to identify insulator damage, missing bolts, corrosion, and foreign objects hanging. For each of the risk areas to be verified, the infrared thermal imaging data is input into the overheating defect detection model based on temperature field analysis to obtain a second identification result. The overheating defect detection model is used to identify abnormal heating points and calculate the relative temperature difference. For each of the aforementioned risk regions to be verified, the ultraviolet imaging data is input into a discharge defect detection model based on photon statistics to obtain a third identification result. The discharge defect detection model is used to identify corona discharge and surface arc. For each of the risk areas to be verified, a defect identification result is determined based on the first identification result, the second identification result, and the third identification result.

[0047] In the above embodiments, the visible light camera, infrared thermal imager and ultraviolet imager carried by the UAV collect data on the target power equipment, such as towers, insulator strings, conductor joints and circuit breakers, in the risk area to be verified in the final UAV inspection mission, and obtain high-resolution visible light images, infrared thermal imaging temperature matrices and ultraviolet discharge photon count images under the same spatiotemporal reference.

[0048] The visual defect detection model is specifically optimized for the appearance features of power equipment. It can automatically identify and label structural anomalies in images, mainly including damage, cracks, and flashover marks of insulators; missing, loose, corroded, and deformed hardware (such as clamps, equalizing rings, and bolts); broken strands, loose strands, and foreign objects hanging on conductors and ground wires; and corrosion, missing parts, and foundation settlement of tower materials.

[0049] The overheating defect detection model performs a detailed analysis of the equipment temperature distribution. By comparing the temperature difference of similar equipment, the temperature difference between the three phases, and the temperature difference between the equipment and the ambient temperature, it identifies thermal anomalies. The core is to locate abnormal heating points, such as wire joints, drain plates, and disconnector contacts, which are overheated due to increased contact resistance. The discharge defect detection model identifies electrical anomalies, namely partial discharge activity, by analyzing photon burst modes, intensity, and spectral characteristics. Specifically, it includes corona discharge and surface arc.

[0050] S7. Based on the defect identification results and risk information, generate maintenance decision suggestions for the monitoring area, and optimize the parameters of the preset power grid digital twin model and defect identification AI model based on the defect identification results, risk information, and actual power grid events.

[0051] In this embodiment of the application, the step of generating maintenance decision suggestions for the monitoring area based on the defect identification results and risk information includes: For any potential defect, the defect identification result is compared with the risk information to obtain a matching value; For any potential defect, if the matching value between the defect identification result and the risk information is greater than a set matching threshold, then the defect feature similarity is obtained by comparing the defect identification result with the corresponding features of each historical case. For any potential defect, the load rate, years of operation, and recent maintenance records of the equipment corresponding to the defect identification result are compared with the equipment status in the same period of each historical case to calculate the status similarity. For any potential defect, the environmental similarity is calculated by comparing the environmental information corresponding to the defect identification result with the environmental conditions of each historical case. For any potential defect, the risk similarity is calculated by comparing it with the risks corresponding to various historical cases based on the failure probability, scope of impact, and development speed in the risk information. For any potential defect, based on the similarity of defect features, state, environment, and risk, a reference historical maintenance decision recommendation is determined; For any potential defect, a maintenance decision recommendation is determined based on historical maintenance decision recommendations and the priority of the defects; For any potential defect, if the matching value between the defect identification result and the risk information is not greater than a set matching threshold, then a maintenance decision recommendation is determined based on expert review.

[0052] In the above implementation, if the matching value is higher than the set threshold, it indicates that the prediction is relatively accurate, and the system enters the automated assisted decision-making process based on historical cases; if the matching value is lower than the threshold, it indicates that the current situation deviates significantly from the prediction, and an expert review report containing multi-source data and contradictions will be automatically generated and pushed to human experts for final adjudication, thus realizing human-machine collaboration.

[0053] In the above implementation, for parameter optimization of the digital twin model, for false negative (missed detection) cases, the sensitivity parameter or fault development rate parameter of the model to the relevant defects is increased through backpropagation or Bayesian update algorithm; for false positive (false alarm) cases, the corresponding parameters are decreased or constraints are added. For parameter optimization of the defect recognition AI model, the correctly identified difficult samples and incorrectly identified samples (missed detection, false detection) from this inspection, along with their accurate annotations, are added to the model's training sample library for incremental training to optimize the parameters of the defect recognition AI model.

[0054] This method integrates multi-source heterogeneous information such as real-time power grid data, meteorological, geographical, public opinion, and disaster early warning data. Based on a high-fidelity power grid digital twin model and a fault propagation knowledge graph, it dynamically extrapolates and accurately predicts the development path, probability, and impact range of potential defects, thus identifying high-risk points before faults occur. This transforms inspection work from indiscriminate periodic coverage into targeted verification actions. Based on the quantitative prioritization of defect risks and combined with UAV resource status, UAV scheduling is achieved through spatiotemporal clustering and multi-objective optimization algorithms, solving the problems of low resource utilization and untimely response in traditional manual planning or simple path planning. UAV swarms collect monitoring data and perform defect identification, accurately identifying defects in risk areas to be verified. Combining defect identification results and risk information to generate maintenance decision recommendations provides a scientific basis for power grid maintenance.

[0055] Figure 2 A schematic diagram of a UAV power line inspection and monitoring system is shown.

[0056] like Figure 2 As shown, a UAV power line inspection and monitoring system 200 mainly includes: The acquisition module 201 is used to acquire multi-source data of the monitoring area, including real-time power grid operation data, meteorological forecast data, geographic information data, public opinion data stream and disaster early warning data of the monitoring area; The potential defect determination module 202 is used to determine the risk information of potential defects in the monitoring area based on multi-source data of the monitoring area, a preset power grid digital twin model and a fault propagation knowledge graph. The risk information includes the path, probability and scope of impact of the potential defect developing into a fault within a preset time window in the future. The priority determination module 203 is used to determine the priority of each potential defect based on the risk information of each potential defect in the monitoring area; The inspection task generation module 204 is used to generate drone inspection tasks based on the priority of each potential defect, the status of available drone resources, and a preset multi-objective optimization algorithm. The drone inspection task includes allocating drone information, inspection path, and execution time for each risk area to be verified. The drone inspection module 205 is used to send the drone inspection task to the corresponding drone cluster, so that the drone cluster can collect data on each of the risk areas to be verified according to the drone inspection task, and obtain the monitoring data of each of the risk areas to be verified, wherein the risk area to be verified is the location area where the potential defect is located. The defect identification module 206 is used to identify defects in the monitoring data of each of the risk areas to be verified based on a preset defect identification AI model, and to obtain the defect identification result of each of the risk areas to be verified. The generation module 207 is used to generate maintenance decision suggestions for the monitoring area based on the defect identification results and risk information.

[0057] In one example, the module in any of the above devices may be one or more integrated circuits configured to implement the above methods, such as one or more application-specific integrated circuits (ASICs), or one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms.

[0058] For example, when modules in a device can be implemented via a processing element scheduler, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor capable of calling programs. Alternatively, these modules can be integrated together as a system-on-a-chip (SOC).

[0059] In this application, various objects such as messages / information / devices / network elements / systems / apparatus / actions / operations / processes / concepts may be named. It is understood that these specific names do not constitute a limitation on the relevant objects. The names may be changed depending on the scenario, context, or usage habits. The understanding of the technical meaning of the technical terms in this application should be mainly determined from their functions and technical effects embodied / performed in the technical solution.

[0060] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0061] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0062] The terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0063] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing application concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions claimed in this application.

Claims

1. A method for unmanned aerial vehicle (UAV) power line inspection and monitoring, characterized in that, include: Acquire multi-source data for the monitoring area, including real-time power grid operation data, meteorological forecast data, geographic information data, public opinion data streams, and disaster early warning data for the monitoring area; Based on multi-source data of the monitoring area, a preset digital twin model of the power grid, and a fault propagation knowledge graph, risk information of potential defects in the monitoring area is determined. The risk information includes the path, probability, and scope of impact of potential defects developing into faults within a preset time window in the future. Based on the risk information of each potential defect in the monitoring area, the priority of each potential defect is determined; Based on the priority of each potential defect, the status of available drone resources, and a preset multi-objective optimization algorithm, a drone inspection task is generated. The drone inspection task includes allocating drone information, inspection path, and execution time to each risk area to be verified. The drone inspection task is sent to the corresponding drone cluster so that the drone cluster can collect data on each of the risk areas to be verified according to the drone inspection task, and obtain the monitoring data of each of the risk areas to be verified, wherein the risk area to be verified is the location area where the potential defect is located. Based on the preset defect identification AI model, defect identification is performed on the monitoring data of each of the risk areas to be verified, and the defect identification result of each of the risk areas to be verified is obtained. Based on the defect identification results and risk information, maintenance decision recommendations for the monitored area are generated.

2. The method for unmanned aerial vehicle (UAV) power line inspection and monitoring according to claim 1, characterized in that, The method for determining potential defect risk information in the monitoring area based on multi-source data of the monitoring area, a pre-set power grid digital twin model, and a fault propagation knowledge graph includes: Semantic analysis of the public opinion data stream in the monitored area is performed based on natural language processing to obtain public opinion events in the monitored area. The public opinion events represent the event type, location, and timestamp of power failure and damage events in the monitored area. By spatiotemporally aligning real-time power grid operation data, weather forecast data, geographic information data, disaster early warning data, and public opinion events, we obtain spatiotemporally aligned real-time power grid operation data, weather forecast data, geographic information data, disaster early warning data, and public opinion events. The real-time operation data of the power grid after spatiotemporal alignment is mapped to the corresponding electrical nodes in the power grid digital twin model, and the electrical state of the power grid digital twin model is updated. The power grid digital twin model is a virtual simulation model established based on the three-dimensional geometric model of the power grid equipment in the monitoring area, the electrical topology connection relationship, the equipment physical parameters and historical operation data. Based on geographic information technology and a preset spatial interpolation algorithm, the spatiotemporally aligned meteorological forecast data, geographic information data, disaster early warning data, and public opinion events are loaded into the spatial locations corresponding to each physical device in the power grid digital twin model to obtain the power grid digital twin of the monitoring area. Driven by the meteorological forecast data, disaster early warning data, and public opinion events, the digital twin of the power grid is simulated based on the preset equipment physical degradation model and the preset fault propagation knowledge graph to determine the risk information of potential defects in the monitoring area. The preset fault propagation knowledge graph defines the chain reaction rules between defects and faults.

3. The method for unmanned aerial vehicle (UAV) power line inspection and monitoring according to claim 2, characterized in that, The semantic analysis of the public opinion data stream in the monitored area based on natural language processing yields public opinion events in the monitored area, including: Based on the power industry dictionary and preset event rule templates, keywords related to the risk of external damage to the power grid are identified from the text of public opinion data streams to obtain the first event information. The keywords represent words describing construction machinery, fire, fallen trees, theft, and geographical location descriptions. Target detection is performed on the images in the public opinion data stream, and the identified targets are matched with the coordinates of the electronic map to obtain the second event information. The targets are cranes, excavators, wildfires, and thick smoke. Based on the first event information and the second event information, the public opinion events in the monitored area are determined.

4. The method for unmanned aerial vehicle (UAV) power line inspection and monitoring according to claim 2, characterized in that, The step of determining the priority of each potential defect based on risk information of each potential defect in the monitoring area includes: Based on the first historical statistical data of the defect types of each potential defect, a first type risk coefficient is determined for each potential defect. The type risk coefficient characterizes the frequency and speed at which such a defect develops into a failure in the historical record. Based on the second historical statistical data of the defect types of each potential defect within their corresponding impact range, the second type risk coefficient corresponding to each potential defect is determined; Based on the safety importance of power grid assets within the scope of each potential defect, a severity coefficient corresponding to each potential defect is determined. The severity coefficient represents the degree of impact on power grid supply after the defect causes a fault. Based on the probability and path of each potential defect developing into a failure within a future preset time window, an urgency coefficient corresponding to each potential defect is determined. The urgency coefficient represents the time urgency at which the defect needs to be dealt with. Based on the public opinion events reflected in the digital twin model of the power grid, which reflect the attention and risk situation in the area where the potential defects are located, the public opinion impact coefficient corresponding to each potential defect is determined. The public opinion impact coefficient characterizes the degree of public attention and pressure to deal with the defect. For any of the aforementioned potential defects, the priority of the potential defect is determined based on the first type risk coefficient, the second risk coefficient, the consequence severity coefficient, the urgency coefficient, the public opinion impact coefficient, and a preset weighted fusion algorithm.

5. The method for unmanned aerial vehicle (UAV) power line inspection and monitoring according to claim 1, characterized in that, The process of generating a drone inspection task based on the priority of each potential defect, the available drone resource status, and a preset multi-objective optimization algorithm includes: S41, based on the preset time window, location and priority of each potential defect, a spatiotemporal clustering algorithm is used to aggregate each defect point into multiple candidate inspection clusters, each candidate inspection cluster including the dominant defect type, the highest risk level and the suggested time window; S42, For each candidate inspection cluster, generate initial inspection task requirements based on the dominant defect type, highest risk level, and preset inspection strategy knowledge base of the candidate inspection cluster. S43. Based on the initial inspection task requirements of each candidate inspection cluster, the status of available UAV resources, airspace restrictions and weather window constraints, with the optimization objectives of maximizing task completion rate and minimizing total energy consumption and task completion time, an improved multi-objective evolutionary algorithm is used to optimize the UAV scheduling scheme to obtain the initial UAV scheduling scheme. S44, Perform Monte Carlo simulation on the initial UAV scheduling scheme, evaluate the initial UAV scheduling scheme by injecting random disturbance events, and generate evaluation results. The simulation environment integrates UAV dynamics model, sensor performance model and dynamic environment interference model. S45, for each candidate inspection cluster, adjust the parameters of the spatiotemporal clustering algorithm, and / or the weights of the multi-objective optimization algorithm and / or the initial inspection task requirements based on the evaluation results, and execute steps S43 and S44 until the evaluation results meet the set requirements or reach the maximum number of iterations, and determine the current UAV scheduling scheme as the final UAV inspection task.

6. The method for unmanned aerial vehicle (UAV) power line inspection and monitoring according to claim 5, characterized in that, Based on the preset time window, location, and priority of each potential defect, a spatiotemporal clustering algorithm is used to aggregate each defect point into multiple candidate inspection clusters, including: Based on the preset time window, location, and priority of each potential defect, a spatiotemporal feature vector is constructed for each defect point; Based on the spatiotemporal feature vector, the weighted spatiotemporal distance between any two defect points is calculated, wherein the weighted spatiotemporal distance is the weighted sum of spatial Euclidean distance, the reciprocal of temporal window overlap, and priority difference; Clustering analysis is performed based on the weighted spatiotemporal distance as a metric and a density-based clustering algorithm, making high-priority defect points the core points, generating multiple clusters, and defining each cluster as a candidate inspection cluster.

7. The method for unmanned aerial vehicle (UAV) power line inspection and monitoring according to claim 1, characterized in that, The monitoring data includes one or more of high-resolution visible light images, infrared thermal imaging data, and ultraviolet imaging data; the defect identification based on the preset defect identification AI model is performed on the monitoring data of each of the risk areas to be verified to obtain the defect identification result for each of the risk areas to be verified, including: For each of the risk areas to be verified, a visible light image is input into a visual defect detection model based on a convolutional neural network to obtain a first recognition result. The visual defect detection model is used to identify insulator damage, missing bolts, corrosion, and foreign objects hanging. For each of the risk areas to be verified, the infrared thermal imaging data is input into the overheating defect detection model based on temperature field analysis to obtain a second identification result. The overheating defect detection model is used to identify abnormal heating points and calculate the relative temperature difference. For each of the aforementioned risk regions to be verified, the ultraviolet imaging data is input into a discharge defect detection model based on photon statistics to obtain a third identification result. The discharge defect detection model is used to identify corona discharge and surface arc. For each of the risk areas to be verified, a defect identification result is determined based on the first identification result, the second identification result, and the third identification result.

8. A method for monitoring and inspecting power lines using unmanned aerial vehicles (UAVs) according to claim 6, characterized in that, The step of generating maintenance decision recommendations for the monitoring area based on the defect identification results and risk information includes: For any potential defect, the defect identification result is compared with the risk information to obtain a matching value; For any potential defect, if the matching value between the defect identification result and the risk information is greater than a set matching threshold, then the defect feature similarity is obtained by comparing the defect identification result with the corresponding features of each historical case. For any potential defect, the load rate, years of operation, and recent maintenance records of the equipment corresponding to the defect identification result are compared with the equipment status in the same period of each historical case to calculate the status similarity. For any potential defect, the environmental similarity is calculated by comparing the environmental information corresponding to the defect identification result with the environmental conditions of each historical case. For any potential defect, the risk similarity is calculated by comparing it with the risks corresponding to various historical cases based on the failure probability, scope of impact, and development speed in the risk information. For any potential defect, based on the similarity of defect features, state, environment, and risk, a reference historical maintenance decision recommendation is determined; For any potential defect, a maintenance decision recommendation is determined based on historical maintenance decision recommendations and the priority of the defects; For any potential defect, if the matching value between the defect identification result and the risk information is not greater than a set matching threshold, then a maintenance decision recommendation is determined based on expert review.

9. The method for unmanned aerial vehicle (UAV) power line inspection and monitoring according to claim 1, characterized in that, Also includes: Based on the defect identification results, risk information, and actual power grid events, the parameters of the preset power grid digital twin model and defect identification AI model are optimized.

10. A UAV power line inspection and monitoring system, characterized in that, include: The acquisition module is used to acquire multi-source data of the monitoring area, including real-time power grid operation data, meteorological forecast data, geographic information data, public opinion data stream and disaster early warning data of the monitoring area; The potential defect determination module is used to determine the risk information of potential defects in the monitoring area based on multi-source data of the monitoring area, a preset power grid digital twin model and a fault propagation knowledge graph. The risk information includes the path, probability and scope of impact of the potential defect developing into a fault within a preset time window in the future. The priority determination module is used to determine the priority of each potential defect based on the risk information of each potential defect in the monitoring area; The inspection task generation module is used to generate drone inspection tasks based on the priority of each potential defect, the status of available drone resources, and a preset multi-objective optimization algorithm. The drone inspection task includes allocating drone information, inspection path, and execution time for each risk area to be verified. The drone inspection module is used to send the drone inspection task to the corresponding drone cluster, so that the drone cluster can collect data on each of the risk areas to be verified according to the drone inspection task, and obtain the monitoring data of each of the risk areas to be verified, wherein the risk area to be verified is the location area where the potential defect is located. The defect identification module is used to identify defects in the monitoring data of each of the risk areas to be verified based on a preset defect identification AI model, and to obtain the defect identification result for each of the risk areas to be verified. The generation module is used to generate maintenance decision suggestions for the monitored area based on the defect identification results and risk information.

Images