A power inspection system and method based on multi-unmanned aerial vehicle cooperation

Abstract

The application relates to the technical field of electric power engineering, in particular to a power inspection system and method based on cooperation of multiple unmanned aerial vehicles, which comprises a task allocation module, a path planning module, a cooperative control module and an evaluation feedback module; the task allocation module is used for dividing an inspection area into multiple subareas according to the geographical position of a power facility inspection point and the endurance of the unmanned aerial vehicles and assigning corresponding inspection tasks to each unmanned aerial vehicle through an improved K-means++ clustering algorithm; compared with the prior art, the application has the defects of high calculation complexity, difficulty in adapting to large-scale scenes and neglecting individual energy constraints of the unmanned aerial vehicles; the application adopts a distributed task allocation module based on the improved K-means++ algorithm, realizes balanced partition through probabilistic weighted selection of initial clustering centers, and has the advantages of significantly improving the inspection efficiency of large-range power lines, optimizing resource utilization and enhancing the self-adaptive capacity of the system in combination with a multi-machine parallel operation mechanism.

Description

Technical Field

[0001] This invention relates to the field of power engineering technology, and in particular to a power inspection system and method based on multi-UAV collaboration. Background Technology

[0002] As a critical infrastructure, the safe and stable operation of power grids is of paramount importance, requiring regular inspections of transmission lines and towers to identify and promptly eliminate potential defects. Traditional power grid inspection methods primarily rely on manual climbing or ground observation using telescopes. These methods are not only inefficient and risky, but also limited by the experience and subjective judgment of inspection personnel, making it difficult to guarantee inspection quality and coverage. In recent years, single-drone inspections have improved work efficiency and personnel safety to some extent. However, with the continuous expansion of the power grid, the limited endurance and load capacity of single drones have become significant bottlenecks when facing large-scale, long-distance transmission corridors. This results in a single flight being unable to complete all inspection tasks, requiring frequent returns to replace batteries, severely restricting inspection efficiency. Therefore, a power grid inspection system based on multi-drone collaboration has emerged, aiming to overcome the limitations of single drones through parallel operations and intelligent collaboration, thereby improving overall inspection efficiency.

[0003] Existing technologies have attempted to incorporate multi-UAV systems into power line inspection tasks to enhance parallel operation capabilities, but most solutions have significant shortcomings. For example, some methods employ centralized task allocation and path planning, treating the entire inspection area as a complex traveling salesman problem or multiple traveling salesman problem. These methods have high computational complexity, making them unsuitable for large-scale scenarios with numerous inspection points, and they lack sufficient consideration of individual UAV energy constraints. The generated paths may prevent some UAVs from safely returning due to insufficient power. In terms of path planning, traditional two-dimensional trajectory planning ignores the elevation changes of power transmission lines in the real geographical environment, while some three-dimensional trajectory planning algorithms fail to effectively integrate real-time obstacle avoidance and collaborative collision avoidance mechanisms among multiple UAVs. Furthermore, most existing systems lack dynamic adjustment capabilities during task execution. When a UAV malfunctions or a new inspection requirement is discovered, online task reallocation and trajectory replanning are not possible, resulting in insufficient system robustness and making it difficult to guarantee inspection efficiency and reliability.

[0004] To address the shortcomings of the existing technologies, this invention proposes a power line inspection system and method based on multi-UAV collaboration. Summary of the Invention

[0005] To overcome the problems mentioned in the background art, the present invention proposes a power line inspection system and method based on multi-UAV collaboration.

[0006] The technical solution of this invention is: a power line inspection system based on multi-UAV collaboration, comprising:

[0007] The task allocation module is used to divide the inspection area into multiple sub-regions based on the geographical location of the power facility inspection points and the drone's endurance, and to assign corresponding inspection tasks to each drone.

[0008] The path planning module is used to plan energy-sensing 3D inspection trajectories for each UAV within an assigned sub-region, based on a bio-inspired behavioral model and an improved simulated annealing algorithm.

[0009] The collaborative control module is used to achieve distributed collaboration, real-time collision avoidance, and trajectory tracking among multiple UAVs by adopting partitioned control based on Voronoi diagrams and signal timing logic specifications.

[0010] The evaluation and feedback module is used to monitor and evaluate the inspection process in real time based on the digital twin model, and to dynamically adjust the task allocation and path planning strategies based on the evaluation results.

[0011] As a preferred option, the improved K-means++ algorithm in the task allocation module has an initial cluster center selection probability that is proportional to the square of the distance from the inspection point to the existing cluster center. The probability calculation formula is as follows:

[0012] ;

[0013] in, Indicates inspection point The probability of being selected as the next initial cluster center. Indicates inspection point Distance to the nearest cluster center This represents the set of all power facilities to be inspected. This represents the checkpoints traversed during the summation process.

[0014] Preferably, the bio-inspired behavioral model in the path planning module simulates the foraging behavior of boobies, including a default inspection mode and a region-restricted search mode triggered when a defect is detected. The improved simulated annealing algorithm in the path planning module incorporates a 2-opt local search operation, which randomly swaps two nodes in the path and accepts the probability of a bad solution. To escape local optima, the improved simulated annealing algorithm also includes a global optimization enhancement mechanism, which includes an adaptive temperature scheduling strategy, a parallel multi-starting point optimization mechanism, and a historical best solution memory pool. The global optimization capability is enhanced through parallel multi-threaded independent evolution and periodic solution crossover.

[0015] As a preferred option, the energy sensing model used in the path planning module is a composite energy consumption model, with the following expression:

[0016] ;

[0017] in, This represents the estimated total energy consumption for the drone to execute the complete inspection trajectory. This represents the total flight time required to complete the trajectory. This represents the base power of the drone at its level flight speed V. For climb and descent rates Corresponding height change power, Yaw angle change rate Corresponding turning power, Wind speed Additional power compensation caused by this.

[0018] As a preferred option, the collaborative control module adopts a distributed control law, the expression of which is:

[0019] ;

[0020] in, This represents the resultant control input vector of UAV i. This represents the centroid orientation control term for UAV i, used to drive the UAV towards its Voronoi unit centroid. Let i represent the obstacle avoidance control term of drone i, and ,in, Represents the proportional control gain matrix. Represents the differential control gain matrix. The geometric matrix representing the Voronoi cell of drone i. This represents the centroid coordinate vector of the Voronoi cell of drone i. Let represent the current position vector of drone i. Represents the current velocity vector of drone i;

[0021] The signal timing logic in the collaborative control module is used to formally express and verify the spatiotemporal constraints in the inspection task, ensuring that the UAV always meets the requirements of safe distance, area coverage, and return-to-base charging. The general mathematical form of the signal timing logic is as follows:

[0022] ;

[0023] in, The formula for representing the timing logic task specification is as follows: Indicates a time range Internal timing operators, Represents workspace constraints. Represents obstacle avoidance. Represents the safe distance between drones. This represents the constraints for task completion. Represents return-to-base constraints, This represents the logical AND operator. Indicates a time range The final timing operator within, The timing operator indicates until.

[0024] As a preferred approach, the digital twin model constructed by the evaluation and feedback module is synchronized in real time with the physical drone system. Sensor data is processed through edge computing nodes, and reinforcement learning algorithms are used to replan and optimize the pre-planned trajectory online based on the real-time status of power facilities and the drone's battery level. The evaluation and feedback module includes a distributed state consensus unit, a conflict prediction and arbitration unit, and a data credibility federated evaluation unit. The distributed state consensus unit achieves decentralized state synchronization based on a lightweight blockchain consensus mechanism, the conflict prediction and arbitration unit generates a fair arbitration scheme based on a Shapley value calculation cooperative game model, and the data credibility federated evaluation unit uses a federated learning framework to collaboratively train a global credibility evaluation model.

[0025] Preferably, a data input layer is also included to receive multi-source heterogeneous data, including GIS geographic information of power facilities, inspection point sets, initial status of UAVs and environmental parameters, and to perform standardized preprocessing on the data. The preprocessing includes spatiotemporal benchmark unification, data cleaning and completion, format standardization and data fusion to form a comprehensive situational awareness layer for initialization of the digital twin model.

[0026] As a preferred option, it also includes an output and application layer, which is connected to the evaluation and feedback module to output optimized trajectories, control commands, evaluation reports, early warning information and decision support schemes in a standardized format, and supports multiple application interfaces to realize the visualization and interactive control of inspection data.

[0027] As a preferred option, the evaluation feedback module also includes a blockchain evidence storage unit, which is used to encrypt and store key inspection data based on blockchain technology to ensure the immutability, reliability and traceability of the data, and to provide a basis for determining responsibility in the inspection process.

[0028] As a preferred embodiment, a power line inspection method based on multi-UAV collaboration is characterized by the following steps:

[0029] S1: Obtain the three-dimensional geographic information and inspection point set of the power facilities to be inspected, and determine the number of available drones and their energy constraints;

[0030] S2: Cluster the inspection points using the improved K-means++ algorithm and assign inspection sub-tasks to each UAV;

[0031] S3: For each UAV's mission set, the inspection path is initialized based on bio-inspired rules and optimized using an improved simulated annealing algorithm that incorporates 2-opt operations to generate an energy-sensing 3D trajectory.

[0032] S4: During the inspection process, based on the distributed collaborative control law and signal timing logic, real-time trajectory tracking, dynamic collision avoidance and collaborative coverage of multiple UAVs are achieved;

[0033] S5: Monitors the inspection status in real time through a digital twin model, assesses path tracking error, power consumption and task completion, and dynamically triggers task reassignment and trajectory replanning based on the assessment results;

[0034] S6: After the inspection task is completed, the system automatically generates a summary document containing the total inspection time, total flight distance, energy consumption analysis and defect detection report, and stores the key inspection data based on blockchain technology.

[0035] The beneficial effects of this invention are:

[0036] 1. Compared with the centralized task allocation scheme of the existing technology, which has the disadvantages of high computational complexity, difficulty in adapting to large-scale scenarios and ignoring the energy constraints of individual UAVs, the present invention adopts a distributed task allocation module based on the improved K-means++ algorithm. It achieves balanced partitioning by selecting the initial cluster center through probability weighting. Combined with the multi-machine parallel operation mechanism, it has the advantages of significantly improving the efficiency of large-scale power line inspection, optimizing resource utilization and enhancing the system's adaptability.

[0037] 2. Compared with existing path planning technologies that are mostly limited to two-dimensional trajectories and do not fully consider three-dimensional terrain and energy constraints, resulting in non-optimal paths and high energy consumption, this invention adopts a path planning scheme that deeply integrates a bio-inspired behavioral model and an improved simulated annealing algorithm. It achieves defect-triggered local search by simulating the foraging behavior of boobies, and introduces an adaptive temperature scheduling and parallel multi-starting point optimization mechanism. It has the advantages of generating a globally energy-aware optimal trajectory, significantly reducing flight energy consumption, and improving adaptability to complex environments.

[0038] 3. Compared with the shortcomings of existing cooperative control methods that rely on centralized commands and lack formal verification, which are prone to collisions and insufficient coverage, this invention adopts a distributed cooperative control strategy based on Voronoi diagram partitioning and signal timing logic. It achieves autonomous motion through centroid tendency and repulsive potential field control laws, and combines spatiotemporal constraint formal verification. It has the advantages of ensuring safe collision avoidance for multiple machines, high-precision trajectory tracking, and enhanced mission reliability.

[0039] 4. Compared with the shortcomings of existing systems that lack dynamic adjustment capabilities and are difficult to cope with sudden failures and environmental changes, this invention adopts an evaluation and feedback mechanism that combines digital twin models with edge intelligence. Through real-time data synchronization and online replanning using reinforcement learning algorithms, and integrating blockchain notarization and federated learning units, it has the advantages of achieving millisecond-level anomaly response, dynamic task optimization, and data trust assurance.

[0040] 5. Compared with the traditional inspection results that rely on manual summarization and are prone to data tampering, this invention automatically generates inspection reports by combining standardized interfaces and visualization engines at the output and application layers with blockchain evidence storage technology. This has the advantages of improving data processing transparency, supporting multi-terminal interactive control, and providing complete technical evidence for quality assessment and responsibility determination. Attached Figure Description

[0041] Figure 1 The diagram shown is a structural schematic of the power inspection system based on multi-UAV collaboration of the present invention.

[0042] Figure 2 The diagram shown is a flowchart of the power line inspection method based on multi-UAV collaboration of the present invention. Detailed Implementation

[0043] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0044] Please see Figure 1 This invention provides an embodiment: a power line inspection system based on multi-UAV collaboration, comprising:

[0045] The task allocation module is used to divide the inspection area into multiple sub-regions based on the geographical location of the power facility inspection points and the drone's endurance, and to assign corresponding inspection tasks to each drone.

[0046] The path planning module is used to plan energy-sensing 3D inspection trajectories for each UAV within an assigned sub-region, based on a bio-inspired behavioral model and an improved simulated annealing algorithm.

[0047] The collaborative control module is used to achieve distributed collaboration, real-time collision avoidance, and trajectory tracking among multiple UAVs by adopting partitioned control based on Voronoi diagrams and signal timing logic specifications.

[0048] The evaluation and feedback module is used to monitor and evaluate the inspection process in real time based on the digital twin model, and to dynamically adjust the task allocation and path planning strategies based on the evaluation results.

[0049] As described above, this invention provides a power line inspection system based on multi-UAV collaboration. The technical solution is as follows: First, the task allocation module uses an improved K-means++ algorithm to intelligently divide the entire inspection area into multiple sub-regions based on the inspection point locations and UAV endurance, and assigns balanced tasks to each UAV. Second, the path planning module, based on a bio-inspired behavioral model and an improved simulated annealing algorithm incorporating 2-opt operations, generates a three-dimensional inspection trajectory for each UAV that balances global optimization and real-time energy consumption. Next, the collaborative control module employs Voronoi diagram-based partitioned control and Signal Sequential Logic (STL) specifications to achieve distributed collaborative motion, real-time dynamic collision avoidance, and precise trajectory tracking among multiple UAVs. Finally, the evaluation and feedback module uses a digital twin model synchronized with the physical system to monitor and evaluate the inspection process in real time, and utilizes reinforcement learning, federated learning, and blockchain storage units to dynamically trigger task and path replanning, ensuring system adaptability and robustness. This solution, through multi-module closed-loop collaboration, achieves fully autonomous inspection from intelligent task allocation to autonomous and safe execution and dynamic optimization adjustment. Its beneficial effects are as follows: it significantly improves the inspection efficiency and coverage of long-distance and large-scale power transmission corridors; through energy sensing paths and real-time collaborative control, it ensures operational safety and flight economy in complex terrain and multi-obstacle environments; the system has high robustness and adaptability, can cope with uncertainties such as sudden changes in UAV power supply and sudden obstacles, reduces human intervention, and provides reliable evidence for inspection data, realizing the upgrade of power inspection from the traditional manual and single-machine automation mode to the multi-machine intelligent collaborative mode.

[0050] As a preferred option, the improved K-means++ algorithm in the task allocation module has an initial cluster center selection probability that is proportional to the square of the distance from the inspection point to the existing cluster center. The probability calculation formula is as follows:

[0051] ;

[0052] in, Indicates inspection point The probability of being selected as the next initial cluster center. Indicates inspection point Distance to the nearest cluster center This represents the set of all power facilities to be inspected. This represents the checkpoints traversed during the summation process.

[0053] As described above, this invention selects initial cluster centers using a probability function. First, a first center is randomly selected, and then subsequent centers are selected iteratively based on the squared distance probability, ensuring that the initial centers are distributed in a dispersed manner. This intelligently divides the inspection area into balanced sub-regions (such as Cluster1 and Cluster2), and assigns inspection tasks with matching loads to each drone. By using probability weighting, the local optimum problem caused by traditional random initialization is avoided, achieving fairness and efficiency in task allocation. Combined with parallel operation of multiple drones, this significantly improves the overall efficiency of large-scale power inspection, while reducing the power risk caused by uneven load and enhancing the robustness and adaptability of the system.

[0054] Preferably, the bio-inspired behavioral model in the path planning module simulates the foraging behavior of boobies, including a default inspection mode and a region-restricted search mode triggered when a defect is detected. The improved simulated annealing algorithm in the path planning module incorporates a 2-opt local search operation, which randomly swaps two nodes in the path and accepts the probability of a bad solution. To escape local optima, the improved simulated annealing algorithm also includes a global optimization enhancement mechanism, which includes an adaptive temperature scheduling strategy, a parallel multi-starting point optimization mechanism, and a historical best solution memory pool. The global optimization capability is enhanced through parallel multi-threaded independent evolution and periodic solution crossover.

[0055] As described above, the path planning module of this invention integrates a bio-inspired behavioral model with an improved simulated annealing algorithm. The bio-inspired model simulates the foraging behavior of boobies, including a default inspection mode (generating an initial path based on the nearest neighbor algorithm) and a region-restricted search mode triggered by defects (such as spiral encrypted local scanning), ensuring the flexibility and coverage of the inspection. The improved simulated annealing algorithm integrates 2-opt local search operations, dynamically escaping local optima by randomly exchanging path nodes and accepting bad solutions with probability. At the same time, it introduces a global optimization enhancement mechanism, including an adaptive temperature scheduling strategy (dynamically adjusting the decay coefficient according to the acceptance rate), a parallel multi-starting point optimization mechanism (multi-threaded independent evolution), and a historical optimal solution memory pool (periodic solution crossover). Through parallel multi-threaded collaboration and elite solution reuse, the global optimization capability is significantly improved. This scheme generates energy-sensing optimal inspection trajectories in complex 3D terrain using a "global-local" hierarchical optimization framework. A bio-inspired model enhances the ability to detect defective areas with precision. An improved simulated annealing algorithm combined with an enhancement mechanism effectively avoids premature convergence in path planning, ensuring the generation of globally energy-efficient optimal paths. Parallel optimization and memory pool mechanisms significantly improve computational efficiency and solution quality, making it suitable for large-scale inspection scenarios. The overall scheme significantly improves the reliability, economy, and adaptability of inspection paths, reduces energy consumption, and ensures full-coverage inspection.

[0056] As a preferred option, the energy sensing model used in the path planning module is a composite energy consumption model, with the following expression:

[0057] ;

[0058] in, This represents the estimated total energy consumption for the drone to execute the complete inspection trajectory. This represents the total flight time required to complete the trajectory. This represents the base power of the drone at its level flight speed V. For climb and descent rates Corresponding height change power, Yaw angle change rate Corresponding turning power, Wind speed Additional power compensation caused by this.

[0059] As described above, this invention employs a composite energy consumption model as the key cost function for path optimization. This model not only considers the basic power consumption during level flight but also precisely quantifies climb / descent power, turning power, and wind speed compensation power, thereby achieving accurate prediction and optimization of total energy consumption in three-dimensional path planning. The model transforms the abstract "energy constraint" into a calculable path cost, enabling the algorithm to automatically avoid high-energy-consuming maneuvers (such as frequent sharp turns and headwind climbs) and plan the optimal trajectory for global energy conservation. By integrating flight time T with multiple power factors, the accuracy of endurance prediction is improved, effectively reducing the risk of flight mission interruption due to power estimation errors. Ultimately, while ensuring full inspection coverage, the limited power of the UAV is maximized, improving the inspection economy and operational radius of a single flight.

[0060] As a preferred option, the collaborative control module adopts a distributed control law, the expression of which is:

[0061] ;

[0062] in, This represents the resultant control input vector of UAV i. This represents the centroid orientation control term for UAV i, used to drive the UAV towards its Voronoi unit centroid. Let i represent the obstacle avoidance control term of drone i, and ,in, Represents the proportional control gain matrix. Represents the differential control gain matrix. The geometric matrix representing the Voronoi cell of drone i. This represents the centroid coordinate vector of the Voronoi cell of drone i. Let represent the current position vector of drone i. Represents the current velocity vector of drone i;

[0063] The signal timing logic in the collaborative control module is used to formally express and verify the spatiotemporal constraints in the inspection task, ensuring that the UAV always meets the requirements of safe distance, area coverage, and return-to-base charging. The general mathematical form of the signal timing logic is as follows:

[0064] ;

[0065] in, The formula for representing the timing logic task specification is as follows: Indicates a time range Internal timing operators, Represents workspace constraints. Represents obstacle avoidance. Represents the safe distance between drones. This represents the constraints for task completion. Represents return-to-base constraints, This represents the logical AND operator. Indicates a time range The final timing operator within, The timing operator indicates until.

[0066] As described above, this invention employs a deep integration of distributed control laws and signal-timing logic (STL) specifications to achieve intelligent collaboration and safety assurance among multiple UAVs. The distributed control law decomposes the combined control input of the UAVs into a centroid-oriented control term and an obstacle avoidance control term. The centroid-oriented control term, based on proportional and differential gain matrices, Voronoi unit geometric properties, and real-time position and velocity information, drives the swarm to move towards their respective regional centroids, ensuring balanced coverage. The obstacle avoidance term dynamically avoids obstacles and other UAVs by repelling potential fields. Simultaneously, the signal-timing logic formally expresses spatiotemporal constraints, using timing operators such as "always" and "finally" to force the UAVs to meet workspace limitations, obstacle avoidance, safe distance maintenance, mission completion, and return-to-home requirements throughout the mission. The beneficial effects of this scheme are: the distributed architecture avoids the single point of failure risk of centralized control, and improves the scalability and real-time response of the system; the STL specification provides verifiable security guarantees for the system, effectively preventing collisions and out-of-bounds behavior, and enhancing robustness; overall, it realizes autonomous collaboration, accurate trajectory tracking and adaptive collision avoidance of multiple machines in a highly dynamic environment, which significantly improves the operational safety and task reliability of power inspection.

[0067] As a preferred approach, the digital twin model constructed by the evaluation and feedback module is synchronized in real time with the physical drone system. Sensor data is processed through edge computing nodes, and reinforcement learning algorithms are used to replan and optimize the pre-planned trajectory online based on the real-time status of power facilities and the drone's battery level. The evaluation and feedback module includes a distributed state consensus unit, a conflict prediction and arbitration unit, and a data credibility federated evaluation unit. The distributed state consensus unit achieves decentralized state synchronization based on a lightweight blockchain consensus mechanism, the conflict prediction and arbitration unit generates a fair arbitration scheme based on a Shapley value calculation cooperative game model, and the data credibility federated evaluation unit uses a federated learning framework to collaboratively train a global credibility evaluation model.

[0068] As described above, this invention constructs a closed-loop feedback system integrating digital twins, edge intelligence, and distributed decision-making. Its core lies in real-time synchronization between the digital twin model and the physical UAV system, utilizing edge computing nodes deployed on the UAVs and ground stations to process multi-source sensor data, and employing reinforcement learning algorithms to perform online replanning and optimization of pre-planned trajectories. This module further integrates three key sub-units: a distributed state consensus unit based on a lightweight blockchain consensus mechanism, building a decentralized network within the UAV cluster to ensure that nodes can still reach consensus on key data such as task status and battery information even when communication with the central server is interrupted; a conflict prediction and arbitration unit coupled with the task allocation and collaborative control module, simulating the benefits and costs of each UAV under different task redistribution strategies based on a Shapley value cooperative game model, generating a fair arbitration scheme that maximizes the overall cluster utility and is accepted by the majority of nodes, rather than executing simple centralized commands; and a data credibility federated evaluation unit interacting with the digital twin model, using a federated learning framework to enable each UAV to detect defects locally. The results are initially assessed for credibility. Then, model gradients, rather than raw data, are encrypted and exchanged within the cluster to collaboratively train a global credibility assessment model. This model is used to identify and label low-confidence detection data in real time to trigger targeted re-inspections. The combination of digital twins and edge computing enables millisecond-level state awareness and response, significantly improving the system's real-time performance. A distributed consensus mechanism endows the system with strong fault tolerance and robustness, ensuring continuous operation in complex battlefield environments. A game theory-based arbitration scheme achieves fair and efficient dynamic task adjustment, avoiding resource waste. The federated learning framework, while ensuring data privacy, improves the overall accuracy and reliability of defect detection through intelligent cluster collaboration. Ultimately, this module upgrades the entire inspection system from a passive response system to an autonomous system with pre-event prediction, in-event arbitration, and post-event evaluation capabilities, significantly enhancing its intelligence and decision-making quality in dealing with uncertain scenarios such as sudden power fluctuations and unexpected obstacles.

[0069] Preferably, a data input layer is also included to receive multi-source heterogeneous data, including GIS geographic information of power facilities, inspection point sets, initial status of UAVs and environmental parameters, and to perform standardized preprocessing on the data. The preprocessing includes spatiotemporal benchmark unification, data cleaning and completion, format standardization and data fusion to form a comprehensive situational awareness layer for initialization of the digital twin model.

[0070] As described above, the data input layer of this invention adopts a multi-channel interface design to receive multi-source heterogeneous data, including GIS geographic information of power facilities (such as tower distribution in GeoJSON format), inspection point sets, initial state of UAVs (location, power, etc.) and environmental parameters (wind speed, obstacle tracks). Through a standardized preprocessing process—including spatiotemporal benchmark unification (using coordinate transformation algorithms to synchronize data to a custom three-dimensional engineering coordinate system and NTP time), data cleaning and completion (threshold filtering of outliers based on physical constraints, and Kalman filtering to interpolate missing data), format standardization (parsing into a unified data object within the system), and data fusion (associating UAV state, environmental field, and obstacle model)—a high-precision comprehensive situational awareness layer is formed, providing a unified and reliable data foundation for the initialization of the digital twin model. The beneficial effects of this technical solution are as follows: preprocessing eliminates heterogeneity and noise from multi-source data, improving data quality and consistency and ensuring that the digital twin model can accurately map the physical world; unified spatiotemporal reference avoids calculation errors caused by coordinate and time misalignment, enhancing the accuracy of path planning and collaborative control; data fusion and correlation achieve seamless integration of multi-dimensional information, supporting intelligent decision-making in real-time dynamic environments, thereby improving the overall robustness, adaptability, and inspection efficiency of the system.

[0071] As a preferred option, it also includes an output and application layer, which is connected to the evaluation and feedback module to output optimized trajectories, control commands, evaluation reports, early warning information and decision support schemes in a standardized format, and supports multiple application interfaces to realize the visualization and interactive control of inspection data.

[0072] As described above, the output and application layer of this invention is closely connected to the evaluation and feedback module, forming a terminal hub for information output. Its technical solution lies in encapsulating core data such as optimized trajectories, control commands, evaluation reports, early warning information, and decision support schemes into structured formats (such as JSON and XML) through standardized data interfaces. It supports multiple application interfaces such as RESTful API and WebSocket, enabling real-time push of inspection data and multi-terminal adaptation. This layer also integrates a visualization engine and interactive control module, dynamically displaying drone trajectories, power status, defect detection results, and environmental parameters through a web interface and mobile application, allowing operators to remotely intervene, adjust parameters, and replan tasks. The beneficial effects of this solution are significant: standardized output ensures cross-platform compatibility and interoperability of data, reducing system integration complexity; multi-interface support provides users with flexible access methods, improving the convenience and real-time nature of inspection management; visualization and interactive control functions enhance situational awareness, enabling maintenance personnel to intuitively monitor overall progress and quickly respond to anomalies, thereby improving the overall decision-making efficiency, operational friendliness, and system scalability of power inspection.

[0073] As a preferred option, the evaluation feedback module also includes a blockchain evidence storage unit, which is used to encrypt and store key inspection data based on blockchain technology to ensure the immutability, reliability and traceability of the data, and to provide a basis for determining responsibility in the inspection process.

[0074] As described above, the blockchain-based evidence storage unit integrated into the evaluation and feedback module employs distributed ledger technology. Its technical solution involves digitizing key data during the inspection process (such as optimization trajectories, defect detection reports, power consumption records, and environmental parameters) using a cryptographic hash algorithm. Based on a consensus mechanism, data blocks are stored in a time-stamped chain across a multi-node network, achieving decentralized encrypted evidence storage. This solution ensures data access control through asymmetric encryption and automatically triggers the evidence storage process using smart contracts, forming an immutable audit trail. Its beneficial effects are significant: the immutability of the blockchain guarantees the integrity and authenticity of inspection data, preventing human tampering and misoperation, and enhancing data credibility; traceability supports precise backtracking from task allocation to execution, facilitating fault location and responsibility determination; simultaneously, distributed storage enhances the system's anti-attack capabilities and data reliability, providing transparent and efficient technical support for power inspection quality management, compliance auditing, and insurance claims, thus improving the overall security and credibility of the system.

[0075] like Figure 2 As shown, this embodiment also provides a power line inspection method based on multi-UAV collaboration, including the following steps:

[0076] S1: Obtain the three-dimensional geographic information and inspection point set of the power facilities to be inspected, and determine the number of available drones and their energy constraints;

[0077] S2: Cluster the inspection points using the improved K-means++ algorithm and assign inspection sub-tasks to each UAV;

[0078] S3: For each UAV's mission set, the inspection path is initialized based on bio-inspired rules and optimized using an improved simulated annealing algorithm that incorporates 2-opt operations to generate an energy-sensing 3D trajectory.

[0079] S4: During the inspection process, based on the distributed collaborative control law and signal timing logic, real-time trajectory tracking, dynamic collision avoidance and collaborative coverage of multiple UAVs are achieved;

[0080] S5: Monitors the inspection status in real time through a digital twin model, assesses path tracking error, power consumption and task completion, and dynamically triggers task reassignment and trajectory replanning based on the assessment results;

[0081] S6: After the inspection task is completed, the system automatically generates a summary document containing the total inspection time, total flight distance, energy consumption analysis and defect detection report, and performs key inspection data based on blockchain technology.

[0082] As described above, this invention constructs a complete intelligent inspection closed loop through six core steps. First, the system acquires the three-dimensional geographic information of power facilities and the set of inspection points, and performs initial configuration in conjunction with the energy constraints of UAVs. Next, the improved K-means++ algorithm is used to intelligently cluster the inspection points to achieve balanced task allocation among multiple UAVs. Subsequently, the path is initialized based on bio-inspired rules and optimized using an improved simulated annealing algorithm that incorporates 2-opt operations to generate the optimal three-dimensional trajectory for energy perception. During the inspection execution phase, distributed cooperative control laws and signal timing logic are used to achieve real-time trajectory tracking, dynamic collision avoidance, and cooperative coverage among multiple UAVs. Simultaneously, a digital twin model monitors and evaluates the inspection status in real time, and dynamically triggers task reassignment and trajectory replanning based on power consumption and abnormal conditions. Finally, the system automatically generates an inspection report and uses blockchain technology to reliably store key data. This solution offers significant benefits: intelligent task allocation and parallel operation mechanisms greatly improve the inspection efficiency of large-scale power lines; energy-sensing path planning and real-time collaborative control ensure flight safety and economy in complex environments; the digital twin-driven dynamic replanning mechanism endows the system with strong robustness and adaptability, effectively responding to sudden failures and environmental changes; and blockchain evidence storage ensures the immutability and traceability of data, providing reliable support for inspection quality assessment and responsibility determination. Overall, it realizes the technological upgrade of power line inspection from the traditional manual mode to intelligent and autonomous operation.

[0083] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A power line inspection system based on multi-UAV collaboration, characterized in that: include: The task allocation module is used to divide the inspection area into multiple sub-regions based on the geographical location of the power facility inspection points and the drone's endurance, and to assign corresponding inspection tasks to each drone. The path planning module is used to plan energy-sensing 3D inspection trajectories for each UAV within an assigned sub-region, based on a bio-inspired behavioral model and an improved simulated annealing algorithm. The collaborative control module is used to achieve distributed collaboration, real-time collision avoidance, and trajectory tracking among multiple UAVs by adopting partitioned control based on Voronoi diagrams and signal timing logic specifications. The evaluation and feedback module is used to monitor and evaluate the inspection process in real time based on the digital twin model, and to dynamically adjust the task allocation and path planning strategies based on the evaluation results.

2. The power line inspection system based on multi-UAV collaboration according to claim 1, characterized in that: The improved K-means++ algorithm in the task allocation module selects the initial cluster center probability in a way that is proportional to the square of the distance from the inspection point to the existing cluster center. The probability calculation formula is as follows: ； in, Indicates inspection point The probability of being selected as the next initial cluster center. Indicates inspection point Distance to the nearest cluster center This represents the set of all power facilities to be inspected. This represents the checkpoints traversed during the summation process.

3. A power line inspection system based on multi-UAV collaboration according to claim 2, characterized in that: The bio-inspired behavioral model in the path planning module simulates the foraging behavior of boobies, including a default inspection mode and a region-restricted search mode triggered when a defect is detected. The improved simulated annealing algorithm in the path planning module incorporates a 2-opt local search operation, randomly swapping two nodes in the path and accepting the probability of a bad solution. To escape local optima, the improved simulated annealing algorithm also includes a global optimization enhancement mechanism, which includes an adaptive temperature scheduling strategy, a parallel multi-starting point optimization mechanism, and a historical best solution memory pool. The global optimization capability is enhanced through parallel multi-threaded independent evolution and periodic solution crossover.

4. A power line inspection system based on multi-UAV collaboration according to claim 3, characterized in that: The energy sensing model used in the path planning module is a composite energy consumption model, expressed as: ； in, This represents the estimated total energy consumption for the drone to execute the complete inspection trajectory. This represents the total flight time required to complete the trajectory. This represents the base power of the drone at its level flight speed V. For climb and descent rates Corresponding height change power, Yaw angle change rate Corresponding turning power, Wind speed Additional power compensation caused by this.

5. A power line inspection system based on multi-UAV collaboration according to claim 4, characterized in that: The collaborative control module uses a distributed control law, the expression of which is: ； in, This represents the resultant control input vector of UAV i. This represents the centroid orientation control term for UAV i, used to drive the UAV towards its Voronoi unit centroid. Let i represent the obstacle avoidance control term of drone i, and ,in, Represents the proportional control gain matrix. Represents the differential control gain matrix. The geometric matrix representing the Voronoi cell of drone i. This represents the centroid coordinate vector of the Voronoi cell of drone i. Let represent the current position vector of drone i. Represents the current velocity vector of drone i; The signal timing logic in the collaborative control module is used to formally express and verify the spatiotemporal constraints in the inspection task, ensuring that the UAV always meets the requirements of safe distance, area coverage, and return-to-base charging. The general mathematical form of the signal timing logic is as follows: ； in, The formula for representing the timing logic task specification is as follows: Indicates a time range Internal timing operators, Represents workspace constraints. Represents obstacle avoidance. Represents the safe distance between drones. This represents the constraints for task completion. Represents return-to-base constraints, This represents the logical AND operator. Indicates a time range The final timing operator within, The timing operator indicates until.

6. A power line inspection system based on multi-UAV collaboration according to claim 5, characterized in that: The evaluation and feedback module constructs a digital twin model that is synchronized in real time with the physical drone system. It processes sensor data through edge computing nodes and uses reinforcement learning algorithms to replan and optimize the pre-planned trajectory online based on the real-time status of power facilities and the drone's battery level. The evaluation and feedback module includes a distributed state consensus unit, a conflict prediction and arbitration unit, and a data credibility federated evaluation unit. The distributed state consensus unit achieves decentralized state synchronization based on a lightweight blockchain consensus mechanism. The conflict prediction and arbitration unit generates a fair arbitration scheme based on a Shapley value calculation cooperative game model. The data credibility federated evaluation unit uses a federated learning framework to collaboratively train a global credibility evaluation model.

7. A power line inspection system based on multi-UAV collaboration according to claim 6, characterized in that: It also includes a data input layer, which receives multi-source heterogeneous data, including GIS geographic information of power facilities, inspection point sets, initial status of UAVs and environmental parameters, and performs standardized preprocessing on the data. The preprocessing includes spatiotemporal benchmark unification, data cleaning and completion, format standardization and data fusion to form a comprehensive situational awareness layer for initialization of the digital twin model.

8. A power line inspection system based on multi-UAV collaboration according to claim 7, characterized in that: It also includes an output and application layer, which connects to the evaluation and feedback module to output optimized trajectories, control commands, evaluation reports, early warning information and decision support schemes in a standardized format, and supports multiple application interfaces to realize the visualization and interactive control of inspection data.

9. A power line inspection system based on multi-UAV collaboration according to claim 8, characterized in that: The assessment and feedback module also includes a blockchain evidence storage unit, which is used to encrypt and store key inspection data based on blockchain technology to ensure the immutability, reliability and traceability of the data, and to provide a basis for the determination of responsibility in the inspection process.

10. A power line inspection method based on multi-UAV collaboration, applied to the system described in any one of claims 1-9, characterized in that: Includes the following steps: S1: Obtain the three-dimensional geographic information and inspection point set of the power facilities to be inspected, and determine the number of available drones and their energy constraints; S2: Cluster the inspection points using the improved K-means++ algorithm and assign inspection sub-tasks to each UAV; S3: For each UAV's mission set, the inspection path is initialized based on bio-inspired rules and optimized using an improved simulated annealing algorithm that incorporates 2-opt operations to generate an energy-sensing 3D trajectory. S4: During the inspection process, based on the distributed collaborative control law and signal timing logic, real-time trajectory tracking, dynamic collision avoidance and collaborative coverage of multiple UAVs are achieved; S5: Monitors the inspection status in real time through a digital twin model, assesses path tracking error, power consumption and task completion, and dynamically triggers task reassignment and trajectory replanning based on the assessment results; S6: After the inspection task is completed, the system automatically generates a summary document containing the total inspection time, total flight distance, energy consumption analysis and defect detection report, and performs key inspection data based on blockchain technology.

Images