Multi-device cooperative transformer substation inspection method and system

Abstract

The invention discloses a multi-device cooperative transformer substation inspection method and system, through deep cooperation and data fusion of an unmanned aerial vehicle and a quadruped robot, on one hand, the problem that a single inspection device cannot effectively cover transformer substation inspection is solved, and on the other hand, the inspection efficiency is improved. Automatic distribution of inspection point positions of the quadruped robot and the unmanned aerial vehicle in the same task can be completed, cooperative distribution of the quadruped robot and the unmanned aerial vehicle is realized, and a cooperative inspection scheme of the quadruped robot and the unmanned aerial vehicle is automatically generated, so that the cooperative capability is improved; meanwhile, through data fusion, the problem of data splitting existing in the traditional technology is also avoided, and the fault detection precision is further improved; therefore, the effectiveness, the precision, the safety and the intelligent level of electric power inspection can be improved, so that the method is very suitable for large-scale application and popularization.

Description

Technical Field

[0001] This invention belongs to the field of power inspection technology, specifically relating to a substation inspection method and system with multi-device collaboration. Background Technology

[0002] Currently, substation inspections are generally conducted manually, relying on sensory evaluation of operating equipment, primarily through sight, touch, hearing, and smell. However, manual inspections are significantly affected by the physiological and psychological state of personnel, their sense of responsibility, the external working environment, experience, and technical skills. This leads to the possibility of missed inspections and undetected defects. Furthermore, operators lack specialized instruments and cannot detect internal equipment defects through simple inspections, such as excessive oil and gas test parameters, overheating in specific equipment parts, or inadequate insulation. Simultaneously, with the increasing number of unmanned substations, many are located far apart. In the event of an accident or after strong winds, heavy snow, or thunderstorms, when the control station is unable to dispatch vehicles for timely inspections, control station operators may not be able to promptly understand the status of on-site equipment and identify potential hazards, thus jeopardizing the safe operation of the power grid. Therefore, manual inspections suffer from low timeliness, low efficiency, and poor reliability.

[0003] Meanwhile, with the development of intelligent inspection technology, robots and drones have gradually been applied to the inspection process of substations. However, existing robots and drones have the following shortcomings during inspection: (1) Existing technologies usually use a single robot or a single drone for inspection. Drones are difficult to hover in complex environments such as densely populated equipment areas and indoor substations, while quadruped robots cannot efficiently cover large areas. Therefore, it is difficult for a single inspection device to achieve effective coverage of substation inspection. (2) Even if there is an inspection method that uses robots and drones in combination, there are problems of weak collaboration and data fragmentation. That is, the robot and drone inspection points cannot be allocated according to different inspection tasks, and collaborative inspection schemes cannot be automatically generated. Furthermore, the data of drones and ground robots are stored independently and lack real-time fusion, resulting in low fault detection accuracy. Therefore, based on the aforementioned shortcomings, how to provide a multi-device collaborative substation inspection method that achieves effective coverage of substation inspection and improves fault detection accuracy through deep collaboration between drones and quadruped robots has become an urgent problem to be solved. Summary of the Invention

[0004] The purpose of this invention is to provide a substation inspection method and system with multi-device collaboration, in order to solve the problems that existing technologies have, such as the difficulty for a single inspection device to effectively cover the substation inspection, and the weak collaboration capability and data fragmentation of the hybrid inspection method.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: Firstly, a method for multi-device collaborative substation inspection is provided, applied to a multi-device collaborative substation inspection system, wherein the substation inspection system includes at least one quadruped robot and at least one unmanned aerial vehicle (UAV), and the method includes: Obtain the equipment ledger information of the substation, and based on the equipment ledger information, configure and associate the inspection points of each piece of equipment to be inspected in the substation to obtain the inspection point configuration and association results. Based on the equipment ledger information and the configuration and association results of inspection points, the input parameters for the inspection plan are generated. The inspection scheme input parameters are input into the decision tree model to output inspection schemes for the equipment to be inspected with different task priorities. The inspection scheme for any task priority includes the inspection equipment assigned to each inspection point in the equipment to be inspected corresponding to that task priority, as well as the execution path of the inspection equipment in the inspection scheme corresponding to that task priority. The inspection equipment at any inspection point is a quadruped robot or a drone. Each inspection plan is sent to its corresponding inspection equipment so that each inspection equipment can perform an inspection of the substation according to the received inspection plan and upload the inspection data. Receive multiple inspection data from the substation within a specified time period, and filter out the inspection data of the points to be inspected from the multiple inspection data. The points to be inspected are the inspection points jointly inspected by drones and quadruped robots in the multiple inspection data. The inspection data of the points to be inspected are fused to obtain fused data, and the inspection results of the points to be inspected are generated based on the fused data.

[0006] Based on the above-disclosed content, this invention employs a hybrid inspection method for substation inspection. Specifically, the substation is equipped with at least one quadruped robot and at least one drone. Simultaneously, by acquiring the substation's equipment ledger information, the inspection points for the drone and quadruped robot are automatically allocated, thus forming inspection schemes with different task priorities. Specifically, this invention first configures and associates inspection points for each piece of equipment to be inspected in the substation based on the equipment ledger information, obtaining the inspection point configuration and association results. Then, based on the inspection point configuration and association results, the input parameters for the inspection scheme are generated. Then... Based on the input parameters of the inspection plan and with the help of a decision tree model, inspection plans for the equipment to be inspected with different task priorities are generated. Each inspection plan for any task priority includes the inspection equipment assigned to each inspection point on the equipment to be inspected for that task priority, as well as the execution path of the inspection equipment in the inspection plan for that task priority. In this way, the present invention realizes the allocation of inspection equipment to each inspection point on the equipment to be inspected with different inspection priorities, that is, it completes the automatic allocation of inspection points for UAVs and quadruped robots in the same task, thereby realizing the collaborative work of the two.

[0007] Next, each inspection device can perform a single inspection of the substation according to its own inspection plan, obtaining inspection data. Then, this invention also performs data fusion on multiple inspection data of the substation within a specified time period, that is, from the multiple inspection data, the inspection data of the inspection points jointly inspected by the drone and the quadruped robot (i.e., the inspection data of the points to be inspected) is selected; then, the selected inspection data is fused to obtain fused data; finally, the inspection results of the points to be inspected can be generated based on the fused data. In this way, this invention fuses the data of the drone and the quadruped robot and uses it to generate inspection results, avoiding the data fragmentation problem existing in traditional technologies, thereby improving the accuracy of fault detection.

[0008] Through the above design, this invention, by deeply collaborating and fusioning data between drones and quadruped robots, solves the problem that a single inspection device cannot effectively cover substation inspections. Furthermore, it enables automatic allocation of inspection points for both quadruped robots and drones in the same task, achieving collaborative allocation and automatically generating a collaborative inspection plan, thereby improving collaborative capabilities. Simultaneously, data fusion avoids the data fragmentation problem inherent in traditional technologies, thus improving fault detection accuracy. Therefore, this invention enhances the effectiveness, accuracy, safety, and intelligence of power inspection, making it highly suitable for large-scale application and promotion.

[0009] In one possible design, based on equipment ledger information, inspection point configuration and association processing are performed for each piece of equipment to be inspected in the substation, resulting in the following inspection point configuration and association results: Based on the equipment ledger information, identify all the inspection components on each piece of equipment to be inspected, and use each inspection component as the initial inspection point. Based on the equipment ledger information, the location attributes of each initial inspection point are marked to obtain multiple inspection points; Based on the location attributes of multiple inspection points, inspection points that are inspected only by drones or quadruped robots are selected from the multiple inspection points as independent points, and inspection points in the same inspection area that are inspected by both drones and quadruped robots are selected as identical points. Based on equipment ledger information, related inspection points are extracted from multiple inspection points, and the extracted inspection points are used as related points to generate related task groups. The inspection point configuration and association results are generated using independent points, identical points, and associated task groups.

[0010] In one possible design, the equipment ledger information includes: equipment type and priority, basic equipment attributes and inspection requirements for each piece of equipment to be inspected. Among them, equipment type and priority include the inspection type of each piece of equipment to be inspected and the priority of each inspection type. Basic equipment attributes include the inspection equipment labels corresponding to different inspection components on each piece of equipment to be inspected. The inspection types include routine inspection, emergency inspection and designated equipment inspection. The inspection plan input parameters are generated based on equipment ledger information and the configuration and association results of inspection points, including: Based on the inspection type and priority of each inspection type of each equipment to be inspected, several inspection tasks and the priority of each inspection task are generated. Among them, the inspection tasks include emergency inspection tasks, designated equipment inspection tasks and regular inspection tasks, and the priority of emergency inspection tasks, designated equipment inspection tasks and regular inspection tasks decreases in that order. Based on several inspection tasks and the priority of each inspection task, a task priority is generated; Based on the inspection equipment labels of the inspection components corresponding to each independent point and each associated point in the associated task group, determine the inspection equipment type corresponding to each independent point and each associated point. Obtain historical inspection data for each identical point in the inspection point configuration and association results, and determine the type of inspection equipment corresponding to each identical point based on the historical inspection data of each identical point. Based on the inspection requirements of each piece of equipment to be inspected, generate personnel configuration information for each independent location, each identical location, and each related location. By utilizing the types of inspection equipment corresponding to each independent point, each identical point, and each related point, as well as the personnel configuration information for each independent point, each identical point, and each related point, task resource requirements are generated. Determine the inspection time for each identical point, and use the inspection time for each identical point to generate the estimated inspection time. The inspection plan input parameters are composed of the task priority, the task resource requirements, the estimated inspection time, and the associated task groups in the inspection point configuration and association results.

[0011] In one possible design, the inspection plan input parameters are input into a decision tree model to output inspection plans for equipment to be inspected with different task priorities, including: Based on the first layer nodes in the decision tree model and according to the task priority, the current optimal task to be executed is determined; By utilizing the second-level nodes in the decision tree model, and based on the task resource requirements and the estimated inspection time, or based on the associated task group and the task resource requirements, an inspection plan for the equipment to be inspected corresponding to the current optimal task is generated. After all inspection tasks in the task priority have been polled, inspection plans for the equipment to be inspected with different task priorities are obtained.

[0012] In one possible design, each inspection area of ​​the substation corresponds to a quadruped robot and / or drone. Utilizing the second-level nodes of a decision tree model, and based on task resource requirements and estimated inspection time, or based on the associated task group and task resource requirements, an inspection plan for the equipment to be inspected corresponding to the current optimal task is generated, including: If the current optimal task to be performed is an emergency inspection task or a regular inspection task, then select each inspection point in the equipment to be inspected that is an emergency inspection or a regular inspection task, and use it as the target point. Filter out independent points and identical points from the target points; Based on the task resource requirements, determine the inspection equipment type corresponding to the independent points in the target points, and the inspection equipment type corresponding to the same points in the target points. Based on the inspection area to which the independent points in the target location belong and the corresponding inspection equipment type, the inspection equipment for each independent point in the target location is determined, and each independent point in the target location is assigned to its respective inspection equipment. The inspection equipment for any independent point is a quadruped robot or drone corresponding to the inspection area to which that independent point belongs. After completing the allocation of independent points in the target location, the inspection equipment for each identical point in the target location is determined by using the inspection area to which each identical point in the target location belongs and the corresponding inspection equipment type. Based on the estimated inspection time, the inspection time of each identical point in the target location is determined, and each identical point in the target location is assigned to its corresponding inspection equipment in order of inspection time from low to high. Based on the target points assigned to the target devices, an execution path for the target devices is generated, wherein the target devices are all the inspection devices corresponding to the target points; By utilizing the execution path of the target device, as well as the inspection equipment and personnel configuration information corresponding to each independent point and each identical point in the target location, an inspection plan for the device to be inspected corresponding to the current optimal execution task is generated.

[0013] In one possible design, utilizing the second-level nodes of the decision tree model, and based on task resource requirements and estimated inspection time, or based on the associated task group and task resource requirements, an inspection plan for the equipment to be inspected corresponding to the current optimal task is generated, including: If the current optimal task to be executed is the specified equipment inspection task, then from the associated task group, each inspection point and the associated point corresponding to each inspection point in the equipment to be inspected with the inspection type of specified equipment inspection are selected, so as to form a target associated group using the selected points. The associated points in the target association group are divided into indoor and outdoor key equipment points and indoor and outdoor associated equipment points; Based on the task resource requirements, the inspection equipment corresponding to the key indoor and outdoor equipment locations is determined, and the key indoor and outdoor equipment locations are assigned to the corresponding inspection equipment. After completing the allocation of key indoor and outdoor equipment locations, based on the task resource requirements, the inspection equipment corresponding to the related indoor and outdoor equipment locations is determined, and the related indoor and outdoor equipment locations are assigned to the corresponding inspection equipment. Each inspection device corresponding to the target association group is designated as a device, and an execution path for each designated device is generated based on the device location assigned to each device. By utilizing the execution paths of each designated device, as well as the inspection equipment and personnel configuration information corresponding to key indoor and outdoor equipment locations and related indoor and outdoor equipment locations, an inspection plan for the equipment to be inspected corresponding to the current optimal execution task is generated.

[0014] In one possible design, the inspection data of the points to be inspected is image data, and the image data includes a first image uploaded by a drone and a second image uploaded by a quadruped robot. This involves data fusion processing of the inspection data for the points to be inspected, obtaining fused data, and generating inspection results for the points to be inspected based on the fused data, including: Image alignment processing is performed on the first image and the second image to obtain a first aligned image and a second aligned image; Perform time alignment processing on the first aligned image and the second aligned image, and determine whether the first aligned image and the second aligned image meet the time constraints; If so, feature stitching is performed on the first aligned image and the second aligned image to obtain a first feature image, wherein the first feature image is used as the fused data; The first feature image is processed by using the CBAM attention network to obtain the second feature image; Based on the second feature image, a temporal feature sequence is generated; The time-series feature sequence is input into the defect detection model to obtain the defect detection results of the points to be inspected, and the defect detection results are used as the inspection results.

[0015] In one possible design, the EXIF ​​information of the first image and the EXIF ​​information of the second image both embed the UTC time at the time of image acquisition. The process of performing time alignment processing on the first and second aligned images and determining whether the first and second aligned images meet the time constraints includes: Extract the UTC time from the EXIF ​​information of the first image, and extract the UTC time from the EXIF ​​information of the second image; Calculate the time difference between the UTC time of the first image and the UTC time of the second image; Determine whether the time difference is less than or equal to a preset duration; If so, then the first aligned image and the second aligned image are determined to satisfy the time constraint.

[0016] In one possible design, the first aligned image and the second aligned image are feature-stitched together to obtain the first feature image, which includes: Feature extraction is performed on the first aligned image to obtain a first feature sub-image, and feature extraction is performed on the second aligned image to obtain a second feature sub-image; The first feature image is obtained by concatenating the first feature sub-image and the second feature sub-image. Accordingly, based on the second feature image, the generation of a temporal feature sequence includes: Using a sliding window, the second feature image is sampled to obtain multiple sampled features, where each sliding window corresponds to a time step; Multiple sampled features are sorted sequentially according to the time steps of the corresponding sliding window to generate the time-series feature sequence.

[0017] Secondly, a multi-device collaborative substation inspection system is provided, including: a substation host and at least one quadruped robot and at least one drone; The substation host is used to obtain the equipment ledger information of the substation, and based on the equipment ledger information, to configure and associate the inspection points of each piece of equipment to be inspected in the substation, and to obtain the inspection point configuration and association results. The substation main unit is used to generate inspection plan input parameters based on equipment ledger information and inspection point configuration and association results; The substation host is used to input the inspection scheme input parameters into the decision tree model to output the inspection schemes for the equipment to be inspected with different task priorities. The inspection scheme for any task priority includes the inspection equipment assigned to each inspection point of the equipment to be inspected corresponding to that task priority, as well as the execution path of the inspection equipment in the inspection scheme corresponding to that task priority. The inspection equipment for any inspection point is a quadruped robot or a drone. The substation main unit is used to distribute various inspection plans to their respective inspection equipment. Each inspection device is used to perform a single inspection of the substation according to the received inspection plan and upload the inspection data. The substation host is used to receive multiple inspection data of the substation within a specified time period, and to filter the inspection data of the points to be inspected from the multiple inspection data. The points to be inspected are the inspection points jointly inspected by drones and quadruped robots in the multiple inspection data. The substation main unit is also used to perform data fusion processing on the inspection data of the points to be inspected, obtain fused data, and generate the inspection results of the points to be inspected based on the fused data.

[0018] Thirdly, a multi-device collaborative substation inspection device is provided. Taking the device as an electronic device as an example, it includes a memory, a processor, and a transceiver that are connected in sequence. The memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer programs and execute the multi-device collaborative substation inspection method as described in the first aspect or any possible design in the first aspect.

[0019] Fourthly, a storage medium is provided, on which instructions are stored, which, when executed on a computer, perform the multi-device collaborative substation inspection method as described in the first aspect or any possible design of the first aspect.

[0020] Fifthly, a computer program product containing instructions is provided, which, when executed on a computer, causes the computer to perform the multi-device collaborative substation inspection method as described in the first aspect or any possible design of the first aspect.

[0021] Beneficial effects: (1) This invention solves the problem that a single inspection device cannot effectively cover the substation inspection by deep collaboration and data fusion between UAV and quadruped robot. On the other hand, it can automatically allocate the inspection points of quadruped robot and UAV in the same task, realize the collaborative allocation between the two, and automatically generate the collaborative inspection plan, thereby improving the collaborative capability. At the same time, through data fusion, it also avoids the problem of data fragmentation in traditional technology, thereby improving the fault detection accuracy. Therefore, this invention can improve the effectiveness, accuracy, safety and intelligence level of power inspection, and is very suitable for large-scale application and promotion. Attached Figure Description

[0022] Figure 1 A flowchart illustrating the steps of a multi-device collaborative substation inspection method provided in an embodiment of the present invention; Figure 2 This is a structural diagram of a multi-device collaborative substation inspection system provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in conjunction with the accompanying drawings and descriptions of the embodiments or the prior art. Obviously, the following description of the structure of the accompanying drawings is 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. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0024] It should be understood that although the terms first, second, etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit, without departing from the scope of the exemplary embodiments of the invention.

[0025] It should be understood that the term "and / or" that may appear in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" that may appear in this document describes another relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " that may appear in this document generally indicates that the related objects before and after it are in an "or" relationship.

[0026] Example: Before describing the substation inspection method provided in this embodiment, a multi-device collaborative substation inspection system is first provided. This system includes a substation host, at least one quadruped robot, and at least one drone. Each quadruped robot and each drone is communicatively connected to the substation host, enabling the distribution of inspection plans and the uploading of inspection data. Simultaneously, each inspection area of ​​the substation corresponds to one quadruped robot and / or drone. Thus, when generating an inspection plan, the corresponding inspection points can be assigned to the quadruped robot or drone in the inspection area corresponding to each piece of equipment to be inspected. This allows the inspection of equipment in the corresponding inspection area to be carried out using the quadruped robot and / or drone corresponding to each inspection area. Of course, this assignment refers to a single task; that is, in one task, a point can only be assigned to a quadruped robot or drone. However, in multiple tasks, the same point may be inspected jointly by a quadruped robot and a drone. For example, a quadruped robot might be used in the first task, while a drone might be used in the second.

[0027] Furthermore, in this embodiment, the aforementioned system adopts a virtual self-organizing network mode. Due to the special nature of substations, the network self-organizing phase mainly relies on edge nodes on drones and quadruped robots. In the fixed mode, the substation host serves as the ground control center node, while in the mobile mode, the mobile work vehicle serves as the ground control center node. The mobile mode refers to the deployment of drones and quadruped robots on mobile work vehicles, while the fixed mode refers to the deployment of drones and quadruped robots within the substation.

[0028] Furthermore, this embodiment also includes a disconnection detection mechanism, triggered by the following conditions: the substation host does not receive a heartbeat packet for one minute consecutively, plus a weak signal warning (signal strength is below the strength threshold). The recovery strategy is as follows: the disconnected node activates the storage mechanism and quickly accesses the storage node after reconnection. That is, when disconnected, data is stored on the onboard end, no data is transmitted back, and the storage node is recorded. In addition, when the battery power is insufficient, the device cannot be located, or other situations prevent the device from continuing inspection, local task priority adjustment, automatic home reconnection, and data recording strategies can be triggered.

[0029] Therefore, based on the aforementioned inspection system, this embodiment provides a multi-device collaborative substation inspection method; see [link to relevant documentation]. Figure 1 As shown in the figure, the multi-device collaborative substation inspection method provided in this embodiment uses a quadruped robot and a drone to jointly inspect the substation, solving the problem that a single inspection device cannot effectively cover the entire substation. Simultaneously, this invention generates inspection point configurations and association results based on the substation's equipment ledger information. Then, based on the inspection point configurations and association results, it generates inspection plan input parameters and uses a decision tree model to generate an inspection plan containing the inspection devices assigned to each inspection point on each device to be inspected and the corresponding execution paths of those devices. Thus, this invention completes the multi-device collaborative inspection of the same task. The automatic allocation of inspection points between drones and quadruped robots can automatically generate collaborative inspection plans, thus achieving deep collaboration between the two. Finally, for inspection data of points jointly inspected by drones and quadruped robots in multiple inspections, this method also performs data fusion to generate inspection results. This avoids the data fragmentation problem of traditional technologies, thereby improving fault detection accuracy. Based on this, this method, through deep collaboration between drones and quadruped robots and real-time data fusion, can improve the efficiency, safety, and intelligence level of power inspection, making it highly suitable for large-scale application and promotion.

[0030] For example, this method can be operated on, but is not limited to, substation mainframes, drones, and quadruped robots. Preferably, this embodiment uses a substation mainframe as an example for illustration. It is understood that the aforementioned execution subject does not constitute a limitation on the embodiments of this application. Accordingly, the operation steps of this method can be, but are not limited to, the steps S1 to S6 below.

[0031] S1. Obtain the equipment ledger information of the substation, and based on the equipment ledger information, configure and associate inspection points for each piece of equipment to be inspected in the substation to obtain the inspection point configuration and association results. In this embodiment, the equipment ledger information is the basic data source for collaborative task allocation, and its content needs to support inspection point configuration, resource requirement analysis, and task association relationship establishment. At the same time, the equipment ledger information can be obtained at preset intervals so that corresponding inspection plans can be generated at each preset interval, thereby conducting substation inspections at each preset interval. In this embodiment, the process of a single inspection of multiple devices in collaboration is illustrated using the equipment ledger information obtained at one time as an example, as shown below.

[0032] Specifically, equipment ledger information may include, but is not limited to: basic equipment attributes, location and environmental attributes, equipment type and priority, relationships between equipment, and inspection requirements for each piece of equipment to be inspected; of course, the aforementioned equipment ledger information is pre-set by staff.

[0033] Furthermore, the basic attributes of the equipment may include, but are not limited to: the ID, name, model, and interval to which the equipment belongs (such as the main control room or main transformer equipment area), installation location (latitude and longitude coordinates, indoor equipment layout information), and the inspection equipment labels corresponding to different inspection components on each of the equipment to be inspected; among them, the inspection equipment labels are used to indicate the type of inspection equipment required for the inspection component, such as inspection equipment label 1 indicating that the type of inspection equipment required is a drone, inspection equipment label 2 indicating that the type of inspection equipment required is a quadruped robot (such as a robot dog), etc.; in this way, the basic attributes of the equipment can serve as a unique identifier for the equipment and locate the physical location of the inspection point and the type of inspection equipment required.

[0034] Meanwhile, location and environmental attributes may include, but are not limited to: the deployment area of ​​the equipment to be inspected (such as indoor or outdoor deployment areas), environmental characteristics: high-altitude equipment / ground equipment, enclosed space / open area, etc.; and the point attribute labeling used for inspection points to help determine the type of equipment to be inspected. For example, if there is an independent outdoor point and the inspection equipment label is 1, then the inspection equipment can be identified as a drone, etc.

[0035] In this embodiment, the equipment type and priority include the inspection type of each device to be inspected and the priority of each inspection type. The inspection types include routine inspection, emergency inspection, and designated equipment inspection (special inspection, referring to user-specified devices to be inspected). Emergency inspection has a higher priority than designated equipment inspection, and designated equipment inspection has a higher priority than routine inspection. That is, emergency inspection must be processed first, then designated equipment inspection, and finally routine inspection. At the same time, inspection type field information can be added to each device to be inspected in advance, such as the inspection type field information of the core device (main transformer) being emergency inspection. In this way, when performing multi-device collaborative inspection, the inspection type of each device to be inspected can be determined according to the inspection type field information of each device.

[0036] Furthermore, the inspection requirements for any piece of equipment to be inspected may include, but are not limited to: inspection items: visible light inspection, infrared inspection, soundprint inspection, etc.; required tools: adding soundprint, adding ultraviolet, etc.; team skills: personnel configuration; whether it is a special inspection, the relationship of related equipment groups such as personnel operation required (the relationship of the entire link from outdoor to indoor of the No. 1 main transformer); the estimated time for each inspection part of each piece of equipment to be inspected (each inspection part is an inspection point), etc.; in this way, the inspection requirements are used to determine the estimated inspection time and the team skills in the resource requirements.

[0037] Finally, the relationships between devices include the associated devices when each device to be inspected is subjected to associated inspection.

[0038] Thus, after obtaining the equipment ledger information, the inspection points can be configured and associated based on this information. The process can be, but is not limited to, the steps S11 to S15 below.

[0039] S11. Based on the equipment ledger information, determine all inspection components on each piece of equipment to be inspected, and use each inspection component as an initial inspection point. In this embodiment, for example, but not limited to, determining all inspection components on each piece of equipment to be inspected based on the aforementioned basic equipment attributes, such as a main transformer being divided into its physical appearance, oil temperature gauge, and breather, etc. Therefore, a main transformer can be divided into "initial inspection point for physical appearance," "initial inspection point for oil temperature gauge reading," "initial inspection point for breather inspection," etc. After obtaining the initial points, the point ID, name, and coordinates will also be recorded. Of course, the above examples are just examples, and this embodiment is not limited to this.

[0040] After the point splitting is completed, point attribute labeling can be performed, as shown in step S12 below.

[0041] S12. Based on the equipment ledger information, mark the location attributes of each initial inspection point to obtain multiple inspection points. In specific applications, this embodiment marks the initial inspection points of each device to be inspected based on the location and environmental data of each device to be inspected, as well as the inspection equipment tags required for different inspection components in the basic attributes of the equipment, to obtain the inspection points of each device to be inspected. For example, the location attributes include independent indoor points, independent outdoor points, and shared inspections. Specifically, for any initial inspection point, it is assumed that the deployment area of ​​the corresponding device to be inspected is indoors, and the environmental characteristics are ground equipment and enclosed space. If there is only one inspection equipment tag for any initial inspection point, then the point attribute of any initial inspection point is marked as an independent indoor point. Alternatively, if the deployment area of ​​the equipment to be inspected corresponding to any initial inspection point is outdoors, and the environmental characteristics are high-altitude equipment and open areas, and there is only one inspection equipment tag for any initial inspection point, then the point attribute of any initial inspection point is marked as an independent outdoor point. Or, if there are two inspection equipment tags corresponding to any initial inspection point, such as tag 1 and tag 2, then the point attribute of any initial inspection point is marked as a joint inspection.

[0042] Thus, after completing the point attribute labeling of each initial inspection point, point differentiation can be performed, as shown in step S13 below.

[0043] S13. Based on the location attributes of multiple inspection points, select inspection points that are inspected only by drones or quadruped robots as independent points, and select inspection points in the same inspection area that are inspected jointly by drones and quadruped robots as identical points. In this embodiment, inspection points with the location attributes of indoor independent points and outdoor independent points are designated as independent points, while points with the location attribute of joint inspection are designated as identical points. At the same time, if more than one identical point has the same location, the identical points with the same location are merged into a new identical point. Thus, one identical point can represent one point or multiple points with the same location.

[0044] After obtaining independent and identical locations, task collaboration relationships can be established, as shown in step S14 below.

[0045] S14. Based on the equipment ledger information, extract the inspection points with related relationships from multiple inspection points, and use the extracted inspection points as related points to generate related task groups. In specific implementation, based on the equipment relationship in the equipment ledger information, determine the related equipment of any equipment to be inspected, and then use the inspection point corresponding to the equipment to be inspected and the inspection point corresponding to its related equipment as related points to generate related task groups.

[0046] After obtaining the associated task group, the aforementioned independent and identical points can be combined to form the inspection point configuration and association results, as shown in step S15 below.

[0047] S15. Using independent locations, identical locations, and associated task groups, generate the inspection point configuration and association results.

[0048] After completing the configuration and association of inspection points for each piece of equipment to be inspected in the substation through the aforementioned steps S11 to S15, the inspection plan generation parameters can be generated based on this, as shown in step S2 below.

[0049] S2. Based on the equipment ledger information and the configuration and association results of inspection points, generate the inspection plan input parameters. In specific applications, the inspection plan input parameters are the key input parameters required by the decision tree. For example, but not limited to, the following steps S21 to S28 can be used to generate the aforementioned key input parameters.

[0050] S21. Based on the inspection type and priority of each inspection type of the equipment to be inspected, generate several inspection tasks and a priority for each inspection task. These inspection tasks include emergency inspection tasks, designated equipment inspection tasks, and regular inspection tasks, with the priorities decreasing sequentially from emergency to designated to regular. In this embodiment, the inspection tasks and their corresponding priorities are generated based on the equipment type and priority. Since the aforementioned inspection types include emergency inspection, regular inspection, and designated equipment inspection, inspection tasks corresponding to these three types are generated accordingly. Then, the priority of the inspection type is used as the priority of the corresponding inspection task. Therefore, the priority of an emergency inspection task is higher than the priority of a designated equipment inspection task, and the priority of a designated equipment inspection task is higher than the priority of a regular inspection task.

[0051] After obtaining the inspection task and its corresponding priority, the task priority can be generated, as shown in step S22 below.

[0052] S22. Based on several inspection tasks and the priority of each inspection task, generate task priorities; in this embodiment, task priorities can be represented as: emergency inspection task - designated equipment inspection task - regular inspection task; thus, after obtaining the task priorities, resource requirements can be constructed, and the process is shown in steps S23 to S26 below.

[0053] S23. Based on the inspection equipment labels of the inspection components corresponding to each independent point and each associated point in the associated task group, determine the inspection equipment type corresponding to each independent point and each associated point. In this embodiment, for any independent point, assuming that the inspection equipment label of its corresponding inspection component is label 1, then the inspection equipment type corresponding to the independent point is a drone. Similarly, if its corresponding inspection equipment label is 2, then its corresponding inspection equipment type is a quadruped robot. Similarly, the method for determining the inspection equipment type required for each associated point is also the same, and will not be repeated here.

[0054] Furthermore, for the same location, it is necessary to combine historical data to determine the type of inspection equipment, and the process is shown in step S24 below.

[0055] S24. Obtain historical inspection data for each identical point in the inspection point configuration and association results, and determine the inspection equipment type corresponding to each identical point based on the historical inspection data. In this embodiment, the historical inspection data for the same point may include, but is not limited to, the inspection equipment used within the historical preset inspection cycle. Thus, this embodiment counts the number of times the inspection equipment used in the historical inspection data is used, and then uses the type of the inspection equipment with the most usage as the inspection equipment type for the same point.

[0056] For example, if drones are used the most frequently in a series of historical inspections, then the type of inspection equipment at the same location is determined to be a drone; that is, at the same time, only a single drone or a single quadruped robot is used for inspection at the same location.

[0057] Similarly, when there are identical points in the associated task group, that is, there are associated points with inspection equipment labels 1 and 2, the method for determining the inspection equipment type can be found in the aforementioned step S24, and will not be repeated here.

[0058] After obtaining the inspection equipment types corresponding to independent locations, identical locations, and related locations, the team skills, i.e. the main personnel configuration, can be determined according to the inspection requirements, as shown in step S25 below.

[0059] S25. Based on the inspection requirements of each piece of equipment to be inspected, generate personnel configuration information for each independent location, each identical location, and each related location. In this embodiment, the process of generating personnel configuration information is illustrated using an independent location as an example. That is, the team skills in the inspection requirements of the equipment to be inspected to which the independent location belongs are used as the team skills (i.e., personnel configuration information) of the independent location. Of course, the method for determining the personnel configuration information of identical locations and related locations is also the same, and will not be repeated here.

[0060] After obtaining the personnel configuration information, the task resource requirements can be formed by combining the types of inspection equipment at each location, as shown in step S26 below.

[0061] S26. Utilize the inspection equipment types corresponding to each independent point, each identical point, and each related point, as well as the personnel configuration information for each independent point, each identical point, and each related point, to generate task resource requirements.

[0062] After generating the task resource requirements, the estimated time for the inspection can be estimated, as shown in step S27 below.

[0063] S27. Determine the inspection time for each identical point and generate the estimated inspection time using the inspection time for each identical point. In this embodiment, as explained above, the inspection requirements for any device to be inspected include the estimated time for each inspection part in each device to be inspected. Therefore, for identical points, the sum of the estimated times for all corresponding inspection parts can be taken as the inspection time for identical points. That is, as explained above, identical points with the same location will be merged. Then, for the merged identical points, the inspection time is the sum of the estimated times for all merged points. For individual identical points, the estimated time for their corresponding inspection parts is taken directly as the inspection time.

[0064] Furthermore, for independent points, the estimated time for the corresponding inspection location is directly taken, while for associated task groups, the inspection time for the associated task group is obtained by adding the sum of the times for all points in the group to the collaborative waiting time (the waiting time can be preset). In this way, the inspection time for independent points, identical points, and associated task groups can be used to form the estimated inspection time. Finally, the inspection plan input parameters can be formed by combining the aforementioned task priority, task resource requirements, and the associated task groups in the inspection point configuration and association results, as shown in step S28 below.

[0065] S28. The inspection plan input parameters are composed using the task priority, the task resource requirements, the estimated inspection time, and the associated task groups in the inspection point configuration and association results.

[0066] Thus, through the aforementioned steps S21 to S28, the key input parameters for the input decision tree can be generated. Then, based on these parameters, inspection plans with different task priorities can be generated, as shown in step S3 below.

[0067] S3. Input the inspection scheme input parameters into the decision tree model to output inspection schemes for the equipment to be inspected with different task priorities. The inspection scheme for any task priority includes the inspection equipment assigned to each inspection point of the equipment to be inspected corresponding to the task priority, and the execution path of the inspection equipment in the inspection scheme corresponding to the task priority. The inspection equipment at any inspection point is a quadruped robot or a drone.

[0068] In practical implementation, the decision tree model outputs different inspection task allocation schemes through multi-level judgments. For example, steps S31 and S32 can be used to generate the inspection schemes corresponding to the aforementioned three inspection tasks.

[0069] S31. Based on the first-level nodes in the decision tree model and according to the task priority, determine the current optimal task to be executed. In this embodiment, the first-level nodes in the decision tree model are used to determine the task priority. The determination condition is: emergency inspection task → designated equipment inspection task → regular inspection task, that is, the priority of the above three tasks decreases in sequence. Therefore, the execution starts from the task with the highest priority. That is, the current optimal task to be executed is the emergency inspection task. After the emergency inspection task is assigned, the current optimal task to be executed is updated to the designated equipment inspection task. Similarly, after the designated equipment inspection task is assigned, the current optimal task to be executed is updated to the regular inspection task (i.e., the routine inspection task).

[0070] Thus, after the task priority is determined, the second-level node can be triggered based on the current optimal task to generate the inspection plan for the current optimal task, as shown in step S32 below.

[0071] S32. Utilize the second-level nodes in the decision tree model, and generate an inspection plan for the equipment to be inspected corresponding to the current optimal task, based on the task resource requirements and the estimated inspection time, or based on the associated task group and the task resource requirements. After polling all inspection tasks in the task priority, inspection plans for the equipment to be inspected with different task priorities are obtained.

[0072] In practical implementation, when the current optimal task to be executed is either an emergency inspection task or a routine inspection task, resource requirements are determined to generate a corresponding inspection plan; when the current optimal task to be executed is a designated equipment inspection task, the associated task group is determined; the generation process for inspection plans for different tasks is as follows: S32a. If the current optimal task to be executed is an emergency inspection task or a regular inspection task, then each inspection point in the equipment to be inspected that is of the type of emergency inspection or regular inspection is selected as the target point. In this embodiment, it is assumed that there are equipment A, equipment B and equipment D to be inspected in the substation. Among them, the inspection type of equipment A and B is emergency inspection. Then, for the emergency inspection task, the inspection points on equipment A and equipment B are used as the target points. Of course, if the current optimal task to be executed is a regular inspection task, the process of determining the target point is the same as the above example, and will not be repeated here.

[0073] After obtaining the target location, the location can be distinguished, as shown in step S32b below.

[0074] S32b. Select independent and identical points from the target points; In this embodiment, for emergency inspection tasks and regular inspection tasks, the point allocation rule is: independent inspection points (indoor / outdoor) → identical inspection points (indoor / outdoor), that is, first allocate independent points, and then allocate identical points, as shown in the following steps S32c to S32f.

[0075] S32c. Based on the task resource requirements, determine the inspection equipment types corresponding to independent points in the target locations, and the inspection equipment types corresponding to identical points in the target locations. In this embodiment, as previously explained, the task resource requirements include the inspection equipment types corresponding to all independent points, identical points, and associated points in the substation. Therefore, when executing the steps, the inspection equipment types corresponding to independent points and identical points can be obtained by directly parsing the task resource requirements. Then, the inspection equipment allocated to each independent point and identical point in the target locations can be determined, as shown in step S32d below.

[0076] S32d. Based on the inspection area to which the independent points in the target location belong and the corresponding inspection equipment type, determine the inspection equipment for each independent point in the target location, and assign each independent point in the target location to its respective inspection equipment. The inspection equipment for any independent point is a quadruped robot or drone corresponding to the inspection area to which that independent point belongs. In this embodiment, it has been explained above that each inspection area corresponds to a drone and / or a quadruped robot. Therefore, after knowing the independent points in the target location and the inspection equipment type of the same point, the inspection equipment can be assigned according to the inspection area to which it belongs.

[0077] For example, suppose that the inspection area of ​​the equipment to be inspected at independent point G1 is area 1, and area 1 corresponds to UAV 01 and quadruped robot 0z1, and the inspection equipment type of independent point G1 is UAV; then the inspection equipment of independent point G1 is UAV 01. Similarly, suppose that the inspection area of ​​the equipment to be inspected at independent point G2 is area 1, and the inspection equipment type of independent point G2 is quadruped robot, then the inspection equipment of independent point G2 is quadruped robot 0z1. In this way, after allocating each independent point in the target points in the aforementioned manner, the same points can be allocated, and the process is shown in step S32e below.

[0078] S32e. After completing the allocation of independent points in the target location, the inspection equipment for each identical point in the target location is determined by using the inspection area to which each identical point belongs and the corresponding inspection equipment type. In this embodiment, the process of determining the inspection equipment for each identical point in the target location is the same as that for independent points, and will not be repeated here. After obtaining the inspection equipment corresponding to each identical point, the identical points can be allocated, and the process is shown in step S32f below.

[0079] S32f. Based on the estimated inspection time, determine the inspection time of each identical point in the target location, and allocate each identical point in the target location to its corresponding inspection equipment in order of inspection time from low to high. In specific implementation, as mentioned above, the estimated inspection time includes the inspection time of each identical point. Therefore, in this embodiment, the points are sorted in ascending order of inspection time to prioritize the allocation of identical points with shorter inspection times. Of course, the allocation process of identical points is the same as that of independent points in the target location. That is, assuming that identical point X1 belongs to the inspection area 1 and its corresponding inspection equipment type is a quadruped robot, then the inspection equipment allocated to identical point X1 is quadruped robot 0z1. In this embodiment, the allocation process of other identical points is also the same, and will not be repeated here.

[0080] Thus, after the allocation of the same points in the target locations is completed, the inspection equipment used for the current optimal task execution can be determined. Then, the execution path of each inspection equipment is generated, as shown in step S32g below.

[0081] S32g. Generate the execution path of the target device based on the target points assigned to the target device, wherein the target device is all the inspection devices corresponding to the target points.

[0082] In this embodiment, assuming that the inspection devices assigned to each independent point and the corresponding identical point in the target area are UAV 01 and quadruped robot 0z1, then the target devices are UAV 01 and quadruped robot 0z1. Specifically, if the points in UAV 01 include: independent point G1, independent point G3, independent point G5, identical point X2, identical point X3, etc., and the points in quadruped robot 0z1 include: independent point G2, independent point G4, independent point G6, identical point X1, identical point X4, etc., then UAV 01 needs to generate an execution path containing independent points G1, G3, G5, identical points X2, identical points X3, etc., while quadruped robot 0z1 needs to generate an execution path containing independent points G2, G4, G6, identical points X1, identical points X4. Specifically, but not limited to, using... The algorithm is used to plan the path for the target device and generate the execution path for the target device; of course... The algorithm is a commonly used algorithm for path planning, and its principle will not be elaborated here.

[0083] Furthermore, the points can be sorted according to the inspection time of each point to generate an execution path. For example, for UAV 01, the independent points assigned to UAV 01 can be sorted in order of inspection time from low to high to obtain the first sequence; then, the same points assigned to UAV 01 can be sorted in the same way to obtain the second sequence; finally, the first sequence and the second sequence can be combined to obtain the execution path.

[0084] After obtaining the execution path, the inspection plans for emergency inspection tasks and routine inspection tasks can be generated by combining the inspection equipment allocated to each independent point and each identical point in the target location. The process is shown in step S32h below.

[0085] S32h. Using the execution path of the target device, as well as the inspection equipment and personnel configuration information corresponding to each independent point and each identical point in the target location, generate the inspection plan for the device to be inspected corresponding to the current optimal execution task; in this embodiment, the personnel configuration information is obtained according to the task resource requirements.

[0086] Thus, through the aforementioned steps S32a to S32h, the corresponding inspection plan can be generated when the current optimal task to be executed is either an emergency inspection task or a regular inspection task.

[0087] If the current optimal task to be executed is the inspection task of a specified device, then the associated task group is determined to generate the corresponding inspection plan. The process is as follows: S32i. If the current optimal task to be executed is the designated equipment inspection task, then from the associated task group, each inspection point and the associated point corresponding to each inspection point in the equipment to be inspected with the inspection type of designated equipment inspection are selected, so as to form a target associated group using the selected points; in this embodiment, the point extraction process of this step is the same as the aforementioned step S32a, and will not be repeated here; after obtaining the target key group, the points can be distinguished, and the process is shown in step S32j below.

[0088] S32j. Divide the associated points in the target association group into indoor and outdoor key equipment points and indoor and outdoor associated equipment points; in this embodiment, indoor and outdoor key equipment points are the inspection points on the body of the equipment to be inspected for the specified equipment inspection, and indoor and outdoor associated equipment points are the associated points in the target association group excluding indoor and outdoor key equipment points.

[0089] Thus, after the division of associated points in the target associated group is completed, the associated points can be allocated according to the rule of first allocating indoor and outdoor key equipment points, and then allocating indoor and outdoor associated equipment points. The process is shown in steps S32k and S32l below.

[0090] S32k. Based on the task resource requirements, determine the inspection equipment corresponding to the key indoor and outdoor equipment locations, and assign the key indoor and outdoor equipment locations to the corresponding inspection equipment. In this embodiment, the process of determining the inspection equipment corresponding to the key indoor and outdoor equipment locations can be referred to the aforementioned steps S32e to S32f. That is, if the key indoor and outdoor equipment locations are independent locations, the inspection equipment will be assigned according to steps S32e and S32d. If the key indoor and outdoor equipment locations are the same location, the inspection equipment will be assigned according to steps S32e and S32f. The principle will not be elaborated here.

[0091] After the allocation of key indoor and outdoor equipment locations is completed, the allocation of related indoor and outdoor equipment locations can be carried out, as shown in step S32l below.

[0092] S32l. After completing the allocation of key indoor and outdoor equipment locations, based on the task resource requirements, determine the inspection equipment corresponding to the indoor and outdoor related equipment locations, and allocate the indoor and outdoor related equipment locations to the corresponding inspection equipment; in this embodiment, the process of determining the inspection equipment corresponding to the indoor and outdoor related equipment locations can be referred to the aforementioned step S32k, and will not be repeated here.

[0093] Thus, after the allocation of each associated point in the target association group is completed, the execution path of the inspection equipment corresponding to the target association group can be generated, as shown in step S32m below.

[0094] S32m. Each inspection device corresponding to the target association group is designated as a device, and an execution path for each designated device is generated based on the device location assigned to each designated device.

[0095] S32n. Utilizing the execution paths of each designated device, as well as the inspection equipment and personnel configuration information corresponding to key indoor and outdoor equipment locations and related indoor and outdoor equipment locations, an inspection plan for the equipment to be inspected corresponding to the current optimal execution task is generated; in this embodiment, an inspection plan can also be adopted. The algorithm is used for path planning for a specified device, which will not be elaborated here.

[0096] Thus, through the aforementioned steps S32 and their sub-steps, the allocation of inspection equipment at inspection points with different task priorities can be completed, thereby automatically allocating points for drones and quadruped robots within the substation, and generating inspection plans corresponding to emergency inspections, designated equipment inspections, and routine inspections; thus, an inspection task can support collaborative inspections using drones and robot dogs, and based on this, deep collaboration between drones and robot dogs can be achieved.

[0097] After obtaining the inspection plans for the equipment to be inspected with different task priorities, the plans can be issued, and the process is shown in step S4 below.

[0098] S4. Distribute each inspection plan to its corresponding inspection equipment so that each inspection equipment can perform an inspection of the substation according to the received inspection plan and upload the inspection data. In this embodiment, assuming that an emergency inspection task requires the use of UAV 01 and quadruped robot 0z1 to inspect the equipment to be inspected, then the execution paths of UAV 01 and quadruped robot 0z1 are distributed respectively, so that UAV 01 and quadruped robot 0z1 can perform inspections according to their respective execution paths and upload their respective inspection data.

[0099] Thus, through the aforementioned steps, a substation inspection can be completed. In addition, this embodiment also includes a data fusion process, which means that the substation may be inspected multiple times within a specified time. Therefore, this embodiment also includes a data fusion-based inspection result generation mechanism, the process of which is shown in steps S5 and S6 below.

[0100] S5. Receive multiple inspection data from the substation within a specified time period, and filter out the inspection data of the points to be inspected from the multiple inspection data. The points to be inspected are the inspection points jointly inspected by the drone and the quadruped robot in the multiple inspection data.

[0101] In this embodiment, although an inspection point can only be inspected by a drone or a robot in a single task, during multiple inspections, an inspection point may be inspected by different inspection devices in different tasks. For example, during the first inspection, the independent point G1 may be inspected by drone 01, while during the second inspection after a preset interval, the independent point G1 may be inspected by quadruped robot 0z1. Therefore, the independent point is jointly inspected by drone and quadruped robot within a specified time period. Based on this, this embodiment uses the inspection points jointly inspected by drone and quadruped robot within a specified time period (e.g., 24 hours) as the points to be inspected. Then, the corresponding inspection data is extracted for data fusion.

[0102] The data fusion process is shown in step S6 below.

[0103] S6. Perform data fusion processing on the inspection data of the points to be inspected to obtain fused data, and generate the inspection results of the points to be inspected based on each fused data. In this embodiment, the inspection data of the points to be inspected is image data, and the image data includes a first image uploaded by a drone and a second image uploaded by a quadruped robot. For example, for an independent point G1, its first image was uploaded by drone 01 during the first inspection (collection time: 10:00 AM on November 23, 2025), while the second image was uploaded by quadruped robot 0z1 during the second inspection (collection time: 11:00 AM on November 23, 2025). Then, the EXIF ​​information of the first and second images will embed various corresponding image acquisition times. Specifically, the UTC time of each image acquisition is embedded so that time constraints can be performed based on the UTC time later.

[0104] Optionally, the data fusion and inspection result generation process may be, but is not limited to, the steps S61 to S66 below.

[0105] S61. Perform image alignment processing on the first image and the second image to obtain a first aligned image and a second aligned image. In this embodiment, for example, but not limited to, the SIFT feature matching algorithm can be used first to match image feature points, and then the homography matrix can be calculated based on the matched feature point pairs. Finally, the second image is aligned with the first image based on the homography matrix to obtain the first aligned image and the second aligned image.

[0106] Optionally, the specific implementation process is as follows: (1) Key point detection: Construct a Gaussian difference pyramid: Perform Gaussian blurring on the image at different scales and calculate the difference between adjacent scale images; (2) Extreme point localization: Detect three-dimensional (spatial + scale) extreme points in the Gaussian pyramid, and remove low contrast points and edge response points through Taylor expansion; (3) Descriptor generation: Take a 16×16 neighborhood centered on the key point, divide it into 4×4 sub-regions, generate an 8-directional gradient histogram for each sub-region, and construct a 128-dimensional descriptor; (4) Matching and optimization: Brute-force matching (Brute-For ce): Calculate the Euclidean distance between the second image and the RGB descriptor, and take the matching pairs with a nearest neighbor / second nearest neighbor ratio of <0.8; (5) RANSAC purification: Estimate the homography matrix H through the random sampling consistency algorithm and remove more than 95% of the mismatch points; (6) Image transformation: Based on the homography matrix, perform perspective transformation on the second image to align the second image to the first image and obtain the second aligned image, while the first image itself is used as the first aligned image; Of course, the aforementioned SIFT feature matching process is a commonly used algorithm for image alignment, and its principle will not be elaborated here.

[0107] After image alignment is completed, time alignment can be performed, as shown in step S62 below.

[0108] S62. Perform time alignment processing on the first aligned image and the second aligned image, and determine whether the first aligned image and the second aligned image meet the time constraint; in specific applications, for example, but not limited to, first extracting the UTC time from the EXIF ​​information of the first image and extracting the UTC time from the EXIF ​​information of the second image; then, calculating the time difference between the UTC time of the first image and the UTC time of the second image; then, determining whether the time difference is less than or equal to a preset duration (the preset duration here is the aforementioned specified duration, i.e., 24 hours); if yes, then determine that the first aligned image and the second aligned image meet the time constraint; otherwise, output an alarm message that the inspection data does not meet the time constraint.

[0109] Once the time constraint is met, feature splicing, also known as multimodal fusion, can be performed, as shown in step S63 below.

[0110] S63. If so, feature concatenation is performed on the first aligned image and the second aligned image to obtain a first feature image, wherein the first feature image serves as the fused data. In specific implementation, for example, but not limited to, feature extraction can be performed on the first aligned image to obtain a first feature sub-image, and feature extraction can be performed on the second aligned image to obtain a second feature sub-image; then, feature concatenation is performed on the first feature sub-image and the second feature sub-image to obtain the first feature image. Specifically, feature extraction can be performed using a YOLOv5s neural network or a ResNet-50 neural network, but not limited to.

[0111] After data fusion is completed, defect identification of the points to be inspected can be performed, as shown in steps S64 to S66 below.

[0112] S64. Using the CBAM attention network, feature extraction processing is performed on the first feature image to obtain the second feature image. In this embodiment, CBAM mainly has two modules: a channel attention module and a spatial attention module. The channel attention module learns the importance of each channel by calculating the average and maximum values ​​of each channel, and weights the channels according to these importance values, so that the network can better utilize the correlation between different channels. The spatial attention module learns the importance of each spatial location by performing average pooling and max pooling along the channel dimension of the feature map, and weights the spatial locations according to these importance values, so that the network can better utilize the correlation between different spatial locations. Finally, by connecting the channel attention module and the spatial attention module, CBAM can perform comprehensive attention weighting processing on the input data, thereby improving the network's expressive power and generalization ability. That is, the CBAM attention network is equivalent to re-extracting features, thereby highlighting important features.

[0113] After obtaining the second feature image, the defect identification of the point to be inspected can be performed based on it, as shown in steps S65 and S66 below.

[0114] S65. Based on the second feature image, generate a temporal feature sequence; in specific implementation, for example, but not limited to, using a sliding window to perform sliding sampling on the second feature image to obtain multiple sampled features (each sliding window corresponds to a time step); then, sort the multiple sampled features in order according to the time steps of the corresponding sliding windows, so as to generate the temporal feature sequence after sorting in order.

[0115] After obtaining the time-series feature sequence, a pre-trained defect detection model can be used to detect defects at the points to be inspected, as shown in step S66 below.

[0116] S66. Input the time-series feature sequence into the defect detection model to obtain the defect detection results of the points to be inspected, and use the defect detection results as the inspection results. In this embodiment, the defect detection model can be, but is not limited to, a Bi-LSTM network, which has two layers of bidirectional LSTM, each layer with 256 hidden units, and each LSTM layer is followed by a Dropout layer with a ratio of 0.3 to prevent overfitting. The output layer is a fully connected layer (512→5) and is activated by the Softmax function. Therefore, the model can output five probability distributions of defect severity (corresponding to the five defect severity levels defined by the power grid). By using the aforementioned steps S6 and their sub-steps, data fusion can be completed, and corresponding inspection results can be generated to achieve accurate identification of faults at different inspection points.

[0117] Therefore, through the multi-device collaborative substation inspection method described in detail in steps S1 to S6 above, this invention, through deep collaboration and data fusion between UAVs and quadruped robots, solves the problem that a single inspection device cannot effectively cover the substation. Furthermore, it enables automatic allocation of inspection points for both quadruped robots and UAVs in the same task, achieving collaborative allocation and automatically generating a collaborative inspection plan. Simultaneously, data fusion avoids the data fragmentation problem inherent in traditional technologies, thereby improving fault detection accuracy. Thus, this invention improves the effectiveness, accuracy, safety, and intelligence of power inspection, making it highly suitable for large-scale application and promotion.

[0118] like Figure 2 As shown, the second aspect of this embodiment provides a hardware system for implementing the multi-device collaborative substation inspection method described in the first aspect of the embodiment, including: a substation host, at least one quadruped robot, and at least one drone; The substation host is used to obtain the equipment ledger information of the substation, and based on the equipment ledger information, to configure and associate the inspection points of each piece of equipment to be inspected in the substation, and obtain the inspection point configuration and association results.

[0119] The substation main unit is used to generate inspection plan input parameters based on equipment ledger information and inspection point configuration and association results; The substation host is used to input the inspection plan input parameters into the decision tree model to output inspection plans for the equipment to be inspected with different task priorities. The inspection plan for any task priority includes the inspection equipment assigned to each inspection point of the equipment to be inspected corresponding to that task priority, as well as the execution path of the inspection equipment in the inspection plan corresponding to that task priority. The inspection equipment at any inspection point is a quadruped robot or a drone.

[0120] The substation main unit is used to distribute various inspection plans to their respective inspection equipment.

[0121] Each inspection device is used to conduct a single inspection of the substation according to the received inspection plan and upload the inspection data.

[0122] The substation host is used to receive multiple inspection data from the substation within a specified time period, and to filter out the inspection data of the points to be inspected from the multiple inspection data. The points to be inspected are the inspection points jointly inspected by drones and quadruped robots in the multiple inspection data.

[0123] The substation main unit is also used to perform data fusion processing on the inspection data of the points to be inspected, obtain fused data, and generate the inspection results of the points to be inspected based on the fused data.

[0124] The working process, working details and technical effects of the system provided in this embodiment can be found in the first aspect of the embodiment, and will not be repeated here.

[0125] like Figure 3 As shown, the third aspect of this embodiment provides a substation inspection device for multi-device collaboration. Taking the device as an electronic device as an example, it includes: a memory, a processor, and a transceiver that are connected in sequence. The memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the substation inspection method for multi-device collaboration as described in the first aspect of the embodiment.

[0126] For specific examples, the memory may include, but is not limited to, random access memory (RAM), read-only memory (ROM), flash memory, first-in-first-out (FIFO) memory, and / or first-in-last-out (FILO) memory, etc.; specifically, the processor may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor may be implemented using at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), PLA (Programmable Logic Array). The processor may also include a main processor and a coprocessor. The main processor, also known as the CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state.

[0127] In some embodiments, the processor may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. For example, the processor may not be limited to microprocessors of the STM32F105 series, reduced instruction set computer (RISC) microprocessors, x86 architecture processors, or processors with integrated neural network processing units (NPUs). The transceiver may be, but is not limited to, a Wi-Fi transceiver, a Bluetooth transceiver, a General Packet Radio Service (GPRS) transceiver, a ZigBee (a low-power LAN protocol based on the IEEE 802.15.4 standard) transceiver, a 3G transceiver, a 4G transceiver, and / or a 5G transceiver. Furthermore, the device may also include, but is not limited to, a power module, a display screen, and other necessary components.

[0128] The working process, working details and technical effects of the electronic device provided in this embodiment can be found in the first aspect of the embodiment, and will not be repeated here.

[0129] The fourth aspect of this embodiment provides a storage medium for storing instructions containing the multi-device collaborative substation inspection method described in the first aspect of the embodiment. That is, the storage medium stores instructions, and when the instructions are run on a computer, the multi-device collaborative substation inspection method described in the first aspect of the embodiment is executed.

[0130] The storage medium refers to a carrier for storing data, which may include, but is not limited to, floppy disks, optical disks, hard disks, flash memory, USB flash drives, and / or memory sticks. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0131] The working process, working details, and technical effects of the storage medium provided in this embodiment can be found in the first aspect of the embodiment, and will not be repeated here.

[0132] The fifth aspect of this embodiment provides a computer program product containing instructions that, when executed on a computer, cause the computer to perform the multi-device collaborative substation inspection method as described in the first aspect of the embodiment, wherein the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device.

[0133] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for substation inspection in multi-device cooperation, characterized in that, The application is applied to a power substation inspection system cooperating with multiple devices, wherein the power substation inspection system comprises at least one quadruped robot and at least one unmanned aerial vehicle, and the method comprises: Obtaining device account information of a power substation, and based on the device account information, performing inspection point position configuration and association processing on each to-be-inspected device in the power substation to obtain an inspection point position configuration and association result; Generating inspection scheme input parameters according to the device account information and the inspection point position configuration and association result; Inputting the inspection scheme input parameters into a decision tree model to output an inspection scheme of a to-be-inspected device with different task priorities, wherein any inspection scheme with a task priority comprises an inspection device assigned to each inspection point position of the to-be-inspected device corresponding to the task priority, and an execution path of the inspection device in the inspection scheme corresponding to the task priority, and the inspection device of any inspection point position is a quadruped robot or an unmanned aerial vehicle; Downlinking each inspection scheme to a corresponding inspection device to enable each inspection device to perform a power substation inspection according to the received inspection scheme and upload inspection data; Receiving multiple inspection data of the power substation within a specified time length, and from the multiple inspection data, screening out inspection data of a to-be-inspected point position, wherein the to-be-inspected point position is an inspection point position commonly inspected by the unmanned aerial vehicle and the quadruped robot in the multiple inspection data; Performing data fusion processing on the inspection data of the to-be-inspected point position to obtain fused data, and generating an inspection result of the to-be-inspected point position according to each fused data.

2. The method of claim 1, wherein, Based on the device account information, performing inspection point position configuration and association processing on each to-be-inspected device in the power substation to obtain an inspection point position configuration and association result, comprising: According to the device account information, determining all inspection components on each to-be-inspected device, and taking each inspection component as an initial inspection point position; Based on the device account information, performing point position attribute labeling on each initial inspection point position to obtain multiple inspection point positions; According to the point position attributes of the multiple inspection point positions, screening out inspection point positions that are only inspected by the unmanned aerial vehicle or the quadruped robot from the multiple inspection point positions as independent point positions, and screening out inspection point positions that are commonly inspected by the unmanned aerial vehicle and the quadruped robot in the same inspection area as the same point positions; Based on the device account information, extracting inspection point positions with an association relationship from the multiple inspection point positions, and taking the extracted inspection point positions as associated point positions to generate an associated task group using the associated point positions; Generating the inspection point position configuration and association result using the independent point positions, the same point positions, and the associated task group.

3. The method of claim 2, wherein, The device account information comprises: device type and priority, device basic attributes, and inspection requirements of each to-be-inspected device, wherein the device type and priority comprise an inspection type of each to-be-inspected device and a priority of each inspection type, the device basic attributes comprise an inspection device label corresponding to different inspection components on each to-be-inspected device, and the inspection type comprises a regular inspection, an emergency inspection, and a designated device inspection; Wherein, generating inspection scheme input parameters according to the device account information and the inspection point position configuration and association result comprises: According to the inspection types of each to-be-inspected device and the priorities of each inspection type, a plurality of inspection tasks and a priority of each inspection task are generated, wherein the plurality of inspection tasks include an emergency inspection task, a designated device inspection task, and a regular inspection task, and the priorities of the emergency inspection task, the designated device inspection task, and the regular inspection task decrease in turn; Based on the plurality of inspection tasks and the priority of each inspection task, a task priority is generated; According to the inspection device tags of the inspection components corresponding to each independent point and each associated point in the associated task group, the inspection device types corresponding to each independent point and each associated point are determined; The historical inspection data of each same point in the inspection point configuration and association result is obtained, and the inspection device type corresponding to each same point is determined according to the historical inspection data of each same point; According to the inspection requirements of each to-be-inspected device, personnel configuration information of each independent point, each same point, and each associated point is generated; The inspection device types corresponding to each independent point, each same point, and each associated point, and the personnel configuration information of each independent point, each same point, and each associated point are used to generate a task resource requirement; The inspection time consumption of each same point is determined, and the inspection estimated time consumption is generated using the inspection time consumption of each same point; The task priority, the task resource requirement, the inspection estimated time consumption, and the associated task group in the inspection point configuration and association result are used to form the inspection scheme input parameter.

4. The method of claim 3, wherein, The inspection scheme input parameter is input into a decision tree model to output an inspection scheme of a to-be-inspected device with different task priorities, including: Based on a first layer node in the decision tree model and according to the task priority, a currently optimal execution task is determined; The second layer node in the decision tree model is used to generate an inspection scheme of a to-be-inspected device corresponding to the currently optimal execution task according to the task resource requirement and the inspection estimated time consumption, or according to the associated task group and the task resource requirement, so as to obtain the inspection scheme of the to-be-inspected device with different task priorities after all the inspection tasks in the task priority are polled.

5. The method of claim 4, wherein, Each inspection area of a substation corresponds to a quadruped robot and / or a drone, wherein the second layer node in the decision tree model is used to generate an inspection scheme of a to-be-inspected device corresponding to the currently optimal execution task according to the task resource requirement and the inspection estimated time consumption, or according to the associated task group and the task resource requirement, including: If the currently optimal execution task is an emergency inspection task or a regular inspection task, each inspection point in the to-be-inspected device with an emergency inspection type or a regular inspection type is selected as a target point; The independent points and the same points are selected from the target points; According to the task resource requirement, the inspection device types corresponding to the independent points in the target points and the inspection device types corresponding to the same points in the target points are determined; determine the inspection device of each independent point in the target point based on the inspection area to which the independent point belongs and the corresponding inspection device type in the target point, and distribute each independent point in the target point to the corresponding inspection device, wherein the inspection device of any independent point is a quadruped robot or a drone corresponding to the inspection area to which the independent point belongs; after completing the distribution of the independent points in the target point, determine the inspection device of each identical point in the target point based on the inspection area to which each identical point belongs and the corresponding inspection device type in the target point; based on the inspection estimated time, determine the inspection time of each identical point in the target point, and distribute each identical point in the target point to the corresponding inspection device in order of low to high inspection time; generate the execution path of the target device according to the target point allocated to the target device, wherein the target device is all inspection devices corresponding to the target point; generate the inspection scheme of the target device corresponding to the current optimal execution task by using the execution path of the target device, and the inspection device and personnel configuration information corresponding to each independent point and each identical point in the target point.

6. The method of claim 4, wherein, generate the inspection scheme of the target device corresponding to the current optimal execution task by using the second layer node in the decision tree model and according to the task resource demand and the inspection estimated time, or according to the associated task group and the task resource demand, including: if the current optimal execution task is a specified device inspection task, filter each inspection point and the corresponding associated point in the target device from the associated task group, wherein the inspection type of the target device is specified device inspection, to form a target associated group by using the filtered points; divide the associated points in the target associated group into indoor and outdoor key device points and indoor and outdoor associated device points; determine the inspection device corresponding to the indoor and outdoor key device points according to the task resource demand, and distribute the indoor and outdoor key device points to the corresponding inspection device; after completing the distribution of the indoor and outdoor key device points, determine the inspection device corresponding to the indoor and outdoor associated device points based on the task resource demand, and distribute the indoor and outdoor associated device points to the corresponding inspection device; use each specified device as the inspection device, and generate the execution path of each specified device according to the device points allocated to each specified device; generate the inspection scheme of the target device corresponding to the current optimal execution task by using the execution path of each specified device, and the inspection device and personnel configuration information corresponding to the indoor and outdoor key device points and the indoor and outdoor associated device points.

7. The method of claim 1, wherein, The inspection data of the to-be-inspected point is image data, and the image data includes first images uploaded by a drone and second images uploaded by a quadruped robot. The inspection data of the to-be-inspected point is processed by data fusion to obtain fused data, and the inspection result of the to-be-inspected point is generated according to each fused data, including: perform image alignment processing on the first images and the second images to obtain first aligned images and second aligned images; The first aligned image and the second aligned image are subjected to time alignment processing, and it is determined whether the first aligned image and the second aligned image satisfy a time constraint; If yes, the first aligned image and the second aligned image are subjected to feature splicing to obtain a first feature image, wherein the first feature image is taken as the fused data; The first feature image is subjected to feature extraction processing by using a CBAM attention network to obtain a second feature image; Based on the second feature image, a time sequence feature sequence is generated; The time sequence feature sequence is input into a defect detection model to obtain a defect detection result of the to-be-inspected point, and the defect detection result is taken as the inspection result.

8. The method of claim 7, wherein, The EXIF information of the first image and the EXIF information of the second image both embed the UTC time when the respective images are collected; The first aligned image and the second aligned image are subjected to time alignment processing, and it is determined whether the first aligned image and the second aligned image satisfy a time constraint, including: The UTC time in the EXIF information of the first image is extracted, and the UTC time in the EXIF information of the second image is extracted; The time difference between the UTC time of the first image and the UTC time of the second image is calculated; It is determined whether the time difference is less than or equal to a preset time length; If yes, it is determined that the first aligned image and the second aligned image satisfy the time constraint.

9. The method of claim 7, wherein, The first aligned image and the second aligned image are subjected to feature splicing to obtain a first feature image, including: The first aligned image is subjected to feature extraction to obtain a first feature sub-image, and the second aligned image is subjected to feature extraction to obtain a second feature sub-image; The first feature sub-image and the second feature sub-image are subjected to feature splicing to obtain the first feature image; Correspondingly, based on the second feature image, a time sequence feature sequence is generated, including: The second feature image is subjected to sliding sampling by using a sliding window to obtain a plurality of sampling features, wherein each sliding window corresponds to a time step; The plurality of sampling features are sequentially sorted according to the time steps of the corresponding sliding windows to generate the time sequence feature sequence after the sequential sorting.

10. A multi-device cooperative substation inspection system, comprising: It includes: A transformer station host and at least one quadruped robot and at least one unmanned aerial vehicle; The transformer station host is configured to acquire device account information of the transformer station, and perform inspection point configuration and association processing on each to-be-inspected device in the transformer station based on the device account information to obtain an inspection point configuration and association result; The transformer station host is configured to generate an inspection scheme input parameter according to the device account information and the inspection point configuration and association result; The transformer station host is configured to input the inspection scheme input parameter into a decision tree model to output an inspection scheme of the to-be-inspected device with different task priorities, wherein each inspection point in any task priority corresponds to a to-be-inspected device in the inspection scheme of the task priority, and the execution path of the to-be-inspected device in the inspection scheme of the task priority, and the to-be-inspected device of any inspection point is a quadruped robot or an unmanned aerial vehicle; The transformer station host is configured to distribute each inspection scheme to a corresponding inspection device. Each inspection device is configured to perform a primary inspection of the substation according to a received inspection scheme and upload inspection data; The substation host is configured to receive multiple inspection data of the substation within a specified time length, and filter out inspection data of a to-be-inspected point from the multiple inspection data, wherein the to-be-inspected point is a common inspection point of the unmanned aerial vehicle and the quadruped robot in the multiple inspection data. The substation host is further configured to perform data fusion processing on the inspection data of the to-be-inspected point to obtain fused data, and generate an inspection result of the to-be-inspected point according to each fused data.

Images