A power distribution network intelligent inspection method, system and computer program product

Abstract

This invention discloses an intelligent inspection method, system, and computer program product for distribution networks, belonging to the field of distribution network operation and maintenance technology. The method constructs and optimizes a dual proprietary recognition model through a lightweight network improved by an attention mechanism combined with knowledge distillation technology; deploys dedicated equipment integrating 5G and AI edge computing, and establishes an algorithm optimization and update mechanism; constructs an automatic drone nest integrating this edge computing equipment, enabling inspection drones to dock, recharge, and communicate; completes multi-machine, multi-task collaborative scheduling algorithm based on the road network and employing integrated spatiotemporal isolation and emergency handling mechanisms; and builds an inspection application module based on the power grid management platform to achieve inspection data sharing and system integration. This invention realizes the intelligence and automation of distribution network inspection, improves inspection identification efficiency, operational safety and flexibility, breaks down information silos, ensures data security, and significantly improves the level of distribution network operation and maintenance management.

Description

Technical Field

[0001] This invention relates to the field of power distribution network operation and maintenance technology, specifically to a power distribution network intelligent inspection method, system, and computer program product. Background Technology

[0002] The distribution network is a crucial link in the power system connecting the transmission network and the user side, and its safe and stable operation is directly related to social production, daily life, and public safety. With the development of my country's social economy and the advancement of urbanization, the scale of the distribution network continues to expand, the number of engineering projects within the line protection zone increases, and the hidden dangers of external damage become increasingly prominent, placing higher demands on the operation and maintenance of the distribution network.

[0003] Inspection and maintenance are key means to ensure the safe operation of the distribution network. Traditional inspections mainly rely on manual on-site inspections and are supplemented by routine tests. This has problems such as low inspection efficiency, high labor intensity, reliance on manual data processing, and delayed fault identification, and can no longer meet the needs of efficient operation and maintenance of the distribution network.

[0004] To address these pain points, drone inspection technology, with its advantages of high mobility and wide coverage, has been gradually applied to power distribution network operation and maintenance. However, existing technologies still have significant limitations: they lack dedicated intelligent identification models for power distribution network defects and external damage hazards, resulting in insufficient identification accuracy and speed; data processing relies on the cloud, leading to high transmission pressure and poor real-time performance; they lack multi-drone collaborative scheduling mechanisms, resulting in unreasonable task allocation and slow emergency response; take-off, landing, and power replenishment rely on manual labor, resulting in low automation levels and equipment utilization; and inspection data is isolated from various power grid management systems, hindering collaborative application capabilities. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method, system and computer program product for intelligent inspection of power distribution networks, so as to realize efficient and intelligent operation and maintenance of power distribution networks.

[0006] To address the aforementioned technical problems, this invention provides an intelligent inspection method for power distribution networks, comprising: Step S1: Construct equipment problem identification model and construction external damage problem identification model based on lightweight network improved by attention mechanism and knowledge distillation technology, and optimize model training by using multi-source data augmentation method and noise augmentation strategy; Step S2: Deploy a power distribution network AI edge computing device that integrates 5G network and AI edge computing to realize real-time video transmission for inspection, and configure algorithm optimization and update mechanisms; Step S3: Construct an automated drone nest to provide docking, recharging and communication services for inspection drones, wherein the automated drone nest integrates the power distribution network AI edge computing device; Step S4: Based on the road network and using a multi-machine multi-task collaborative scheduling algorithm that integrates spatiotemporal isolation and emergency response mechanisms, multiple inspection drones are scheduled for tasks. Step S5: Construct a distribution intelligent inspection application module based on the power grid management platform to realize inspection data sharing and system integration.

[0007] Preferably, in step S1, the lightweight network improved based on the attention mechanism is a YOLO+MobileNet lightweight network, which uses the MobileNet V3 network combined with the CSPdarknet53 feature layer in YOLO V4 for feature extraction.

[0008] Preferably, in step S1, the effective feature layers of the last three shapes of the MobileNet V3 model are extracted and fused with the effective feature layers of CSPdarknet53 to form the network structure of the lightweight network; the multi-source data augmentation method includes rotation, translation, scaling, and random occlusion operations, and the MOSAIC augmentation technology is combined synchronously during model training to optimize the training process, and the training loss value is calculated through the CIoU loss function to adjust the model parameters.

[0009] Preferably, in step S2, the distribution network AI edge computing device uses a dedicated power network card to achieve real-time video transmission and penetration between internal and external networks; the algorithm optimization and update mechanism uses a multimodal large model to reduce the initial training load of the equipment problem identification model and the construction external damage problem identification model, and iteratively optimizes the two models through continuous updates of the identification material library.

[0010] Preferably, in step S3, the automated drone nest adopts a modular design, including a docking module, a power replenishment module, and a communication module; the power replenishment module is used to support the charging, replacement, and intelligent wake-up functions of the inspection drone's battery; the communication module uses TCP / IP protocol combined with 5G network to establish a communication and data transmission scheme between the automated drone nest and the inspection drone to ensure data transmission stability.

[0011] Preferably, in step S4, the multi-machine multi-task collaborative scheduling algorithm realizes the scheduling of multiple inspection drones based on the road network, and integrates data fusion and analysis functions; the algorithm uses a genetic algorithm or particle swarm algorithm to realize the optimal allocation of inspection tasks, and includes navigation decision, satellite positioning anomaly handling, visual navigation, aircraft type adaptation, health monitoring, visual assisted shooting and fixed route remote take-off and landing sub-algorithms.

[0012] Preferably, in step S4, the spatiotemporal isolation mechanism is achieved through airspace grid segmentation and highly refined management, and sets safe isolation distances for multiple inspection drones in the time and space dimensions; the emergency handling mechanism automatically judges the fault or abnormal situation of the inspection drone based on preset rules, and quickly adjusts the inspection task allocation according to the judgment result.

[0013] Preferably, in step S5, the power distribution intelligent inspection application module achieves system integration with the defect management system, the external force damage prevention visa management system, the streaming media platform, edge services, and the power HarmonyOS ecosystem, and the system integration adopts a microservice architecture to complete the splitting, integration, and orchestration of services.

[0014] The present invention also provides a power distribution network intelligent inspection system, including multiple inspection drones, power distribution network AI edge computing equipment, automatic drone nests, multi-machine scheduling module and power distribution intelligent inspection application module; The power distribution network AI edge computing device integrates 5G network and AI edge computing capabilities to realize real-time video transmission for inspection, and is equipped with algorithm optimization and update mechanisms. The automated drone nest is used to provide docking, recharging and communication services for the inspection drone, and the automated drone nest integrates the power distribution network AI edge computing device; The multi-machine scheduling module has a built-in multi-machine multi-task collaborative scheduling algorithm that integrates a spatiotemporal isolation mechanism and an emergency handling mechanism, which is used to schedule inspection tasks for multiple inspection drones. The intelligent power distribution inspection application module is built on the power grid management platform and is used to realize inspection data sharing and system integration. The system constructs equipment problem identification models and construction external damage problem identification models based on lightweight networks improved by attention mechanism and knowledge distillation technology, and uses multi-source data augmentation methods and noise augmentation strategies to complete the training and optimization of the models.

[0015] The present invention also provides a computer program product, including computer instructions, which instruct computer equipment to perform operations corresponding to the intelligent inspection method for the power distribution network.

[0016] The beneficial effects of this invention are as follows: It effectively solves many technical pain points of traditional inspection methods and existing drone inspections. By constructing a distribution network-specific intelligent identification model through an improved lightweight network, combined with multi-source data enhancement and loss function optimization, it significantly improves the efficiency and accuracy of identifying equipment defects and construction-related external damage hazards in the distribution network, significantly reducing manual intervention. Relying on intelligent optimization algorithms to achieve unified scheduling of multiple machines and tasks, coupled with spatiotemporal isolation and emergency handling mechanisms, it avoids task conflicts and enables rapid response to emergencies, effectively improving the safety and flexibility of inspection operations. Simultaneously, the distribution intelligent inspection application module achieves seamless integration with power grid management platforms, the Dianhong ecosystem, and other systems, completely breaking down information silos. Furthermore, it ensures the integrity and security of data transmission and storage through multiple security solutions, including domestic cryptographic technology, access control, and data backup, supporting concurrent access by multiple users. This comprehensively improves the intelligence and automation level of distribution network operation and maintenance management, providing reliable technical support for the safe and stable operation of the distribution network. Attached Figure Description

[0017] 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 only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating an intelligent inspection method for a power distribution network according to an embodiment of the present invention.

[0019] Figure 2 This is a schematic diagram of the implementation architecture of an intelligent inspection method for power distribution networks according to Embodiment 1 of the present invention. Detailed Implementation

[0020] The following description of the embodiments is taken with reference to the accompanying drawings, which illustrate specific embodiments in which the invention can be implemented.

[0021] Please refer to Figure 1 As shown, Embodiment 1 of the present invention provides a method for intelligent inspection of a power distribution network, comprising: Step S1: Construct equipment problem identification model and construction external damage problem identification model based on lightweight network improved by attention mechanism and knowledge distillation technology, and optimize model training by using multi-source data augmentation method and noise augmentation strategy; Step S2: Deploy a power distribution network AI edge computing device that integrates 5G network and AI edge computing to realize real-time video transmission for inspection, and configure algorithm optimization and update mechanisms; Step S3: Construct an automated drone nest to provide docking, recharging and communication services for inspection drones, wherein the automated drone nest integrates the power distribution network AI edge computing device; Step S4: Based on the road network and using a multi-machine multi-task collaborative scheduling algorithm that integrates spatiotemporal isolation and emergency response mechanisms, multiple inspection drones are scheduled for tasks. Step S5: Construct a distribution intelligent inspection application module based on the power grid management platform to realize inspection data sharing and system integration.

[0022] The intelligent distribution network inspection method of this invention is adapted to an intelligent UAV inspection system composed of multiple inspection UAVs, forming an integrated intelligent distribution network inspection solution of "intelligent identification - edge computing - automatic UAV nesting - multi-UAV scheduling - platform integration". This solves the technical problems of low identification accuracy, poor real-time data processing, low automation level of UAVs, unreasonable multi-UAV scheduling, and information silos in existing distribution network inspections, achieving efficient, intelligent, and automated operation and maintenance of distribution network inspections. The following is a combination of... Figure 2 The architecture shown provides a detailed explanation of each step.

[0023] Step S1 is used to build and optimize the dual proprietary identification model for power distribution inspection.

[0024] Specifically, step S1 constructs a power distribution inspection equipment problem identification model and a construction damage problem identification model based on a lightweight network improved by attention mechanism and knowledge distillation technology. It also uses multi-source data augmentation methods and noise enhancement strategies to optimize the model training process. Through the dedicated construction and training optimization of the model, the accuracy and speed of identifying power distribution equipment defects and construction damage hazards are improved, solving the technical problems of lack of dedicated identification models and poor identification effect in existing inspections.

[0025] In this embodiment of the invention, the lightweight network based on the attention mechanism improvement is specifically the YOLO+MobileNet lightweight network. This lightweight network uses the MobileNet V3 network combined with the CSPdarknet53 feature layer in YOLO V4 for feature extraction. Specifically, it extracts the effective feature layers of the last three shapes of the MobileNet V3 model and merges them with the effective feature layers of CSPdarknet53 to form a new network structure. Combined with knowledge distillation technology, it achieves both lightweight and high accuracy of the model, which is suitable for the computing power requirements of inspection drones and edge computing devices, while ensuring the accuracy of recognition.

[0026] Multi-source data augmentation methods specifically include rotation, translation, scaling, and random occlusion operations. During model training, MOSAIC augmentation technology and noise enhancement strategies are combined simultaneously to form a multi-dimensional model training optimization strategy. Among them, MOSAIC augmentation technology stitches together multiple images by randomly scaling, cropping, and arranging them to form training images, and dynamically selects the type of training image according to the actual needs of small object recognition, effectively improving the model's ability to identify defects in small-sized equipment and minor external damage hazards in the power distribution network. The noise enhancement strategy improves the model's anti-interference ability and adapts to the image recognition needs in complex outdoor inspection environments.

[0027] In the loss calculation stage of model training, the CIoU loss function is used to calculate the training loss value. The model parameters are adjusted through real-time feedback of the loss value to further optimize the model training effect. This allows the constructed equipment problem identification model and construction external damage problem identification model to accurately match the actual scenario of power distribution network inspection, and to accurately identify equipment problems such as broken strands in power distribution lines, damaged insulators, corroded hardware, and foreign objects in conductors, as well as construction external damage problems such as unauthorized construction machinery and foreign objects intruding into the line protection zone.

[0028] Step S2 is used to deploy AI edge computing devices for the power distribution network and establish an algorithm update mechanism.

[0029] Specifically, step S2 involves deploying distribution network AI edge computing equipment that integrates 5G network and AI edge computing to achieve real-time video transmission of inspection data and establish an algorithm optimization and update mechanism. By replacing traditional cloud computing with edge computing, it solves the technical problems of high data transmission pressure, poor real-time performance, and low algorithm iteration efficiency in inspections.

[0030] The distribution network AI edge computing device is a dedicated edge computing hardware adapted for outdoor inspection scenarios of distribution networks. It deeply integrates the high-speed transmission characteristics of 5G networks with the local data processing capabilities of AI edge computing, pushing the analysis and processing of inspection data to the edge, effectively reducing the amount of data transmission in the cloud and improving the real-time performance of data processing. In the real-time video transmission stage, a dedicated power network card is used to transmit inspection images and video data and achieve penetration between internal and external networks. Relying on the specialization and stability of the dedicated power network, it avoids the problems of poor public network signal and transmission interruption in outdoor inspection scenarios, ensuring that the inspection data collected by the inspection drone can be transmitted to the edge computing device for processing in real time and stably.

[0031] The algorithm optimization and update mechanism is designed for the dual proprietary recognition model constructed in step S1. Specifically, it utilizes a multimodal large model to reduce the initial training load of the dual proprietary recognition model, thereby reducing the computational and time costs of model training. At the same time, it establishes a continuous update mechanism for the recognition material library, continuously adding images / videos of new equipment problems and construction damage hazards collected during the inspection by the inspection drone into the recognition material library. By updating the material library, the dual proprietary recognition model is iteratively optimized, enabling the model to continuously adapt to new defects and new hazard types that constantly emerge in power distribution network inspections, ensuring the long-term effectiveness of the recognition model.

[0032] Step S3 is used to build an automated nesting system for integrated edge computing devices.

[0033] Specifically, step S3 constructs an automated drone nest to provide docking, recharging, and communication services for inspection drones. The automated drone nest integrates the distribution network AI edge computing device from step S2, enabling unattended docking, recharging, and data interaction for inspection drones. This solves the technical problems of existing inspection drones relying on manual labor for inspection take-off and landing and recharging, having low automation levels, and insufficient equipment utilization.

[0034] The automated drone nest adopts a modular design concept, comprising three main modules: a docking module, a power supply module, and a communication module. These modules work together to achieve fully unattended operation and maintenance of the inspection drones: the docking module provides stable docking and take-off / landing positions for the inspection drones, adapting to the docking needs of various drone models; the power supply module supports charging, replacement, and intelligent wake-up functions for the inspection drone batteries, enabling automatic detection, charging, and replacement of batteries, and intelligent wake-up of the drones based on inspection task requirements, significantly improving the drones' continuous operation capability; the communication module establishes a stable communication link between the automated drone nest, the inspection drones, and the power grid management platform, ensuring real-time transmission of inspection data and dispatch commands.

[0035] An efficient communication and data transmission scheme is established between the automated drone nest and the inspection drone. The communication module uses TCP / IP protocol combined with 5G network to achieve data transmission. The combination of dual protocols / networks ensures the stability and real-time performance of data transmission between the nest and the inspection drone. At the same time, the automated drone nest integrates power distribution network AI edge computing equipment, enabling the nest to not only have docking and recharging functions, but also to realize local edge processing of inspection data. Inspection data collected by the inspection drone can be directly transmitted to the edge computing equipment in the nest for analysis and identification, further improving the efficiency of inspection data processing and realizing an edge-end closed loop of "inspection-processing-feedback".

[0036] Step S4 implements the task scheduling of inspection drones based on the multi-machine multi-task collaborative scheduling algorithm.

[0037] Specifically, step S4 uses a multi-machine, multi-task collaborative scheduling algorithm based on the road network and integrating spatiotemporal isolation and emergency response mechanisms to schedule inspection tasks for multiple inspection drones. This addresses the technical problems of existing inspection drones lacking multi-machine collaborative scheduling mechanisms, having unreasonable task allocation, and slow emergency response, thereby achieving optimal allocation of inspection resources and rapid response to emergencies.

[0038] The multi-machine, multi-task collaborative scheduling algorithm integrates data fusion and analysis functions. Based on the road network distribution in the area where the power distribution network is located, the scheduling logic is designed to adapt to actual inspection scenarios along the road network, making the inspection route planning and task allocation of inspection drones more aligned with the actual situation on site. The algorithm incorporates sub-algorithms such as navigation decision-making, satellite positioning anomaly handling, visual navigation, drone model adaptation, health monitoring, visual-assisted shooting, and fixed-route remote take-off and landing. These sub-algorithms work together to achieve intelligent scheduling of inspection drones throughout the entire process: the navigation decision-making and visual navigation sub-algorithms ensure the accuracy of the inspection drone's inspection route; the satellite positioning anomaly handling sub-algorithm achieves accurate positioning of the inspection drone when satellite positioning fails; the drone model adaptation sub-algorithm matches the appropriate inspection drone model based on the difficulty of the inspection task and the regional environment; the health monitoring sub-algorithm monitors the battery, fuselage, and communication status of the inspection drone in real time; the visual-assisted shooting sub-algorithm guides the inspection drone to accurately capture defects / potential hazards; and the fixed-route remote take-off and landing sub-algorithm enables cross-regional inspection scheduling of the inspection drone.

[0039] In the optimal allocation stage of inspection tasks, an intelligent optimization algorithm is used to search for the optimal scheduling solution. Specifically, the intelligent optimization algorithm is either a genetic algorithm or a particle swarm optimization algorithm. If a genetic algorithm is used, the search for the optimal solution of the inspection task is achieved through steps such as parameter encoding, initial population setting, fitness function design, genetic operations, and control parameter setting. If a particle swarm optimization algorithm is used, the search for the optimal solution of the inspection task is achieved by adjusting the inertia weight, individual learning factor, and social learning factor, and optimizing the particle movement trajectory based on the particle's historical best position and the population's historical best position.

[0040] The spatiotemporal isolation mechanism integrated into the multi-machine, multi-task collaborative scheduling algorithm is achieved through airspace grid segmentation and highly refined management. It sets safe isolation distances for multiple inspection drones performing inspection tasks in both time and space dimensions, avoiding airspace conflicts and collisions during multi-machine inspections and ensuring the safety of multi-machine collaborative inspections. The integrated emergency response mechanism automatically judges the faults or abnormal conditions of inspection drones (such as low battery, fuselage failure, communication interruption, etc.) based on preset rules, and quickly adjusts the inspection task allocation according to the judgment results. For example, it dispatches nearby inspection drones to take over the inspection tasks of the faulty inspection drone, realizing the automation and rapid response of inspection emergencies.

[0041] Step S5 is used to build the intelligent power distribution inspection application module and realize system integration.

[0042] Specifically, step S5 involves building a distribution intelligent inspection application module based on the power grid management platform. This module enables the sharing of power grid data and the integration with relevant power grid systems, solving the information silo problem between existing inspection data and the power grid management system, and realizing the full-process sharing and collaborative application of inspection data.

[0043] The distribution intelligent inspection application module is built on the existing power grid management platform's hardware and software foundation. Its core function is to realize Dianhong data sharing. It shares data such as equipment status, defect / hazard information, and inspection trajectory collected by inspection drones during inspections, as well as various data generated by edge computing devices, automatic drone nests, and multi-machine scheduling algorithms, after unifying the format according to Dianhong data standards. This allows various business systems of the power grid to easily access inspection data.

[0044] To ensure information security during data sharing, the distribution intelligent inspection application module employs encryption protection measures for secure data storage and transmission. Specifically, it utilizes domestically developed cryptographic technology for encrypted data storage and transmission, and combines this with identity authentication, access control, data leakage prevention gateways, local backup, and disaster recovery technologies to form a multi-layered information security system, preventing the leakage, loss, or tampering of inspection data. Furthermore, the application module establishes a reasonable user interaction mechanism, adapting to the operating habits of power grid maintenance personnel, enabling the visualization, querying, statistics, and reporting of inspection data, thereby improving the operational efficiency of maintenance personnel.

[0045] In the system integration phase, the distribution intelligent inspection application module completes comprehensive integration with power grid-related business systems and ecosystems. Specifically, this includes the defect management system, the external damage prevention visa management system, the streaming media platform, edge services, and the Dianhong ecosystem. The system integration is implemented using a microservice architecture, which completes the service decomposition, integration, and orchestration operations between various systems / platforms. This allows inspection data to flow seamlessly between systems. For example, identified equipment defect information is automatically synchronized to the defect management system, and construction-related external damage hazard information is automatically synchronized to the external damage prevention visa management system. This achieves deep integration of inspection data with power grid operation and maintenance business, fully exploring the application value of inspection data.

[0046] As can be seen from the above, the intelligent distribution network inspection method of the present invention realizes the intelligence, automation, and efficiency of the entire distribution network inspection process. All technical indicators meet the requirements of refined operation and maintenance of distribution networks: the identification accuracy of distribution inspection equipment defects and construction-related external damage hazards is ≥90%, and the output time of a single image recognition result is ≤1 second, which greatly improves the identification efficiency and accuracy of defects / hazards; the intelligent wake-up time of the automatic hive for inspection drones is ≤60 seconds, realizing the rapid response operation of inspection drones; the intelligent scheduling module can schedule a maximum of ≥40 inspection drones and hangar equipment per square kilometer in a single area, and the output time of the scheduling result is ≤30 seconds, realizing the optimal allocation and rapid scheduling of inspection resources.

[0047] Corresponding to the intelligent inspection method for power distribution networks in Embodiment 1 of the present invention, Embodiment 2 of the present invention also provides an intelligent inspection system for power distribution networks, including multiple inspection drones, power distribution network AI edge computing devices, automatic drone nests, multi-machine scheduling modules, and intelligent inspection application modules for power distribution networks; The power distribution network AI edge computing device integrates 5G network and AI edge computing capabilities to realize real-time video transmission for inspection, and is equipped with algorithm optimization and update mechanisms. The automated drone nest is used to provide docking, recharging and communication services for the inspection drone, and the automated drone nest integrates the power distribution network AI edge computing device; The multi-machine scheduling module has a built-in multi-machine multi-task collaborative scheduling algorithm that integrates a spatiotemporal isolation mechanism and an emergency handling mechanism, which is used to schedule inspection tasks for multiple inspection drones. The intelligent power distribution inspection application module is built on the power grid management platform and is used to realize inspection data sharing and system integration. The system constructs equipment problem identification models and construction external damage problem identification models based on lightweight networks improved by attention mechanism and knowledge distillation technology, and uses multi-source data augmentation methods and noise augmentation strategies to complete the training and optimization of the models.

[0048] Corresponding to the intelligent inspection method for power distribution networks in Embodiment 1 of the present invention, Embodiment 3 of the present invention also provides a computer program product, including computer instructions, which instruct computer equipment to perform the operation corresponding to the intelligent inspection method for power distribution networks.

[0049] Preferably, the computer device is a hardware device with data processing, computing and communication capabilities, including at least a processor and a memory, and may also be configured with hardware components such as communication modules and input / output modules according to actual needs. The components establish data interaction links through system bus or other communication connection methods to collaboratively execute the operations corresponding to the intelligent inspection method of the power distribution network.

[0050] The processor can be a central processing unit (CPU), or other general-purpose processors, 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, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or the processor can be any conventional processor. The processor acts as the control center of the computer equipment, connecting various components of the computer equipment through various interfaces and lines, and is used to call and execute computer instructions stored in the memory to realize the corresponding operations of each step of the intelligent inspection method of the power distribution network.

[0051] The memory mainly includes a program storage area and a data storage area. The program storage area can store the operating system, computer instructions required to implement the intelligent inspection method of the distribution network, and application programs required for at least one function to run. The data storage area can store inspection data, model training data, scheduling data, system integration interaction data, and other related data generated during the execution of the intelligent inspection method of the distribution network. In addition, the memory can be a high-speed random access memory, or a non-volatile memory, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, or other volatile solid-state storage devices, to provide stable data storage and retrieval support for the processor's operation and processing.

[0052] The computer instructions are stored in the memory in the form of computer programs. They are machine instructions that can be recognized, parsed and executed by the processor. The instruction logic corresponds one-to-one with each step of the distribution network intelligent inspection method. They can be transmitted and loaded between computer devices along with the computer program. The processor reads and executes the computer instructions to drive the computer devices to complete all operations corresponding to the distribution network intelligent inspection method, such as model building and optimization, edge computing deployment, multi-machine scheduling and platform integration.

[0053] It should be noted that the above-mentioned devices may include, but are not limited to, processors and memory, as will be understood by those skilled in the art.

[0054] For the working principle and process of the above embodiments, please refer to the description of Embodiment 1 of the present invention, which will not be repeated here.

[0055] Compared with existing technologies, this invention has the following significant advantages: It effectively solves many technical pain points of traditional inspection methods and existing drone-based inspections. By constructing a distribution network-specific intelligent identification model through an improved lightweight network, combined with multi-source data enhancement and loss function optimization, it significantly improves the efficiency and accuracy of identifying equipment defects and construction-related external damage hazards in the distribution network, significantly reducing manual intervention. Relying on intelligent optimization algorithms to achieve unified scheduling of multiple machines and tasks, coupled with spatiotemporal isolation and emergency handling mechanisms, it avoids task conflicts and enables rapid response to emergencies, effectively improving the safety and flexibility of inspection operations. Simultaneously, the distribution intelligent inspection application module achieves seamless integration with power grid management platforms, the Dianhong ecosystem, and other systems, completely breaking down information silos. Furthermore, it ensures the integrity and security of data transmission and storage through multiple security solutions, including domestic cryptographic technology, access control, and data backup, supporting concurrent access by multiple users. This comprehensively improves the intelligence and automation level of distribution network operation and maintenance management, providing reliable technical support for the safe and stable operation of the distribution network.

[0056] The above description is merely a preferred embodiment of the present invention and should not be construed as limiting the scope of the invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for intelligent inspection of power distribution networks, characterized in that, include: Step S1: Construct equipment problem identification model and construction external damage problem identification model based on lightweight network improved by attention mechanism and knowledge distillation technology, and optimize model training by using multi-source data augmentation method and noise augmentation strategy; Step S2: Deploy a power distribution network AI edge computing device that integrates 5G network and AI edge computing to realize real-time video transmission for inspection, and configure algorithm optimization and update mechanisms; Step S3: Construct an automated drone nest to provide docking, recharging and communication services for inspection drones, wherein the automated drone nest integrates the power distribution network AI edge computing device; Step S4: Based on the road network and using a multi-machine multi-task collaborative scheduling algorithm that integrates spatiotemporal isolation and emergency response mechanisms, multiple inspection drones are scheduled for tasks. Step S5: Construct a distribution intelligent inspection application module based on the power grid management platform to realize inspection data sharing and system integration.

2. The method according to claim 1, characterized in that, In step S1, the lightweight network improved based on the attention mechanism is the YOLO+MobileNet lightweight network, which uses the MobileNet V3 network combined with the CSPdarknet53 feature layer in YOLO V4 for feature extraction.

3. The method according to claim 2, characterized in that, In step S1, the effective feature layers of the last three shapes of the MobileNet V3 model are extracted and fused with the effective feature layers of CSPdarknet53 to form the network structure of the lightweight network. The multi-source data augmentation method includes rotation, translation, scaling, and random occlusion operations. During model training, the MOSAIC augmentation technique is combined to optimize the training process, and the training loss value is calculated through the CIoU loss function to adjust the model parameters.

4. The method according to claim 1, characterized in that, In step S2, the distribution network AI edge computing device uses a dedicated power network card to achieve real-time video transmission and penetration between internal and external networks; the algorithm optimization and update mechanism uses a multimodal large model to reduce the initial training load of the equipment problem identification model and the construction external damage problem identification model, and iteratively optimizes the two models through continuous updates of the identification material library.

5. The method according to claim 1, characterized in that, In step S3, the automated drone nest adopts a modular design, including a docking module, a power replenishment module, and a communication module. The power replenishment module is used to support the charging, replacement, and intelligent wake-up functions of the inspection drone's battery. The communication module uses the TCP / IP protocol combined with a 5G network to establish a communication and data transmission scheme between the automated drone nest and the inspection drone to ensure data transmission stability.

6. The method according to claim 1, characterized in that, In step S4, the multi-machine multi-task collaborative scheduling algorithm realizes the scheduling of multiple inspection drones based on the road network, and integrates data fusion and analysis functions; the algorithm uses genetic algorithm or particle swarm algorithm to realize the optimal allocation of inspection tasks, and includes navigation decision, satellite positioning anomaly handling, visual navigation, model adaptation, health monitoring, visual assisted shooting and fixed route remote take-off and landing sub-algorithms.

7. The method according to claim 6, characterized in that, In step S4, the spatiotemporal isolation mechanism is achieved through airspace grid segmentation and highly refined management, and sets safe isolation distances for multiple inspection drones in the time and space dimensions; the emergency handling mechanism automatically judges the fault or abnormal situation of the inspection drones based on preset rules, and quickly adjusts the inspection task allocation according to the judgment results.

8. The method according to claim 1, characterized in that, In step S5, the power distribution intelligent inspection application module achieves system integration with the defect management system, the external force damage prevention visa management system, the streaming media platform, edge services, and the power HarmonyOS ecosystem. The system integration adopts a microservice architecture to complete the splitting, integration, and orchestration of services.

9. A power distribution network intelligent inspection system, characterized in that, This includes multiple inspection drones, power distribution network AI edge computing equipment, automatic drone nests, multi-drone scheduling modules, and power distribution intelligent inspection application modules; The power distribution network AI edge computing device integrates 5G network and AI edge computing capabilities to realize real-time video transmission for inspection, and is equipped with algorithm optimization and update mechanisms. The automated drone nest is used to provide docking, recharging and communication services for the inspection drone, and the automated drone nest integrates the power distribution network AI edge computing device; The multi-machine scheduling module has a built-in multi-machine multi-task collaborative scheduling algorithm that integrates a spatiotemporal isolation mechanism and an emergency handling mechanism, which is used to schedule inspection tasks for multiple inspection drones. The intelligent power distribution inspection application module is built on the power grid management platform and is used to realize inspection data sharing and system integration. The system constructs equipment problem identification models and construction external damage problem identification models based on lightweight networks improved by attention mechanism and knowledge distillation technology, and uses multi-source data augmentation methods and noise augmentation strategies to complete the training and optimization of the models.

10. A computer program product, characterized in that, It includes computer instructions that instruct computer equipment to perform operations corresponding to the intelligent inspection method for power distribution networks as described in any one of claims 1 to 8.

Images