Intelligent inspection auxiliary system for substations in the context of the Internet of Things for power

Abstract

This invention discloses an intelligent substation inspection auxiliary system in the context of the power Internet of Things (IoT), relating to the field of intelligent inspection control technology. The development unit acquires the infrastructure of the target substation and develops intelligent inspection modules by defining task scenarios—inspection topologies. One type of inspection control unit performs one type of inspection task judgment, switches the target inspection topology, and executes topology-driven facility inspection control under a forward protocol. A second type of inspection control unit transmits self-inspection risk data to an edge computing gateway, triggering lightweight inspection analysis embedded in the gateway and executing topology-driven facility inspection control under a reverse protocol. This addresses the technical problems in existing technologies, such as the difficulty in achieving proactive perception, rapid response, and risk closed-loop management, and the inability to ensure both inspection efficiency and accuracy while maintaining adaptive and flexible deployment. The goal is to achieve proactive perception, rapid response, and risk closed-loop management, while balancing inspection efficiency and accuracy under adaptive and flexible deployment.

Description

Technical Field

[0001] This invention relates to the field of intelligent inspection and control technology, specifically to an intelligent inspection auxiliary system for substations in the context of the Internet of Things in the power industry. Background Technology

[0002] In existing power system inspection practices, substation inspections primarily rely on manual fixed-point inspections or semi-automated robotic inspections. While manual inspections offer some flexibility and judgment, they are limited by human resources, the working environment, and the complexity of the tasks. To improve efficiency, some systems have introduced image recognition, edge computing, or IoT sensing devices, but the following technical bottlenecks remain: First, existing systems mostly employ static task modes, lacking the ability to dynamically adapt to the inspection tasks and equipment topology, making it difficult to cope with complex scenarios and unexpected events; second, inspection paths and priorities cannot be adaptively adjusted; and third, current communication protocols are mostly based on traditional passive transmission architectures, making it difficult to support efficient control tasks.

[0003] In summary, there is currently a lack of a fully functional intelligent inspection system for substations, making it difficult to achieve proactive perception, rapid response, and closed-loop risk management, and unable to ensure both inspection efficiency and accuracy while maintaining adaptive and flexible deployment. Summary of the Invention

[0004] This application provides a substation intelligent inspection auxiliary system in the context of the power Internet of Things, which is used to address the technical problems in existing technologies that make it difficult to achieve proactive perception, rapid response and risk closed-loop management, and to ensure that inspection efficiency and accuracy can be balanced on the basis of adaptive and flexible deployment.

[0005] In view of the above problems, this application provides a substation intelligent inspection auxiliary system in the power Internet of Things environment. The system includes: a development unit, used to acquire the infrastructure of the target substation, and develop an intelligent inspection module within the substation intelligent inspection auxiliary system by defining a task scenario-inspection topology; a first-class inspection control unit, used to determine if an inspection task is uploaded to the intelligent inspection module as a first-class inspection task, switch to the target inspection topology through task interpretation and scenario-based topology matching, and execute topology-driven facility inspection control under the forward protocol using a synaptic IoT protocol; and a second-class inspection control unit, used to transmit risk data found in the self-inspection of the inspected infrastructure to the edge computing gateway, reverse-engineer the risk event and generate a second-class inspection task, trigger the lightweight inspection module embedded in the gateway, determine the event inspection topology and execute topology-driven facility inspection control under the reverse protocol; wherein, the inspection topology is optimized and adjusted through self-healing networking.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] The substation intelligent inspection auxiliary system provided in this application embodiment under the power Internet of Things environment includes a development unit for acquiring the infrastructure of the target substation, defining a task scenario-inspection topology, and developing an intelligent inspection module within the substation intelligent inspection auxiliary system. A first-class inspection control unit is used to determine if an inspection task is uploaded to the intelligent inspection module as a first-class inspection task. Through task interpretation and scenario-based topology matching, it switches to the target inspection topology, assists with a synaptic IoT protocol, and executes topology-driven facility inspection control under the forward protocol. A second-class inspection control unit is used to transmit risk data detected by the self-inspection of the inspected infrastructure to the edge computing gateway, reverse-engineer the risk event, generate a second-class inspection task, trigger a lightweight inspection module embedded in the gateway, determine the event inspection topology, and execute topology-driven facility inspection control under the reverse protocol. This addresses the technical problems in existing technologies where it is difficult to achieve proactive perception, rapid response, and risk closed-loop management, and where it is impossible to ensure both inspection efficiency and accuracy while maintaining adaptive and flexible deployment. This system aims to achieve proactive perception, rapid response, and risk closed-loop management, while maintaining both inspection efficiency and accuracy while maintaining adaptive and flexible deployment. Attached Figure Description

[0008] Figure 1 This application provides a logical flow diagram of a substation intelligent inspection auxiliary system in the context of the power Internet of Things.

[0009] Figure 2 This application provides a schematic diagram of the structure of a substation intelligent inspection auxiliary system in the context of the Internet of Things for power.

[0010] Explanation of reference numerals in the attached drawings: Development unit 11, Class I inspection control unit 12, Class II inspection control unit 13. Detailed Implementation

[0011] This application provides a substation intelligent inspection auxiliary system in the context of the power Internet of Things, which addresses the technical problems in existing technologies that make it difficult to achieve proactive perception, rapid response, and risk closed-loop management, and that cannot ensure both inspection efficiency and accuracy while maintaining adaptive and flexible deployment.

[0012] Example: Figure 1 , Figure 2 As shown, this application provides a substation intelligent inspection auxiliary system in the context of the power Internet of Things, the system comprising:

[0013] Development Unit 11 is used to acquire the infrastructure of the target substation and develop an intelligent inspection module within the substation intelligent inspection auxiliary system by defining the task scenario - inspection topology.

[0014] In this embodiment, the infrastructure of the target substation is acquired. This infrastructure includes various types of power transmission, control, measurement, and protection equipment, access terminals, and their physical connection structures. In one feasible embodiment, the development unit 11 establishes a data connection with the substation asset management system or real-time monitoring platform, and retrieves structured equipment files and operating status information to obtain a complete list of substation facilities and their layout topology. Preferably, the acquisition process supports data synchronization and dynamic updating functions to ensure that the acquired equipment information remains consistent with the actual situation on site.

[0015] Furthermore, after acquiring the infrastructure data, development unit 11 performs the definition operation of the task scenario - inspection topology. Specifically, the task scenario is a set of task execution conditions constructed around a specific inspection target, such as equipment aging detection, high temperature anomaly early warning, and communication status verification; the inspection topology refers to the logical structure and control path between inspection objects in the corresponding scenario, reflecting the dependencies and action sequence between devices during the inspection process.

[0016] Optionally, the mapping relationship between the task scenario and the inspection topology is established through a rule engine. In a specific implementation process provided in this application, the development unit 11 divides the inspection units according to the substation infrastructure type, such as GIS bays, busbar connections, main transformer sections, etc., extracts the associated equipment set of each unit, constructs the corresponding logical nodes, and generates an executable inspection topology structure.

[0017] Furthermore, to enhance the intelligence and practicality of the task scenario-inspection topology construction, development unit 11 further introduces multi-source data such as inspection history records, equipment health records and alarm data for fusion analysis.

[0018] For example, when historical fluctuations in main transformer oil temperature and three-phase current imbalance are detected simultaneously, a task scenario for early warning of main transformer anomalies can be generated, and a set of inspection topologies with the main transformer body, bushings, cooling fans, and monitoring units as core nodes can be matched. This structure reflects which inspection facilities should be driven in this specific scenario, and which facilities should be prioritized for tasks such as motion recognition, temperature sampling, or status verification.

[0019] Finally, based on the mapping between the task scenario and the inspection topology, an intelligent inspection module is constructed. In this application, this module serves as the core unit of the system execution layer, including functions such as inspection task scheduling, path control, data interaction, and risk feedback, and supports dynamic topology switching and task evolution configuration during subsequent inspection processes.

[0020] The above methods enable a comprehensive deployment process from static facility data to dynamic task execution logic, providing fundamental support for the operation of the substation intelligent inspection auxiliary system.

[0021] Furthermore, the development unit 11 includes:

[0022] The facility classification unit is used to identify inspection infrastructure and substation infrastructure for the infrastructure, wherein the inspection infrastructure is identified by an inspection domain based on the substation infrastructure, and the inspection infrastructure includes substation access facilities and dynamic inspection facilities; the scenario reconstruction unit is used to perform topology reconstruction based on the substation scenario for the inspection infrastructure and define the task scenario-inspection topology.

[0023] In this embodiment of the application, the facility classification unit is used to perform structured classification operations on the infrastructure in order to clearly define the functional roles of various types of equipment in the substation.

[0024] Specifically, the infrastructure includes power equipment (e.g., circuit breakers, disconnect switches, voltage transformers, transformers, etc.), control and measurement devices, communication nodes, and their installation environment information. The facility classification unit first divides the infrastructure into two main categories based on the functional attributes, deployment location, and communication interface characteristics of the equipment: substation infrastructure and inspection infrastructure.

[0025] Among them, substation infrastructure refers to the core equipment that undertakes the transmission and control of primary and secondary power systems, and its functions do not have mobility or active sensing capabilities, such as busbars, main transformers, and switchgear; while inspection infrastructure refers to auxiliary facilities with status acquisition, information transmission or inspection execution capabilities, including but not limited to access terminals, sensing nodes, and mobile inspection equipment (such as inspection robots and drones).

[0026] Furthermore, after the classification is completed, the facility classification unit further identifies the corresponding inspection area for the inspected infrastructure, that is, the monitoring range of the substation infrastructure covered by the facility.

[0027] In one optional embodiment, the identification of the inspection domain is dynamically set based on the facility coverage radius, sensor parameter matching relationship, and physical deployment path. For example, a wireless inspection facility accessed in a GIS interval area has an inspection domain that includes equipment such as circuit breakers, current transformers, and disconnectors within that interval.

[0028] Specifically, inspection infrastructure includes two types: substation access facilities, such as fixed cameras, infrared thermal imaging modules, wiring terminals, etc., which are mainly used for local status monitoring and signal aggregation; and dynamic inspection facilities, such as track robots, wall-climbing detection devices, flying inspection units, etc., which have the ability to move and perform tasks between multiple inspection domains and are suitable for large-scale or high-frequency task scenarios.

[0029] Subsequently, the scenario reconstruction unit performs a topology reconstruction operation on the inspected infrastructure based on the substation scenario. The substation scenario refers to a set of task environments formed under specific operating conditions or early warning backgrounds, such as: equipment insulation status inspection after thunderstorms, bus system re-inspection after a short-term trip, etc.

[0030] Optionally, the scenario reconstruction unit, based on scenario-specific characteristics, mines the structure and logic of the availability of inspection infrastructure under different inspection scenarios, realizes the topology cascade reconstruction of scenario-specific inspection coverage relationships, and forms a logical inspection structure under the task scenario. The topology reconstruction process reflects the correlation strength between devices, the priority of control paths, and the order of execution processes, ensuring that the generated inspection topology reflects both the physical structure and the task logic.

[0031] Finally, the scene reconstruction unit converts the topology reconstruction results into a task scene-inspection topology mapping, which serves as the core basis for subsequent inspection control and module generation. Through the above facility classification and scene topology reconstruction process, a logical transformation from static structure to dynamic task-driven operation is achieved, improving the scene fit and configuration flexibility of inspection-driven operation.

[0032] Furthermore, the scene reconstruction unit includes:

[0033] The clustering unit is used to retrieve the inspection records of the target substation, perform scenario-based clustering, and determine M record groups; the mining unit is used to mine the scenario inspection facilities-inspection guides within the M record groups, with a preset ratio within each group as a constraint; the determination unit is used to determine the inspection topology based on the scenario inspection facilities-inspection guides; and the association unit is used to define task scenarios for the M record groups and perform the association between the task scenarios and the inspection topology.

[0034] In this embodiment, the clustering unit is used to retrieve inspection record data generated by the target substation in different operating cycles and perform scenario-oriented classification. The inspection records include, but are not limited to, multi-dimensional data fields such as inspection task number, execution time, execution path, facility status data, alarm information, and task response results.

[0035] Preferably, to improve the accuracy of scenario-based clustering, the clustering unit prioritizes extracting record entries with complete structure and good data timeliness, and performs unsupervised or semi-supervised clustering operations based on the task feature vector.

[0036] In one feasible implementation of the scenario-based clustering, a feature space is constructed based on the similarity of inspection tasks in dimensions such as facility path, task objective, execution frequency, and associated alarm type. Algorithms such as K-means, hierarchical clustering, or DBSCAN can be selected, but are not limited to, to divide the task into M record groups. Each record group represents a scenario task evolution trajectory with common inspection logic, where M is a positive integer.

[0037] Subsequently, the mining unit is used to further mine the structure of scene inspection facilities and inspection guidance in the M record groups, that is, to analyze the execution path of inspection tasks and the facility response relationship within each group, and to construct a path guidance map from the task trigger point to the response target facility. In specific implementation, this process is based on parameters such as the frequency of inspection tasks, facility response intensity, and path stability within the group. A preset ratio is used as a screening constraint in the data within the group, for example, the proportion of high-frequency paths is not less than 80%, to identify highly correlated facility sequences and extract them as inspection guidance relationships.

[0038] For example, in the scenario of detecting abnormal hot spots on a busbar, if multiple tasks are concentrated in the busbar area and highly repetitively involve temperature-sensing cameras, thermal imaging devices, GIS bay switches, and bus tie controllers, then this facility chain constitutes the scenario guidance path for this group.

[0039] Furthermore, the determining unit receives the scene inspection facility-inspection guide as input, and constructs an inspection topology graph with node-edge weight attributes based on the physical location relationships between facilities and the logical order of task execution. Optionally, each node in the topology represents an identifiable inspection infrastructure, and edges represent execution flows or data dependencies that occur in the task. The weight of the edges can be generated by indicators such as path frequency and importance score. This topology serves as the smallest executable inspection structural unit in the scene, used for subsequent task-driven and resource scheduling.

[0040] Furthermore, based on the M record groups, the association unit defines a corresponding task scenario label for each group and maps and binds it to the determined inspection topology. The task scenario label is generated by comprehensively considering the mined alarm type, task objective, and historical response results.

[0041] Examples include high-voltage switchgear re-inspection scenarios and cable temperature rise monitoring scenarios. In one optional embodiment provided in this application, the association between the task scenario and the inspection topology is stored through a structured index, supporting retrieval by task characteristics and rapid generation of matching execution structures, serving as one of the key bases for subsequent inspection control unit execution scheduling, path generation, and module adaptation.

[0042] In summary, by constructing a highly compatible and adaptable task topology, we can provide data-driven intelligent support capabilities for the intelligent inspection system.

[0043] Furthermore, the development unit includes:

[0044] The first deployment unit is used to interpret the inspection task and deploy the first decision node, wherein the interpretation dimensions include at least the molecular scale, component scale, and system scale; the second deployment unit is used to deploy the second decision node based on task scenario matching and inspection topology switching; the third deployment unit is used to deploy the third decision node based on the control decision of topology inspection parameters; the module construction unit is used to construct an inspection decision chain based on the first decision node, the second decision node, and the third decision node, and perform data-driven training until convergence, as the intelligent inspection module.

[0045] In this embodiment, the first deployment unit is used to perform multi-scale interpretation operations on the input inspection task, and deploy the first decision node accordingly. The inspection task interpretation proposed in this application refers to converting the task objective described in natural language or the structured task instructions into a set of decision parameters that the system can recognize, so as to achieve a refined decomposition of the task logic.

[0046] In one optional embodiment provided in this application, the interpretation process is based on a multi-dimensional framework, with interpretation dimensions including at least three levels: molecular scale, component scale, and system scale. The molecular scale represents the smallest equipment unit involved in the task, such as the detection requirements of bolts, contacts, and interface points; the component scale represents the operational status analysis at the specific component level, such as switching mechanisms, cooling fans, and connecting busbars; and the system scale corresponds to the judgment of the coordinated behavior of the entire equipment system, such as the coordinated state of the bus system and the connectivity of the power supply link.

[0047] The first deployment unit maps the aforementioned multi-scale task elements into structured vector form, thereby generating the first decision node deployed in the control model. The node's function is task interpretation and scale mapping.

[0048] Furthermore, the second deployment unit is used to deploy a second decision node after the task interpretation is completed, based on the mapping relationship between the task scenario matched by the task and the corresponding inspection topology.

[0049] Task scenario matching refers to identifying the target scenario most similar to the current task from the task scenario library (the set of clustered task scenarios mined in the previous steps) and extracting the topology bound to that scenario. Inspection topology switching refers to dynamically activating the appropriate inspection path and execution node based on different scenario switching logic. The second deployment unit deploys a second decision node based on this logic. The node's function is to complete the decision binding from task scenario to inspection path and support real-time adjustment of the topology during task flow.

[0050] Furthermore, the third deployment unit is used to deploy a third decision node based on key control parameters in the inspection topology. These control parameters include node evaluation weights, execution order priorities, equipment status criterion thresholds, and control action scheduling rules in the inspection path. Implementably, this deployment process, combined with the operational capabilities of the inspection facilities and communication feedback latency in the actual scenario, generates a set of strategies that can be used for real-time control. Based on this, the third decision node executes control behaviors such as path optimization, equipment scheduling, and status feedback triggering, thereby achieving comprehensive decision support for the execution strategies of each node within the topology.

[0051] In summary, the deployment of the node functional logic of each decision node has been completed. Furthermore, in order to ensure its automated driving capability and decision convergence, logical association and training are performed on it.

[0052] Specifically, the module construction unit logically connects the first, second, and third decision nodes to construct a complete inspection decision chain. This decision chain is structurally multi-layered, with different nodes corresponding to different task understanding, path matching, and control logic. The module construction unit further employs a data-driven training mechanism to perform end-to-end optimization training on the decision chain. The training data comes from historical inspection task execution data, equipment response logs, and inspection feedback results, and is iteratively updated based on supervised learning or reinforcement learning methods. After final training convergence, the resulting decision chain constitutes an intelligent inspection module with adaptive and generalizable capabilities, supporting efficient response and precise control of various tasks.

[0053] Furthermore, the intelligent inspection module establishes a communication loop with the edge computing gateway and the inspection infrastructure; wherein, inspection scheduling and management are carried out by deploying a synaptic-like IoT protocol, wherein the synaptic-like IoT protocol is an event-driven IoT interaction protocol, including forward and reverse protocols.

[0054] In this embodiment of the application, the intelligent inspection module establishes a stable, low-latency communication loop with the edge computing gateway and the inspection infrastructure. Optionally, the communication loop supports bidirectional status synchronization and command distribution functions.

[0055] Specifically, the intelligent inspection module, as the core decision-making body of the system, acquires uploaded task information in real time and forwards the generated control commands to the specific execution terminals through the edge computing gateway, realizing closed-loop control of the system. The communication loop supports multi-protocol access and task priority scheduling mechanism, and can automatically select the optimal transmission path according to the time sensitivity, data volume and security level of the task.

[0056] Preferably, in the above communication structure, to improve the adaptability and response efficiency of inspection tasks, a synaptic-like IoT protocol is introduced. This protocol simulates the signal transmission mechanism of biological neural synapses and uses an event-driven interaction method for task scheduling and data response. This type of protocol features dynamic routing, state wake-up, and collaborative feedback, and can automatically initiate relevant task chains when the inspection facility's state changes abruptly or when external events are triggered, avoiding the resource waste and response delays caused by traditional periodic polling.

[0057] Furthermore, the aforementioned synaptic-like IoT protocol is specifically divided into a forward protocol and a reverse protocol. The forward protocol is suitable for proactive inspection tasks, where the intelligent inspection module initiates the task, issues instructions to the inspection infrastructure, and performs status collection and target confirmation according to a preset path. The forward protocol emphasizes the forward transmission mechanism of task-driven and path control, and is commonly used in scenarios such as routine inspections and periodic maintenance.

[0058] Similarly, the reverse protocol is suitable for passive response tasks. When an inspection infrastructure detects an anomaly, such as abnormal temperature, arc discharge signal, or loose cable, it reports it back through the edge computing gateway, triggering the intelligent inspection module to generate a corresponding emergency task and quickly deploy the execution path. The reverse protocol emphasizes the event perception and retrospective deployment response mechanism, and is suitable for emergency scenarios such as fault tracing and anomaly verification.

[0059] In summary, a closed-loop linkage from state triggering to task response has been achieved, effectively improving the dynamic adaptability and intelligence level of substation inspection, and providing a technical foundation for realizing highly reliable inspection in the power Internet of Things environment.

[0060] Furthermore, the inspection tasks include a first type of inspection task and a second type of inspection task; wherein, the first type of inspection task is an active inspection task of the intelligent inspection module-inspection infrastructure, and the second type of inspection task is a passive inspection task of the inspection infrastructure-edge computing gateway-inspection infrastructure; wherein, the passive inspection task is guided by self-inspection risk triggering events.

[0061] Among them, the first type of inspection task adopts a forward protocol based on a synaptic IoT protocol, and the second type of inspection task adopts a reverse protocol based on a synaptic IoT protocol.

[0062] In this embodiment, the inspection tasks can be divided into two categories based on their triggering method and task path. The first category of inspection tasks is initiated by the intelligent inspection module, and its execution path is typically intelligent inspection module → inspection infrastructure.

[0063] Specifically, based on preset inspection strategies or periodic scheduling rules, combined with the task scenario and inspection topology, task instructions are automatically generated and sent to the corresponding inspection infrastructure, such as fixed sensors, track-based inspection robots, or infrared imaging devices. This task type is mainly used for routine inspections, periodic checks, or strategy-coverage tasks, and features clear task planning, controllable paths, and pre-execution.

[0064] In contrast, the second type of inspection task is a passive inspection task initiated by the inspection infrastructure itself in detecting abnormal states. Its complete path is inspection infrastructure → edge computing gateway → inspection infrastructure.

[0065] Specifically, some sensing nodes deployed in the system, such as temperature sensors, partial discharge detectors, and image recognition terminals, have real-time self-testing capabilities. When they detect that a device's status exceeds a preset safety threshold, such as excessive cable temperature rise, loose connections, or enhanced partial discharge signals, they immediately upload the abnormal data to the edge computing gateway. Upon receiving this data, the edge computing gateway uses its embedded event recognition model to perform anomaly analysis and task generation, triggering the corresponding scene topology and inspection path, and issuing a new round of targeted verification tasks to the executing devices to complete risk confirmation and emergency response operations.

[0066] In one feasible embodiment, the event recognition model is a simplified derivation function component for sample training. For example, the state vector is used as the input sample and the substation risk event is used as the output sample. The retrieved historical data records are integrated, and supervised training is performed until convergence, that is, the preset accuracy is met, and the completed event recognition model is determined.

[0067] Therefore, the second type of inspection task is guided by self-inspection risk triggering events, and has the characteristics of rapid response, flexible path and dynamic task generation. It is widely used in scenarios such as temporary inspection, fault tracing and anomaly re-inspection.

[0068] In the embodiments of this application, regarding the protocol mechanism, the aforementioned inspection task adopts a forward protocol based on a synaptic IoT protocol. The forward protocol constructs a forward command link from the intelligent inspection module to the inspection execution end, emphasizing task scheduling priority, inspection path order, and data flow control between facilities. This protocol features mechanisms such as task broadcasting, path wake-up, and sequence verification, making it suitable for a clearly structured proactive inspection process.

[0069] Correspondingly, the second type of inspection task adopts a reverse protocol based on a synaptic IoT protocol. This protocol simulates the reverse feedback mechanism of neurons, supporting event-driven data backpropagation, path selection, and task reallocation. Through the reverse protocol, it is possible to quickly reverse-schedule resources, construct emergency paths, and execute closed-loop handling based on abnormal events, thereby significantly improving the response capability to sudden failures or potential risks.

[0070] In summary, by distinguishing between active and passive path types and matching forward and reverse communication protocols, the inspection task achieves an efficient, dynamic, and adaptive task linkage mechanism between the intelligent inspection module and the edge computing architecture.

[0071] A type of inspection control unit 12 is used to upload an inspection task to the intelligent inspection module, determine it as a type of inspection task, and switch to the target inspection topology through task interpretation and scenario-based topology matching. It also uses a synaptic IoT protocol to execute topology-driven facility inspection control under the forward protocol.

[0072] In this embodiment of the application, a type of inspection control unit 12 is used to identify whether the task meets the judgment conditions of a type of inspection task after receiving the inspection task upload instruction, and execute the corresponding inspection logic if it is confirmed as a type of inspection task.

[0073] Among them, the first type of inspection task is a task actively generated by the system based on a preset cycle, strategy-driven or scheduling plan. The task instructions usually come from the internal planning area of ​​the central control platform or intelligent inspection module.

[0074] Specifically, after an inspection task is uploaded to the intelligent inspection module, the first inspection control unit 12 first calls the task identification logic to analyze the task type, trigger source, task tag, and execution priority of the task to determine whether it was actively generated by the intelligent inspection module or subjectively uploaded by the terminal device, and whether it does not contain external abnormal event trigger fields. If the above judgment logic is met, it is confirmed as a first-class inspection task.

[0075] Subsequently, after confirming the task type, the inspection control unit 12 calls the first decision node to perform structured interpretation of the task content. Specifically, the task interpretation process is completed based on a multi-dimensional rule set, including: inspection target identification (such as target equipment type, area number), inspection parameter extraction (such as sampling frequency, accuracy requirements, task duration), and anomaly tolerance range setting, etc. The interpretation result forms a task attribute vector, which is passed to the second decision node as input parameters.

[0076] Furthermore, after receiving the decoded task attribute vector, the second decision node performs a matching retrieval operation in a preset task scenario-inspection topology mapping library to find the target inspection topology that most closely matches the task characteristics. The target inspection topology is a directed graph structure reflecting the task execution path, containing inspection facilities represented by nodes and execution logic represented by edges. After matching is complete, the system switches to the target inspection topology, allowing inspection tasks to be executed according to a specific path order and prioritizing facilities with higher weights.

[0077] Then, the third decision node is triggered to convert the determined target inspection topology into readable parameter control information with specific inspection facilities as the main body.

[0078] Subsequently, the inspection control unit 12 invokes the forward protocol in the synaptic IoT protocol to construct a control link that propagates unidirectionally from the intelligent inspection module to the inspection infrastructure. The forward protocol supports three communication modes: command issuance, status request, and feedback reception, and features lightweight design, low latency, and high command consistency.

[0079] Specifically, one type of inspection control unit schedules corresponding facilities to perform data collection, action execution, or visual recognition operations between nodes on the inspection path according to the topological order. During the execution process, it makes intermediate adjustments based on feedback information and ensures the path is closed-loop.

[0080] In summary, the first-class inspection control unit 12 achieves accurate identification and efficient execution of proactive inspection tasks through the processing flow of task identification → interpretation → topology matching → forward protocol control, and constructs an intelligent inspection control link with tasks as the core, topology as the structure, and protocols as the communication.

[0081] The second-class inspection control unit 13 is used to transmit risk data from the self-inspection of the inspection infrastructure to the edge computing gateway, reverse the risk event and generate a second-class inspection task, trigger the lightweight inspection module embedded in the gateway, determine the event inspection topology and execute the topology-driven facility inspection control under the reverse protocol, wherein the inspection topology is optimized and adjusted through self-healing networking.

[0082] In this embodiment, the Class II inspection control unit 13 is used to identify and process the self-inspection risk data generated by the inspection infrastructure during the operation of the substation, and generate Class II inspection tasks that are highly correlated with the event, so as to realize rapid response and closed-loop handling of sudden anomalies.

[0083] For example, when a certain inspection infrastructure, such as a temperature sensor, partial discharge probe, visual monitoring node, or self-testing unit embedded in the substation infrastructure, detects abnormal data during normal operation, such as a sudden temperature rise, current fluctuation, or suspicious deformation in the image, the abnormal information is marked as risk data and uploaded to the edge computing gateway in real time through the communication loop.

[0084] Subsequently, after the edge computing gateway receives the risk data, the second-class inspection control unit 13 first calls the embedded event recognition model to perform structured parsing of the data, identify the abnormal trigger source, the scope of impact and the characteristics of related equipment, and reverse-engineer possible risk events based on historical scene data, equipment health models or alarm modes.

[0085] For example, if the temperature in a certain area continues to rise and is accompanied by frequent fluctuations in the signal of a certain type of connector, an abnormal event in the contact resistance of medium-voltage cables can be deduced from this, and this event can be used as input to generate a targeted risk inspection task.

[0086] Subsequently, the second-class inspection control unit 13 invokes the lightweight inspection module within the gateway, triggering the task distribution mechanism. The lightweight inspection module is an edge version of the intelligent inspection module, compressed and trained using an attention transfer mechanism, possessing low resource consumption, rapid deployment, and rapid response capabilities. Upon acquiring a risk event, this module quickly determines the corresponding event inspection topology based on its facility relationships and spatial distribution characteristics. That is, it expands outward from the event device as the central node to potentially affected related facilities, forming a highly concentrated regional task chain.

[0087] After the task topology is constructed, the Class II inspection control unit 13 schedules tasks according to the reverse protocol in the synaptic IoT protocol. The reverse protocol differs from the forward approach where commands are issued from the central point outwards; it is event-driven, transmitting distributed commands from the edge to the inspection side, emphasizing response speed, path resilience, and fault tolerance. Following the sequence in the event topology, relevant inspection facilities are scheduled to perform operations such as image capture, partial discharge detection, thermal imaging, or action verification, and the execution results are transmitted back in real time to support risk confirmation and extended judgment.

[0088] In a preferred embodiment provided in this application, during the execution of Type I and Type II inspections, if problems such as node communication failure, equipment malfunction, or insufficient power consumption occur along the path, a self-healing networking mechanism will be automatically invoked to optimize and adjust the current inspection topology in real time. The self-healing mechanism, based on redundant connections and state sharing between facility nodes, automatically reconstructs the task path or schedules backup equipment to ensure a complete closed loop for the inspection task.

[0089] For example, if the original path of robot A fails, robot B, which has redundant access capabilities, can be scheduled to continue the remaining tasks to ensure that the tasks are not interrupted.

[0090] In summary, the Class II inspection control unit 13 realizes a complete control chain from risk data triggering, event identification, lightweight scheduling to topology adaptive control and task closed-loop execution. It has event-driven, edge autonomy and path recovery capabilities, providing efficient, flexible and real-time anomaly handling capabilities for substation intelligent inspection systems.

[0091] Furthermore, the system also includes:

[0092] The mechanism introduction unit is used to introduce a self-healing networking mechanism for the inspection infrastructure; the mechanism triggering unit is used to trigger the self-healing networking mechanism based on the inspection facility status based on the inspection topology, and perform inspection optimization and adjustment.

[0093] In this embodiment, the mechanism introduction unit is used to embed a self-healing networking mechanism into the inspection infrastructure network structure during the inspection deployment phase. The self-healing networking mechanism refers to establishing a dynamic communication and task scheduling system among inspection equipment nodes with redundant connections, state sharing, and path self-adjustment capabilities. In an optional embodiment, this mechanism is borrowed from distributed fault-tolerant systems, allowing for the automatic selection of alternative paths or nodes in the event of local communication interruptions, equipment failures, or abnormal task execution, thereby ensuring the continuity and stability of inspection tasks.

[0094] In an optional embodiment, the mechanism introduction unit implements the deployment of the mechanism in the following ways: First, in the initial configuration of the system, an adjacency table is established for each inspection infrastructure node, and its redundant communication links and alternative functional nodes are marked; second, status broadcasting and heartbeat detection are introduced at the node level to ensure that the operating status, task execution progress and availability of each inspection infrastructure can be perceived in real time; finally, a path variability rule base is established between the edge computing gateway and the intelligent inspection module to support topology dynamic reconstruction operations.

[0095] Furthermore, the mechanism triggering unit is used to dynamically assess the continuity of task execution based on the operating status of each facility in the inspection topology during the execution of the inspection task, and to trigger the self-healing networking mechanism when execution anomalies or structural degradation are detected. The triggering judgment is based on a combination of multiple conditions, such as node communication timeout, status update frequency below a threshold, execution feedback anomalies, or path prediction mismatch.

[0096] For example, if a track robot in the inspection path fails to upload task feedback data within a predetermined time, or if the uploaded data indicates that the current path task cannot be continued, the mechanism triggering unit will determine that there is a risk of breakage in the inspection topology and immediately activate the self-healing mechanism.

[0097] Subsequently, after the self-healing mechanism is triggered, the current topology will be reassessed based on preset redundant connections and task priority rules, and the inspection path will be optimized and adjusted. This adjustment includes, but is not limited to, the following: switching to a backup node to execute the task (e.g., using a second drone to inspect the same area), reconstructing the task path (e.g., avoiding failed nodes and detouring to the target device), and reallocating tasks (e.g., transferring some tasks to other idle inspection units). After the optimization and adjustment are completed, the updated topology will be written into the current task flow as the new execution path, ensuring that the task is successfully executed in a closed loop within a dynamic network structure with a certain degree of fault tolerance.

[0098] In summary, the system possesses online fault tolerance, autonomous recovery, and path self-reconfiguration capabilities during the inspection process, effectively improving the robustness and system stability of inspection tasks in complex power Internet of Things environments.

[0099] Furthermore, the second type of inspection control unit includes:

[0100] The lightweight training unit is used to perform lightweight training on the intelligent inspection module using local event-oriented attention transfer to determine the lightweight inspection module; the module deployment unit is used to deploy the lightweight inspection module to the edge computing gateway.

[0101] In this embodiment, the lightweight training unit is used to perform structural compression and task transfer on the intelligent inspection module to meet the constraints on resource consumption, execution latency, and response speed in the edge computing environment.

[0102] Specifically, this unit uses a local event-oriented attention transfer mechanism as its core training strategy. Local event-oriented refers to high-frequency risk events or recurring task scenarios collected during task execution, such as frequent temperature rise alarms in a certain GIS interval area or frequent failures of the main transformer cooling fan. After identifying such local events, they are designated as key areas for the model to enhance the edge module's ability to discriminate in critical scenarios.

[0103] The attention transfer mechanism involves transferring the attention weight map from the original intelligent inspection module to the target lightweight model during the model lightweighting process, retaining its feature recognition ability for highly relevant regions, and eliminating redundant structures and low-contribution parameters.

[0104] In a feasible training process, the lightweight training unit constructs a dual-network structure: the main model retains the complete parameter structure, while the lightweight model is simplified through parameter pruning, channel sparsity, knowledge distillation, and other methods. The two are trained under attention map matching and output consistency constraints under shared input, thereby enabling the lightweight model to significantly reduce model size and computational complexity while maintaining the core performance of the main model.

[0105] After training, the determined lightweight inspection module is solidified as a deployable component, featuring fast loading, low-power execution, and optimized event response. It also possesses basic intelligent inspection capabilities such as task scheduling, path generation, and feedback processing in specific scenarios. This lightweight inspection module is suitable for resource-constrained deployment environments such as edge computing gateways, and can independently complete risk event response and local task handling without frequent reliance on the main control.

[0106] Furthermore, the module deployment unit is responsible for deploying the generated lightweight inspection module to the edge computing gateway node. The deployment process includes three stages: resource initialization, module registration, and communication interface binding.

[0107] Specifically, firstly, the module deployment unit initializes the operating environment according to the edge gateway hardware specifications, including configuring memory usage limits, computing resource scheduling weights, and network access permissions; secondly, it registers the lightweight module with the edge system's task scheduling platform to ensure that it has the authority and execution capability to receive abnormal data, trigger tasks, and provide feedback results; finally, by binding a communication interface with a synaptic IoT protocol, it ensures that the lightweight module can interact bidirectionally with the inspection infrastructure to achieve event-driven autonomous response.

[0108] In summary, the system has achieved a functional migration and capability decentralization from central intelligence to edge response nodes, and has built an edge intelligent inspection system with regional autonomy, rapid response, and resource sensitivity, effectively improving the overall real-time performance and reliability of the system.

[0109] Furthermore, after performing facility inspection and control, the system also includes:

[0110] The data feedback unit is used to acquire task inspection data and send it back to the edge computing gateway; the conflict handling unit is used to perform data conflict analysis and conflict resolution decisions on the task inspection data based on the source of inspections, and determine the valid inspection data, wherein the conflict resolution decision method is data analysis resolution or targeted re-inspection resolution; the risk management unit is used to send back the valid inspection data for substation risk location and alarm management.

[0111] In this embodiment, the data feedback unit collects and uploads the inspection data generated by the inspection infrastructure after the inspection task is completed, ensuring real-time monitoring of the actual execution status and results during the inspection process. The inspection data includes, but is not limited to, device status values ​​(such as voltage, current, and temperature), image information, behavior logs, abnormal event records, and task completion markers. The data feedback unit transmits this data in a structured format to the edge computing gateway via a high-speed communication channel, ensuring that data is not lost due to communication interruptions or delays. In a preferred embodiment, breakpoint resumption and data verification mechanisms are supported to improve data integrity and security.

[0112] Subsequently, the conflict resolution unit performs consistency checks and conflict resolution on the returned task inspection data. Its core logic is based on the principle of inspection homogeneity, analyzing whether multiple inspection facilities generate conflicting data within the same task cycle, on the same inspection object, or at the same task node. Specifically, conflict types include, but are not limited to: numerical differences in the status of the same equipment (e.g., equipment A uploads a temperature of 65℃, while equipment B uploads 80℃), inconsistent image recognition results, and conflicting task completion status markers. To address this, the conflict resolution unit constructs a data index table for the task cycle and uses timestamp, equipment number, and task identifier as a joint primary key to compare and analyze the data sources.

[0113] Furthermore, regarding conflict resolution strategies, two conflict resolution decision-making methods are supported: the first is data analysis resolution, which uses statistical methods, such as multi-source fusion, weighted averaging, and trend extrapolation, to analyze and process conflict data and generate unified and valid data items; the second is targeted re-inspection resolution, which automatically triggers a rescheduling mechanism when the conflict level or scope exceeds a preset threshold, instructing designated inspection facilities or backup channels to quickly re-inspect the conflict nodes, prioritizing the latest collected results as the final valid data. This processing flow ensures data consistency and accuracy under a multi-path, multi-terminal redundant inspection architecture.

[0114] Furthermore, the risk management unit analyzes and processes the valid inspection data filtered by the conflict resolution unit, and completes risk location and alarm management operations for substation equipment or areas. Specifically, by invoking a multimodal risk assessment model based on fault feature pattern recognition, status trend prediction, and scenario simulation, it correlates and interprets valid data to determine whether it constitutes a warning signal for equipment abnormality, operational deviation, or potential fault. If a risk event is identified, an alarm response process is triggered according to preset alarm rules, including alarm level classification (such as yellow warning, red fault), notification of responsible persons, and generation of maintenance suggestions, and the results are fed back to the intelligent inspection module and the substation monitoring center to form a closed-loop system control.

[0115] Optionally, the multimodal risk assessment model is a functional unit that takes detection data as input and outputs real-time faults and predicted faults. It can employ a neural network architecture and be constructed using sample training.

[0116] In summary, a complete post-processing chain has been established, from inspection data collection and conflict resolution to risk warning and response. This enables data-driven management of highly reliable inspection results and coordinated risk handling, significantly improving the operational stability of the intelligent inspection system and the predictability of substation operation and maintenance management.

[0117] The intelligent substation inspection auxiliary system in the power Internet of Things environment provided in this application has the following technical effects:

[0118] 1. Task Scenario - Inspection Topology Definition: Distinguishing between routine inspections and substation infrastructure, clustering and mining scenario features based on historical inspection records, and associating task scenarios with inspection topologies. This enables scenario-based pre-planning of inspection paths, improving inspection targeting, reducing redundant operations, and enhancing inspection efficiency. Deploying a decision chain encompassing task interpretation, topology switching, and parameter control, and achieving multi-scale task analysis and response through data-driven training. Supporting intelligent parsing and execution of complex inspection tasks, ensuring the accuracy and automation level of the inspection process.

[0119] 2. Two-way Protocol-Driven Inspection: Employing a synaptic-like IoT protocol, two types of tasks involve proactive inspection via a forward protocol and passive inspection triggered by risk data via a reverse protocol. This dual inspection mechanism, combining proactive prevention and passive response, ensures timely risk detection and handling, enhancing system reliability. Self-Healing Network Optimization: A self-healing network mechanism is triggered based on the status of the inspected facilities, dynamically adjusting the inspection topology. This adapts in real-time to equipment anomalies or environmental changes, ensuring the stability of the inspection link and reducing the risk of missed inspections.

[0120] 3. Edge Computing and Data Processing: The edge computing gateway deploys a lightweight inspection module to perform conflict analysis and resolution on the transmitted data, locate risks, and issue alarms. This enables rapid local data processing, reduces transmission latency, accurately identifies risk sources, and improves the substation's safety management capabilities.

[0121] Through the foregoing detailed description of the substation intelligent inspection auxiliary system in the power Internet of Things environment, those skilled in the art can clearly understand the substation intelligent inspection auxiliary system in the power Internet of Things environment in this embodiment. As for the device disclosed in the embodiment, since it corresponds to the system disclosed in the embodiment, the description is relatively simple, and relevant parts can be referred to the system section description.

[0122] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A substation intelligent inspection auxiliary system in a power Internet of Things environment, characterized in that, The system comprises: a development unit configured to obtain infrastructure of a target substation, develop an intelligent inspection module in a substation intelligent inspection auxiliary system by defining a task scene-inspection topology, and a first-class inspection control unit configured to, if a task is uploaded to the intelligent inspection module, determine that the task is a first-class inspection task, match the task with the scene topology, switch to a target inspection topology, assist a synaptic-like protocol, and execute facility inspection control driven by the topology under a forward protocol; a second-class inspection control unit configured to, if there is self-inspection of the inspection infrastructure to risk data, transmit the risk data to an edge computing gateway, inversely deduce a risk event and generate a second-class inspection task, trigger a light inspection module embedded in the gateway, determine an event inspection topology, and execute facility inspection control driven by the topology under a reverse protocol; wherein the inspection topology is optimized and adjusted through self-healing networking; wherein the development unit comprises: a first deployment unit configured to deploy a first decision node by task interpretation, wherein the interpretation dimension comprises at least a molecular scale, a component scale, and a system scale; a second deployment unit configured to deploy a second decision node by task scene matching and inspection topology switching; a third deployment unit configured to deploy a third decision node by control decision of the topology inspection parameters; a module construction unit configured to construct an inspection decision chain according to the first decision node, the second decision node, and the third decision node, and perform data-driven training until convergence, serving as the intelligent inspection module. 2.The substation intelligent inspection assisting system in a power internet of things environment of claim 1, wherein, The development unit comprises: a facility classification unit configured to determine inspection infrastructure and substation infrastructure for the infrastructure, wherein the inspection infrastructure is identified with an inspection domain based on the substation infrastructure, and the inspection infrastructure includes substation access infrastructure and dynamic inspection infrastructure; a scene reconstruction unit configured to perform topology reconstruction based on a substation scene for the inspection infrastructure, and define the task scene-inspection topology. 3.The substation intelligent inspection assisting system in a power internet of things environment of claim 2, wherein, The scene reconstruction unit comprises: a clustering unit configured to call inspection records of a target substation, perform scene clustering, and determine M record groups; a mining unit configured to mine scene inspection facilities-inspection orientations within a group for the M record groups with a preset ratio in the group as a constraint; a determination unit configured to determine an inspection topology according to the scene inspection facilities-inspection orientations; an association unit configured to define a task scene for the M record groups, and perform association of the task scene and the inspection topology. 4.The substation intelligent inspection assisting system in a power internet of things environment of claim 1, wherein, The intelligent inspection module, the edge computing gateway, and the inspection infrastructure establish a communication loop; wherein inspection scheduling management is performed by deploying a synaptic-like protocol, wherein the synaptic-like protocol is an event-driven type of industrial internet of things interaction protocol, including a forward protocol and a reverse protocol. 5.The substation intelligent inspection assisting system in a power internet of things environment of claim 4, wherein, The inspection task includes a first-class inspection task and a second-class inspection task; wherein the first-class inspection task is an active inspection task of the intelligent inspection module-inspection infrastructure, and the second-class inspection task is a passive inspection task of the inspection infrastructure-edge computing gateway-inspection infrastructure; wherein the passive inspection task is triggered by a self-checking risk trigger event. 6.The substation intelligent inspection assisting system in a power internet of things environment of claim 5, wherein, The first type of inspection task adopts a forward protocol based on a synapse-like protocol, and the second type of inspection task adopts a reverse protocol based on a synapse-like protocol.

7. The intelligent inspection auxiliary system for transformer substation in power internet of things environment of claim 1, wherein, The system further comprises: A mechanism introduction unit is configured to introduce a self-healing networking mechanism for the inspection infrastructure. A mechanism triggering unit is configured to trigger the self-healing networking mechanism to perform inspection optimization adjustment based on the state of the inspection infrastructure. 8.The substation intelligent inspection assisting system in a power internet of things environment of claim 1, wherein, The second type of inspection control unit comprises: A lightweight training unit is configured to perform lightweight training on the intelligent inspection module based on local event-oriented attention migration to determine a lightweight inspection module. A module deployment unit is configured to deploy the lightweight inspection module to an edge computing gateway. 9.The substation intelligent inspection assisting system in a power internet of things environment of claim 1, wherein, After performing the facility inspection control, the system further comprises: A data back transmission unit is configured to acquire task inspection data and transmit the task inspection data to the edge computing gateway. A conflict processing unit is configured to perform data conflict analysis and conflict resolution strategy on the task inspection data based on inspection homology to determine effective inspection data, wherein the conflict resolution strategy is data analysis resolution or directional re-inspection resolution. A risk management unit is configured to perform substation risk positioning and alarm management on the effective inspection data.

Images