Multi-agent heterogeneous inspection system and method for new energy power transmission channel

Abstract

The invention relates to the technical field of intelligent operation and maintenance of new energy power transmission channels, in particular to a multi-agent heterogeneous inspection system and method of a new energy power transmission channel, a cloud twinborn management platform constructs a digital twinborn model, and generates an inspection task set and risk prediction information; the edge collaboration node receives the information, arranges tasks based on the states of the multi-type inspection agents, and generates a distribution instruction; the multi-type inspection agent responds to the instruction and collects multi-modal sensing data; the edge node fuses the processed data to generate defect alarm information and uploads the defect alarm information; and the cloud platform updates parameters of the digital twin model or the defect detection model according to the alarm information. Intelligent arrangement of inspection tasks, multi-source data fusion analysis and model dynamic optimization are realized, and the inspection efficiency and defect identification accuracy of the new energy transmission channel are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance technology for new energy power transmission channels, specifically to a multi-agent heterogeneous inspection system and method for new energy power transmission channels. Background Technology

[0002] Currently, the rapid development of new energy power generation has led to a significant increase in the number of power transmission channels. These transmission lines traverse complex terrains and variable weather conditions, making manual inspection methods inefficient and subject to significant environmental limitations, which cannot meet the real-time safety inspection needs of large-scale power transmission channels.

[0003] Existing UAV or ground vehicle inspection systems for power transmission lines mostly operate with single-type equipment and lack collaborative mechanisms among multiple agents, making it difficult to achieve task autonomy and dynamic optimization in long-distance power transmission channels. At the same time, positioning accuracy is easily affected by factors such as wind disturbance and electromagnetic interference, leading to inspection path drift and image registration errors.

[0004] In addition, existing inspection systems mostly rely on defect detection models based on a single data source, making it difficult to effectively fuse multimodal data such as visible light, infrared, ultraviolet, and lidar. When communication bandwidth is limited or edge node energy is insufficient, the system often experiences task interruption or detection delay, failing to achieve continuity and adaptive optimization of inspection tasks.

[0005] Therefore, there is an urgent need to propose a new inspection system and method with multi-subject collaboration, self-organized task allocation, anti-interference positioning fusion and energy closed-loop management, so as to realize intelligent, continuous and highly reliable inspection of new energy transmission channels. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a multi-agent heterogeneous inspection system and method for new energy power transmission channels, which addresses the shortcomings of the prior art. This system solves the technical problem that existing inspection systems rely heavily on single data source defect detection models and have difficulty in effectively fusing multimodal data such as visible light, infrared, ultraviolet and lidar.

[0007] The objective of this invention is achieved through the following technical solutions: In a first aspect, the present invention provides a multi-agent heterogeneous inspection system for new energy transmission channels, comprising: A cloud-based twin management platform is used to establish a digital twin model of the power transmission channel and generate inspection task sets and risk prediction information based on the digital twin model. At least one edge collaboration node is communicatively connected to the cloud twin management platform to receive inspection task sets and risk prediction information, and to perform task orchestration on the inspection task sets based on the status information of multiple types of inspection intelligent agents, and generate task allocation instructions. A multi-type inspection intelligent agent is communicatively connected to the at least one edge collaborative node. The inspection intelligent agent is used to respond to the task allocation instruction and collect multimodal sensing data of the power transmission channel. The at least one edge collaboration node is also used to perform fusion processing on the multimodal sensing data to generate defect alarm information, and upload the defect alarm information to the cloud twin management platform; The cloud-based twin management platform is also used to update the digital twin model and / or the model parameters used for defect detection based on the defect alarm information.

[0008] As a further improvement of the present invention, the edge collaboration node includes a task orchestration submodule, a location fusion submodule, and a communication scheduling submodule; The task orchestration submodule is used to orchestrate the inspection task set based on the state information of the multi-type inspection agents using a decentralized bidding algorithm. The positioning fusion submodule is used to tightly couple and fuse the positioning source data uploaded by each of the inspection intelligent agents to obtain the pose information of the inspection intelligent agents; The communication scheduling submodule is used to process feature stream compression of the multimodal sensing data and / or migration of the inspection task when the network performance is detected to be lower than a preset threshold.

[0009] As a further improvement of the present invention, the positioning source data includes at least two of the following: satellite positioning data, inertial measurement unit data, ultra-wideband ranging data, and visual anchor data; The positioning fusion submodule is specifically used to fuse the positioning source data using the extended Kalman filter algorithm, and to adaptively adjust the geometric constraint weights in the fusion process when environmental interference is detected.

[0010] As a further improvement of the present invention, each of the inspection intelligent agents is equipped with a multimodal detection device, which includes at least two of the following: a visible light camera unit, an infrared thermal imaging unit, an ultraviolet corona detection unit, and a lidar ranging unit.

[0011] As a further improvement of the present invention, each of the inspection intelligent agents is provided with an energy state monitoring unit, the energy state monitoring unit comprising: Receive energy status data, which includes at least one of real-time monitoring of battery power, payload power, and mission progress; Predict the remaining battery life of the corresponding inspection agent based on the energy state data; When the remaining battery life is lower than a preset threshold, the corresponding inspection agent is controlled to issue a task transfer request, and another inspection agent that meets the takeover conditions is scheduled to take over the inspection task.

[0012] As a further improvement of the present invention, it also includes an energy supply device deployed along the power transmission channel; When the inspection agent relinquishes its inspection task due to energy status, the at least one edge collaboration node is also used to guide the inspection agent to move to the target energy replenishment device for energy replenishment.

[0013] As a further improvement of the present invention, the cloud-based twin management platform includes a model aggregation module; The edge collaboration node is also used to train the locally deployed defect detection model based on the defect detection data acquired locally, obtain the model gradient data, and upload it to the model aggregation module; The model aggregation module is used to aggregate the model gradient data from different edge collaborative nodes through a federated learning mechanism, update the global defect detection model, and send the updated model parameters back to each edge collaborative node.

[0014] As a further improvement of the present invention, the edge collaboration node is also used for: The multimodal perception data collected by each of the inspection intelligent agents is initially detected to identify suspected defect features; If the suspected defect features are identified, then different types of inspection agents in the inspection agent are triggered to conduct collaborative review and verification. Perform time-series consistency verification on the review and verification results, and generate valid defect alarm information with a preset length of evidence chain.

[0015] Secondly, the present invention provides a multi-agent heterogeneous inspection method for new energy transmission channels, comprising: The cloud-based twin management platform generates inspection task sets based on the digital twin model of the power transmission channel; At least one edge collaboration node performs task orchestration on the inspection task set based on the state information of multiple types of inspection agents, obtains task allocation instructions, and sends them to the multiple types of inspection agents. The multi-type inspection intelligent agents respond to the task allocation instructions, execute inspection tasks, and collect multimodal perception data; The multimodal sensing data is fused using edge collaborative nodes to generate defect alarm information and upload it to the cloud twin management platform; Update the digital twin model and / or the model parameters used for defect detection based on the defect alarm information.

[0016] As a further improvement of the present invention, the edge collaborative node, based on the state information of multiple agents, performs task orchestration on the inspection task set, including: Obtain the status information of each inspection agent among the multiple types of inspection agents, wherein the status information includes at least two of the following: location information, energy status information, payload capacity information, and communication quality information; Based on the state information and the task attribute information of each task in the inspection task set, the output value of each inspection agent for different tasks is calculated by a decentralized bidding algorithm. Based on the principle of minimum output value, the inspection tasks are assigned to the corresponding inspection agents, forming the task assignment instructions.

[0017] Thirdly, the present invention provides a computer-readable storage medium for storing one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform the above-described multi-agent heterogeneous inspection method for new energy transmission channels.

[0018] Fourthly, the present invention provides a computing device, comprising: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include steps for performing the steps in the above-described multi-agent heterogeneous inspection method for new energy transmission channels.

[0019] The beneficial effects of this invention are as follows: This invention provides a multi-agent heterogeneous inspection system for new energy transmission channels. In this inspection system, the cloud-based twin management platform generates inspection task sets and risk prediction information by establishing a digital twin model of the transmission channel, realizing dynamic planning of inspection tasks and forward-looking prediction of risks; edge collaborative nodes intelligently orchestrate the inspection task sets based on the status information of multiple types of inspection agents and generate task allocation instructions, improving the rationality of task allocation and execution efficiency; multiple types of inspection agents respond to instructions and collect multimodal perception data, including drones, inspection vehicles, and conductor robots, achieving multi-dimensional and comprehensive inspection data collection based on multiple types of inspection agents; edge collaborative nodes fuse the multimodal perception data to generate defect alarm information and upload it, improving the accuracy of defect identification and the timeliness of alarms; the cloud-based twin management platform updates the parameters of the digital twin model and defect detection model based on the defect alarm information, and through multi-level collaboration of cloud, edge, and terminal, ultimately achieves a synergistic improvement in the inspection efficiency, defect identification accuracy, and system adaptability of new energy transmission channels. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of a multi-agent heterogeneous inspection system for a new energy transmission channel according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating a multi-agent heterogeneous inspection method for a new energy transmission channel according to an embodiment of the present invention. Figure 3 This is an internal structural diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0022] To make the objectives and technical solutions of this invention clearer and easier to understand, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0023] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments. The described embodiments are only some embodiments of the present invention, and not all embodiments.

[0024] Example 1 This embodiment provides a multi-agent heterogeneous inspection system for new energy power transmission channels. The inspection system includes: A cloud-based digital twin management platform is used to establish a digital twin model of the power transmission channel and generate inspection task sets and risk prediction information based on the digital twin model. At least one edge collaboration node is connected to the cloud-based digital twin management platform to receive the inspection task sets and risk prediction information, and to orchestrate the inspection task sets based on the status information of multiple types of inspection agents, generating task allocation instructions. Multiple types of inspection agents, including at least two of aerial drones, ground inspection vehicles, and conductor climbing robots, are connected to the at least one edge collaboration node to respond to task allocation instructions and collect multimodal sensing data of the power transmission channel. At least one edge collaboration node is also used to fuse the multimodal sensing data to generate defect alarm information and upload the defect alarm information to the cloud-based digital twin management platform. The cloud-based digital twin management platform is also used to update the digital twin model and / or model parameters used for defect detection based on the defect alarm information.

[0025] This embodiment utilizes a digital twin model established through a cloud-based twin management platform to provide accurate virtual mapping and risk prediction benchmarks for inspection operations. Edge collaboration nodes dynamically orchestrate tasks based on real-time perception of the status of various types of inspection agents, significantly improving the overall system's task execution efficiency and resource utilization. The execution layer, composed of heterogeneous agents such as aerial drones, ground inspection vehicles, and conductor-climbing robots, can collaborate under unified instructions, fully leveraging their respective advantages in spatial coverage and close-range detection, thereby achieving efficient and complementary data acquisition of the complex structure and full-domain space of the power transmission channel. Edge collaboration nodes fuse the multimodal perception data collected by the agents to generate defect alarm information. This process integrates the characteristics of different sensor data, enhancing the accuracy and reliability of defect identification and reducing false alarms and false negatives. The cloud-based twin management platform updates the parameters of the digital twin model and / or defect detection model based on the uploaded defect alarm information, forming a data-driven closed loop. This enables the system to continuously learn and evolve, constantly adapting to environmental changes and new defect patterns, ultimately achieving a synergistic improvement in inspection accuracy, efficiency, and system adaptability.

[0026] The edge collaborative node includes a task orchestration submodule, a positioning fusion submodule, and a communication scheduling submodule. The task orchestration submodule orchestrates inspection tasks based on the state information of multiple types of inspection agents using a decentralized bidding algorithm. The positioning fusion submodule tightly fuses the positioning source data uploaded by each inspection agent to obtain the pose information of the inspection agent. The communication scheduling submodule handles feature stream compression of multimodal perception data and / or the migration of inspection tasks when network performance is detected to be below a preset threshold. The positioning source data includes at least two of the following: satellite positioning data, inertial measurement unit data, ultra-wideband ranging data, and visual anchor data. Specifically, the positioning fusion submodule uses an extended Kalman filter algorithm to fuse the positioning source data and adaptively adjusts the geometric constraint weights during the fusion process when environmental interference is detected.

[0027] In order to acquire multi-source data, each of the inspection intelligent agents is equipped with a multimodal detection device, which includes at least two of the following: a visible light camera unit, an infrared thermal imaging unit, an ultraviolet corona detection unit, and a lidar ranging unit.

[0028] Each of the aforementioned inspection agents is equipped with an energy status monitoring unit, which includes: The system receives energy status data, which includes at least one of real-time battery power, payload power, and task progress. Based on the energy status data, it predicts the remaining runtime of the corresponding inspection agent. When the remaining runtime is lower than a preset threshold, it controls the corresponding inspection agent to issue a task relinquishment request and schedules another inspection agent that meets the takeover conditions to take over the inspection task. This ensures that the task can be completed on time.

[0029] This embodiment also includes energy replenishment devices deployed along the power transmission channel; when the inspection agent relinquishes its inspection task due to energy status, at least one edge collaboration node is also used to guide the inspection agent to move to the target energy replenishment device for energy replenishment.

[0030] The cloud-based twin management platform includes a model aggregation module; the edge collaboration nodes are also used to train the locally deployed defect detection model based on locally acquired defect detection data, obtain model gradient data, and upload it to the model aggregation module; the model aggregation module is used to aggregate the model gradient data from different edge collaboration nodes through a federated learning mechanism, update the global defect detection model, and send the updated model parameters back to each edge collaboration node.

[0031] Example 2 This embodiment provides a multi-agent heterogeneous inspection system for new energy power transmission channels, such as... Figure 1 As shown, the system adopts a cloud-edge-device collaborative architecture, consisting of a cloud twin management platform, edge collaborative nodes, and multiple types of inspection intelligent agents.

[0032] The cloud-based twin management platform is used to establish a digital twin model of the power transmission channel, and to virtually model and map the operating parameters of structural objects such as lines, towers, fittings, and insulators. Edge collaboration nodes are deployed in the field or substation area and are responsible for task allocation, data fusion and local anomaly handling. The inspection intelligent agents include aerial drones, ground inspection vehicles and conductor climbing robots, which realize information interaction and task collaboration through self-organizing networks.

[0033] Each intelligent agent is equipped with an anti-interference localization and attitude fusion unit, an energy state monitoring unit, and a multimodal detection device, enabling autonomous inspection of power transmission channels under different terrain and weather conditions. The system automatically generates a task map based on cloud-based risk prediction results, and performs decentralized bidding scheduling through edge nodes to achieve dynamic optimal task allocation.

[0034] In addition, the system supports energy status monitoring and task transfer functions. When the power or task load of a certain intelligent agent is lower than the threshold, the task can be automatically transferred to a nearby node and energy can be replenished in the battery swapping compartment at the base or top of the tower, forming an energy closed loop and task continuity.

[0035] To make the technical solution of the present invention clearer, the following will be combined with Figure 1 The system's structure and the collaborative relationships between its functional modules are described in detail. Each module can be implemented as an independent hardware unit or logically partitioned using software algorithms on the same processing platform. Through orderly interaction and data flow between modules, a closed-loop control system is achieved, encompassing the entire process from inspection task generation, location fusion, data acquisition, analysis and decision-making to energy management. The following section, using a structural diagram, describes each component module and its function in detail.

[0036] like Figure 1 As shown, the multi-agent heterogeneous inspection system for new energy transmission channels in this embodiment mainly includes: a cloud-based twin management platform, edge collaborative nodes, a group of inspection agents, a communication and timing system, and an energy management and replenishment device. The modules interact and coordinate data and tasks through encrypted communication protocols and synchronized timing signals.

[0037] The cloud-based twin management platform serves as the core decision-making and overall management center of the system, enabling virtual mapping of power transmission channels, task scheduling, risk prediction, and model self-evolution updates.

[0038] The platform includes the following three core functional modules: The digital twin modeling module constructs digital twins of key components such as transmission lines, towers, hardware, conductors, and insulators based on GIS geographic information data, transmission line structural parameters, and historical operating status data. The twin model maintains parameter mapping consistency with the field equipment and reflects changes in the line's operating status through dynamic simulation.

[0039] The risk prediction module utilizes multimodal inspection data and historical defect samples received from the cloud to calculate risk indicators such as conductor sag, icing thickness, and insulation aging through deep neural networks. The prediction results are mapped to a twin model in probabilistic form, enabling visualization and dynamic alarm of line node-level risks.

[0040] The model aggregation module executes a federated aggregation algorithm across multiple sites, weighting and fusing the gradients of the detection models uploaded by edge nodes to form a globally optimized detection model. The updated model parameters are then fed back to each edge node, enabling the system to periodically learn itself.

[0041] Through the collaboration of the above modules, the cloud platform can realize closed-loop decision control from virtual modeling to policy feedback, enabling the inspection system to maintain adaptive operation under different environmental and load conditions.

[0042] Edge collaboration nodes are deployed along power transmission lines or in substation areas to receive cloud commands, execute task orchestration, and perform data fusion.

[0043] This node contains the following three sub-modules: The task orchestration submodule uses an improved decentralized bidding algorithm to allocate tasks based on the inspection task map and multi-agent capability descriptions distributed from the cloud. Each agent calculates a bidding function based on the task target distance, risk level, and remaining energy, and edge nodes autonomously allocate tasks according to the minimum bid principle.

[0044] The positioning fusion submodule integrates RTK satellite positioning, IMU inertial navigation, UWB ranging, and visual anchor signals, employing an extended Kalman filter algorithm for tightly coupled calculations. When anomalies such as wind speed or electromagnetic interference are detected, the geometric constraint weights are automatically increased to suppress drift errors, outputting high-precision pose information.

[0045] The communication scheduling submodule is used to coordinate communication tasks between multiple agents and the cloud. When communication bandwidth decreases or latency increases, the system automatically activates feature stream compression and task migration mechanisms, retaining only key feature data for reporting, thereby maintaining stable and real-time data transmission.

[0046] Edge collaboration nodes effectively reduce cloud computing load and improve system robustness through localized task control and data preprocessing.

[0047] The inspection intelligent agent group consists of aerial drones, ground inspection vehicles, and conductor climbing robots. Each type of agent has different operational characteristics and can work together to complete the full-area inspection of the power transmission channel.

[0048] The aerial drone is responsible for inspecting the top of the tower and the section it crosses. It is equipped with a visible light camera, an infrared thermal imager and an ultraviolet corona detection device, and supports automatic path planning and hovering detection.

[0049] Ground inspection vehicles are mainly used for inspection tasks of tower foundations, conductor accessories and grounding lead areas. They are equipped with lidar and multi-modal cameras and have obstacle avoidance and autonomous path planning functions.

[0050] The wire climbing robot is responsible for close-range inspection of the middle section of the wire and the hardware connection points. Equipped with infrared and acoustic detection modules, it can achieve automatic attitude leveling and anti-sway control in environments with changing wind speeds.

[0051] All agents are equipped with built-in anti-disturbance localization and attitude fusion units, multimodal detection devices, and energy state monitoring units, which can automatically perform task migration and fault self-checks under the command of edge nodes.

[0052] The communication and timing system is used to enable data transmission and time synchronization between intelligent agents and between edge nodes and the cloud.

[0053] The system adopts an encrypted wireless communication protocol and a high-precision time synchronization mechanism, and supports three transmission modes: 5G, Mesh and satellite communication.

[0054] When the communication quality of any path deteriorates, the system automatically switches to the backup link and recalibrates the clock to ensure that multi-source data is processed synchronously under the same time reference.

[0055] Energy management and recharging devices are used to ensure the continuity of inspection tasks.

[0056] The system monitors the agent's battery level, power consumption, and task progress in real time, and calculates the remaining time based on a predictive model. When energy is insufficient, it issues a task relinquishment request, and a nearby agent automatically takes over the task.

[0057] The original intelligent agent is powered by an inductive power supply base at the base or top of the tower, with a power supply of up to 10 kW and a battery swapping time of no more than 30 seconds.

[0058] This mechanism enables closed-loop energy control for inspection tasks, preventing inspection interruptions due to insufficient power.

[0059] The above modules together constitute a complete control system from cloud modeling to agent execution.

[0060] The system achieves continuous and intelligent inspection tasks through a cycle of cloud-based risk prediction, edge node task orchestration, multi-agent collaborative execution, and energy closed-loop feedback.

[0061] When the detection model is updated at the edge node, the cloud twin synchronously performs model parameter correction, realizing the system's self-learning and self-evolution.

[0062] Example 3 This embodiment provides a multi-agent heterogeneous inspection method for new energy transmission channels based on the system implemented in Embodiment 1. For example... Figure 2 As shown, the method includes the following steps: the cloud-based twin management platform generates an inspection task set based on the digital twin model of the power transmission channel; at least one edge collaborative node performs task orchestration on the inspection task set based on the state information of multiple types of inspection agents, obtains task allocation instructions, and sends them to the multiple types of inspection agents; the multiple types of inspection agents respond to the task allocation instructions, execute inspection tasks, and collect multimodal sensing data; the edge collaborative node performs fusion processing on the multimodal sensing data, generates defect alarm information, and uploads it to the cloud-based twin management platform; based on the defect alarm information, the digital twin model and / or the model parameters used for defect detection are updated.

[0063] For details on the specific execution process, calculation logic, and coordination mechanism of each step, please refer to the detailed description section of the aforementioned Example 2.

[0064] To better understand the working principle of this invention, the following will be combined with Figure 2The execution flow of the inspection method of the present invention is described in detail. This method relies on the hardware and software structure of the system described in Example 1, and achieves full-process task autonomy, data fusion, and model self-evolution through the collaboration of cloud, edge, and multi-agent terminals. Each step can be executed sequentially according to a set time sequence, or dynamically adjusted based on environmental conditions or communication requirements. The main steps of this method and their implementation logic are described below to fully reveal the technical essence of the present invention.

[0065] like Figure 2 As shown, the multi-agent heterogeneous inspection method for new energy transmission channels provided in this embodiment is based on the system structure described in Embodiment 1, and mainly includes the following steps: Step S1: Task Generation and Multi-Agent Autonomous Orchestration Before the inspection task begins, the cloud-based twin management platform generates an inspection task set based on the risk prediction results of the digital twin and the line operation and maintenance plan.

[0066] Each task unit includes the target tower number, geographical coordinates, inspection priority, expected operation time, and risk weight information.

[0067] Edge collaboration nodes periodically broadcast the current multi-agent capability vector set (C={location, power, sensor type, payload margin, communication quality, health}) and execute an improved decentralized bidding algorithm.

[0068] Each agent calculates its bid function based on the distance to the task objective, the task risk coefficient, the time cost, and the remaining energy.

[0069] in, For the target distance, To estimate the execution time, For risk coefficient, The percentage of remaining energy. These are the weighting coefficients.

[0070] Edge nodes allocate tasks based on the lowest bid principle, generating the optimal matching scheme. When a node experiences a communication interruption or equipment failure, its tasks automatically enter a bidding reallocation process, achieving task autonomy and dynamic recovery.

[0071] This step enables self-organization and adaptive task orchestration among multiple entities, ensuring optimal coverage of inspection tasks in both spatial and temporal dimensions.

[0072] Step S2: Anti-interference localization fusion and spatiotemporal alignment During task execution, the system performs multi-source fusion of the positioning and attitude information of each intelligent agent to ensure the spatial accuracy of the inspection path and detection data.

[0073] Each agent collects RTK or PPP satellite positioning data, UWB ranging signals, IMU inertial information, and visual anchor recognition results.

[0074] Edge collaborative nodes employ an extended Kalman filter algorithm to achieve tight coupling and fusion, and automatically increase the weight of geometric constraint terms to suppress drift in high wind speed or electromagnetic interference environments.

[0075] The clock signals of all agents and edge nodes are uniformly calibrated to the time reference of the timing layer, forming a cross-agent time synchronization and spatial alignment system.

[0076] The pose accuracy of the system output is stable within 10 cm, providing a high-precision reference frame for subsequent multimodal detection.

[0077] This step achieves anti-interference positioning and spatiotemporal integrated calibration through multi-source data fusion and dynamic weight adjustment.

[0078] Step S3: Multimodal Detection and Cascaded Evidence Generation Each intelligent agent collects multimodal data such as visible light, infrared, ultraviolet and lidar according to the task requirements, and processes it synchronously through edge nodes.

[0079] The edge nodes first perform preliminary detection based on single-frame data to identify abnormal features such as conductor damage, loose fittings, insulator cracks, and corona discharge.

[0080] When a suspected defect is detected, the system automatically triggers a cross-subject verification mechanism: when the thermal image of the aerial drone is abnormal, the climbing robot performs acoustic detection; when the conductor is geometrically distorted, the ground inspection vehicle performs three-dimensional re-measurement; when the corona is abnormal, ultraviolet and visible light images are jointly verified.

[0081] After the detection results are verified for temporal consistency, a defect evidence chain is generated. A valid alarm is output only when the evidence chain length is ≥3 and the consistency confidence is higher than the threshold.

[0082] The final results are uploaded to a cloud-based twin platform for risk correction and model training.

[0083] This step enables cross-subject multimodal fusion detection and highly reliable evidence chain generation, significantly reducing the false detection rate and false negative rate.

[0084] Step S4: Energy State Prediction and Mission Transfer During the inspection process, the system continuously monitors the energy status of each intelligent agent, including battery level, load power, and task progress.

[0085] The energy prediction module calculates the remaining battery life based on the current energy consumption curve and the mission distance. When the battery life is less than the 20-minute threshold, the system automatically issues a mission handover request.

[0086] The neighboring agent comprehensively assesses whether to take over the task based on geographical distance, communication quality, and energy reserves. After the original agent exits the inspection, it performs a rapid energy replenishment operation at the tower base induction energy replenishment station or the tower top battery swapping compartment, with a replenishment power of 10 kW and a battery swapping time controlled within 30 seconds.

[0087] This step, through energy prediction and automatic transfer mechanisms, ensures the continuity of inspection tasks and closed-loop energy operation, avoiding task interruption and resource waste.

[0088] Step S5: Cloud-based twin collaboration and model aggregation update After the inspection is completed, the test results and energy data are uploaded to the cloud-based twin management platform.

[0089] The cloud-based twin compares inspection data with the virtual model and performs parameter corrections, including key parameters such as conductor sag, icing thickness, thermoelectric coupling, and hardware degradation rate.

[0090] The risk prediction module calculates the probability of anomaly propagation based on the LSTM network, and automatically adjusts the inspection frequency of the corresponding node when the risk value exceeds 0.6.

[0091] Each edge node periodically uploads the gradient of the lightweight detection model, and differential privacy-preserving federated aggregation computation is performed in the cloud to generate new detection model parameters and send them back to the edge nodes, thereby realizing continuous optimization of the detection algorithm and system self-evolution.

[0092] When communication bandwidth is limited or latency increases, the system automatically activates feature stream compression and task migration mechanisms, uploading only key feature vectors and offloading computationally intensive tasks to idle nodes for execution, thereby maintaining the system's real-time performance and robustness.

[0093] This step achieves closed-loop optimization of cloud-edge-device, enabling the system to maintain high efficiency, stability, and self-learning capabilities in multi-task collaborative scenarios.

[0094] Through the coordinated execution of steps S1 to S5 above, the present invention realizes the entire process of multi-agent heterogeneous collaborative inspection of new energy transmission channels.

[0095] This method can perform task self-organization, high-precision positioning alignment, multimodal defect identification, energy closed-loop management, and model self-evolution optimization under complex terrain and dynamic weather conditions.

[0096] Compared with traditional single-unit inspection methods, this method has significant advantages in terms of inspection coverage, identification accuracy and system sustainability, and is suitable for intelligent safety operation and maintenance of large-scale new energy transmission channels.

[0097] Example 4 In another embodiment of the present invention, a computer-readable storage medium is provided as a storage component within a terminal device, the function of which is to store programs and data. It should be noted that the computer-readable storage medium here encompasses not only the built-in storage components of the terminal device but also extended storage components supported by the device. Essentially, it is a tangible medium capable of containing or storing programs that can be invoked by or in conjunction with an instruction execution system, device, or apparatus. This storage medium provides storage areas for the terminal's operating system and stores one or more instructions suitable for processor loading and execution, which can constitute one or more computer programs containing program code.

[0098] Specifically, examples of computer-readable storage media (a non-exclusive list) include: electrical connections with one or more wires, portable disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, optical fibers, portable optical disc read-only memory, optical storage devices, magnetic storage devices, or any reasonable combination of the above types.

[0099] The storage medium may also include data signals propagated as part of a baseband portion or a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any reasonable combination of both. Furthermore, computer-readable storage medium may also refer to other readable media besides conventional readable storage media, capable of sending, propagating, or transmitting programs for use or operation by an instruction execution system, apparatus, or device. Program code on the storage medium can be transmitted via any suitable medium, including but not limited to wireless, wired, optical fiber, or any reasonable combination thereof.

[0100] The program code used to implement the operations of this invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C. The execution modes of the program code include: running entirely on the user's computing device, running partially on the user's device as a standalone software package, running partially in a distributed manner on both the user's device and a remote computing device, or running entirely on a remote computing device or server. When a remote computing device is involved, the device can be connected to the user's computing device via any type of network such as a local area network (LAN) or a wide area network (WAN), or connected to an external computing device via the Internet through an Internet service provider.

[0101] The processor is capable of loading and executing one or more instructions stored in a computer-readable storage medium to implement the corresponding steps of the multi-agent heterogeneous inspection method for new energy transmission channels described in Embodiment 1.

[0102] Example 5 Figure 3This is a schematic diagram of a computer device provided according to an embodiment of the present invention.

[0103] Please see Figure 3 The terminal device is a computer device. In this embodiment, the computer device 60 includes a processor 61, a memory 62, and a computer program 63 stored in the memory 62 and executable on the processor 61. When executed by the processor 61, the computer program 63 implements the multi-agent heterogeneous inspection method for the new energy transmission channel in this embodiment. To avoid repetition, these details are not elaborated here. Alternatively, when executed by the processor 61, the computer program 63 implements the functions of each model / unit in the computational system constituting the multi-agent heterogeneous inspection system for the new energy transmission channel in this embodiment. To avoid repetition, these details are not elaborated here.

[0104] Computer device 60 can be a desktop computer, laptop, handheld computer, cloud server, or other computing device. Computer device 60 may include, but is not limited to, a processor 61 and a memory 62. Those skilled in the art will understand that... Figure 3 This is merely an example of computer device 60 and does not constitute a limitation on computer device 60. It may include more or fewer components than shown, or combine certain components, or different components. For example, computer device may also include input / output devices, network access devices, buses, etc.

[0105] The processor 61 may be a central processing unit (CPU), or other general-purpose processors, CPUs, graphics processing units (GPUs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, quantum computing-based data processing logic units, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0106] The memory 62 can be an internal storage unit of the computer device 60, such as a hard disk or RAM of the computer device 60. The memory 62 can also be an external storage device of the computer device 60, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on the computer device 60.

[0107] Furthermore, the memory 62 may include both internal storage units of the computer device 60 and external storage devices. The memory 62 is used to store computer programs and other programs and data required by the computer device. The memory 62 can also be used to temporarily store data that has been output or will be output.

[0108] Any references to memory, databases, or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (Read-Only Memory). Memory includes ROM, magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0109] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

Claims

1. A multi-agent heterogeneous inspection system for new energy power transmission channels, characterized in that, include: A cloud-based twin management platform is used to establish a digital twin model of the power transmission channel and generate inspection task sets and risk prediction information based on the digital twin model. At least one edge collaboration node is communicatively connected to the cloud twin management platform to receive inspection task sets and risk prediction information, and to perform task orchestration on the inspection task sets based on the status information of multiple types of inspection intelligent agents, and generate task allocation instructions. A multi-type inspection intelligent agent is communicatively connected to at least one of the aforementioned edge collaborative nodes. The inspection intelligent agent is used to respond to the task allocation instruction and collect multimodal sensing data of the power transmission channel. The at least one edge collaboration node is also used to perform fusion processing on the multimodal sensing data to generate defect alarm information, and upload the defect alarm information to the cloud twin management platform; The cloud-based twin management platform is also used to update the digital twin model and / or the model parameters used for defect detection based on the defect alarm information.

2. The multi-agent heterogeneous inspection system for new energy transmission channels according to claim 1, characterized in that, The edge collaboration node includes a task orchestration submodule, a location fusion submodule, and a communication scheduling submodule; The task orchestration submodule is used to orchestrate the inspection task set based on the state information of the multi-type inspection agents using a decentralized bidding algorithm. The positioning fusion submodule is used to tightly couple and fuse the positioning source data uploaded by each of the inspection intelligent agents to obtain the pose information of the inspection intelligent agents; The communication scheduling submodule is used to process feature stream compression of the multimodal sensing data and / or migration of the inspection task when the network performance is detected to be lower than a preset threshold.

3. The multi-agent heterogeneous inspection system for new energy transmission channels according to claim 2, characterized in that, The positioning source data includes at least two of the following: satellite positioning data, inertial measurement unit data, ultra-wideband ranging data, and visual anchor data. The positioning fusion submodule is specifically used to fuse the positioning source data using the extended Kalman filter algorithm, and to adaptively adjust the geometric constraint weights in the fusion process when environmental interference is detected.

4. The multi-agent heterogeneous inspection system for new energy transmission channels according to claim 1, characterized in that, Each of the aforementioned inspection intelligent agents is equipped with a multimodal detection device, which includes at least two of the following: a visible light camera unit, an infrared thermal imaging unit, an ultraviolet corona detection unit, and a lidar ranging unit.

5. The multi-agent heterogeneous inspection system for new energy transmission channels according to claim 4, characterized in that, Each of the aforementioned inspection intelligent agents is equipped with an energy status monitoring unit, which includes: Receive energy status data, which includes at least one of real-time monitoring of battery power, payload power, and mission progress; Predict the remaining battery life of the corresponding inspection agent based on the energy state data; When the remaining battery life is lower than a preset threshold, the corresponding inspection agent is controlled to issue a task transfer request, and another inspection agent that meets the takeover conditions is scheduled to take over the inspection task.

6. The multi-agent heterogeneous inspection system for new energy transmission channels according to claim 5, characterized in that, It also includes energy replenishment devices deployed along the power transmission channels; When the inspection agent relinquishes its inspection task due to energy status, the at least one edge collaboration node is also used to guide the inspection agent to move to the target energy replenishment device for energy replenishment.

7. The multi-agent heterogeneous inspection system for new energy transmission channels according to claim 1, characterized in that, The cloud-based twin management platform includes a model aggregation module; The edge collaboration node is also used to train the locally deployed defect detection model based on the defect detection data acquired locally, obtain the model gradient data, and upload it to the model aggregation module; The model aggregation module is used to aggregate the model gradient data from different edge collaborative nodes through a federated learning mechanism, update the global defect detection model, and send the updated model parameters back to each edge collaborative node.

8. The multi-agent heterogeneous inspection system for new energy transmission channels according to claim 1, characterized in that, The edge collaboration node is also used for: The multimodal perception data collected by each of the inspection intelligent agents is initially detected to identify suspected defect features; If the suspected defect features are identified, then different types of inspection agents in the inspection agent are triggered to conduct collaborative review and verification. Perform time-series consistency verification on the review and verification results, and generate valid defect alarm information with a preset length of evidence chain.

9. A multi-agent heterogeneous inspection method for new energy transmission channels, characterized in that, include: The cloud-based twin management platform generates inspection task sets based on the digital twin model of the power transmission channel; At least one edge collaboration node performs task orchestration on the inspection task set based on the state information of multiple types of inspection agents, obtains task allocation instructions, and sends them to the multiple types of inspection agents. The multi-type inspection intelligent agents respond to the task allocation instructions, execute inspection tasks, and collect multimodal perception data; The multimodal sensing data is fused using edge collaborative nodes to generate defect alarm information and upload it to the cloud twin management platform; Update the digital twin model and / or the model parameters used for defect detection based on the defect alarm information.

10. The multi-agent heterogeneous inspection method for new energy transmission channels according to claim 9, characterized in that, The edge collaboration node, based on the state information of multiple agents, orchestrates the inspection task set, including: Obtain the status information of each inspection agent among the multiple types of inspection agents, wherein the status information includes at least two of the following: location information, energy status information, payload capacity information, and communication quality information; Based on the state information and the task attribute information of each task in the inspection task set, the output value of each inspection agent for different tasks is calculated by a decentralized bidding algorithm. Based on the principle of minimum output value, the inspection tasks are assigned to the corresponding inspection agents, forming the task assignment instructions.

Images