Inspection data processing method and system of smart substation integrated monitoring system

Abstract

The application discloses a kind of intelligent substation comprehensive monitoring system's tour data processing method and system, it is related to intelligent substation image tour and monitoring data processing technical field.For the problems that existing tour system needs artificial one by one to set image acquisition device preset position, artificial binding tour content and identification rule, and fixed tour strategy is difficult to adapt to equipment state change, the present application obtains the scanning image data obtained by scanning the substation tour area by image acquisition device, determines the corresponding equipment identification, and generates the tour task information containing preset position configuration and identification rule;Determine the execution priority according to equipment state information, generate tour result;Further, the effectiveness of the tour result is verified in combination with online monitoring data, and secondary review shooting is triggered when the credibility is insufficient.Thereby it can reduce the artificial participation in tour task configuration process, improve the adaptability of preset position and field equipment arrangement.

Description

Technical Field

[0001] This application relates to the field of image inspection and data processing technology for intelligent substations, and in particular to a method and system for processing inspection data of an integrated monitoring system for intelligent substations. Background Technology

[0002] During the operation and maintenance of smart substations, image acquisition devices with rotation and zoom capabilities are typically used to inspect circuit breakers, disconnectors, meters, transformers, surge arresters, insulators, and other equipment components. To ensure that the image acquisition devices can capture inspection images that meet recognition requirements at predetermined times, it is usually necessary to configure a preset position for each piece of equipment to be inspected, and associate this preset position with the corresponding equipment identifier, inspection content, and recognition rules.

[0003] When existing inspection systems are deployed at new substations or the location of image acquisition devices changes, technicians typically need to manually adjust the shooting direction and imaging range of each device. After observing the target equipment component, a preset position is saved, and the correspondence between the preset position and equipment identification, inspection content, and recognition rules is manually established. Due to differences in equipment layout, orientation, obstruction relationships, and image acquisition device installation locations at different substations, existing inspection task configurations are difficult to reuse directly, leading to repeated configuration after new substation deployments or equipment adjustments.

[0004] Furthermore, once configured, inspection tasks are typically executed in a fixed sequence and at a fixed frequency. It is difficult to adjust the execution schedule based on changes in equipment status reflected by online monitoring data, equipment alarm information, or historical defect information, resulting in a low degree of matching between the execution method of the inspection tasks and the actual state of the equipment. Summary of the Invention

[0005] This application provides a method and system for processing inspection data in an intelligent substation integrated monitoring system. This method reduces the reliance on manually setting preset positions and manually binding identification rules during the configuration of intelligent substation inspection services, enabling the execution of inspection tasks to match the actual status of the equipment, and forming an automatic review closed loop when the reliability of the inspection results is insufficient.

[0006] Firstly, this application provides a method for processing inspection data in an integrated monitoring system for intelligent substations. This method is applied to inspection management equipment. The inspection management equipment can be a substation-side edge computing device, a centralized control server, or a processing device composed of both.

[0007] The inspection and management equipment acquires scanned image data and substation equipment information obtained by the image acquisition device scanning the substation inspection area; it determines the equipment components, their categories, and their regions within the scanned image data based on the scanned image data, and then determines the corresponding equipment identifier based on the category, region information, and substation equipment information; based on the region information and category, it generates inspection task information corresponding to the equipment identifier, which includes preset position configurations of the image acquisition device and recognition rules for the equipment components; it acquires the equipment status information corresponding to the equipment identifier, and determines the execution strategy for the inspection task information based on the equipment status information; according to the execution strategy, it controls the image acquisition device to acquire inspection images according to the preset position configuration, and processes the inspection images using the recognition rules to obtain the inspection results corresponding to the equipment identifier.

[0008] Through the above processing, the equipment components in the scanned image can be associated with the equipment identifiers in the substation equipment information. The equipment identifiers, preset position configurations, and identification rules can form executable inspection task information. The equipment status information can also further influence the execution strategy of the inspection task information, thereby enabling the inspection task to automatically adapt to the on-site equipment layout and operating status.

[0009] In one possible implementation, the scanned image data includes multiple scanned images acquired by the image acquisition device at multiple horizontal and multiple vertical scanning positions. The inspection and management equipment performs image registration and image stitching on adjacent scanned images to obtain panoramic image data; it then performs equipment component identification on the panoramic image data to obtain the equipment component category, target image region, and identification confidence level corresponding to the identified target; based on the equipment component detection confidence threshold, it determines the equipment component, and based on the target image region, it determines the position and region size of the equipment component.

[0010] In one possible implementation, the substation equipment information includes a substation equipment ledger, which records equipment identification, equipment category, and equipment physical location. The inspection and management equipment spatially aligns panoramic image data with the substation equipment ledger, determines candidate physical locations corresponding to equipment components based on location mapping relationships, and identifies the equipment based on category matching and location matching.

[0011] In one possible implementation, the preset position configuration includes horizontal rotation parameters, vertical rotation parameters, and zoom parameters. The inspection and management equipment determines the preset position configuration based on the parameter mapping relationship between image position and rotation / zoom parameters, the regional center position of equipment components, regional size ratio, and target imaging requirements, and corrects the preset position configuration through image verification.

[0012] In one possible implementation, the inspection management equipment establishes a correspondence between equipment component categories, inspection content, and identification rules. The inspection management equipment determines the inspection content and identification rules based on the equipment component category, and associates the equipment identifier, preset position configuration, inspection content, and identification rules to generate inspection task information.

[0013] In one possible implementation, equipment status information includes at least one of online monitoring data, equipment alarm information, and historical defect information. The inspection and management system determines the extent of monitoring data exceeding limits, alarm event levels, and historical defect levels. The acquired data is standardized, and the abnormal status level is determined. For data types not included in the standard data set, a default status record is established, and this data type is excluded from the comprehensive processing of valid status quantities. This ensures that the abnormal status level can still be determined even when only one or two of the aforementioned data are acquired.

[0014] In one possible implementation, the execution strategy includes at least one of execution priority, execution frequency, image acquisition parameters, and processing parameters for recognition rules. When the execution strategy includes execution priority, the higher the abnormal state level, the higher the execution priority of the patrol task information; when the execution strategy includes execution frequency, the higher the abnormal state level, the higher the execution frequency of the patrol task information; the execution strategy can also increase image resolution, increase the number of consecutive acquisition frames, or select more stringent result confirmation conditions based on the abnormal state level.

[0015] In one possible implementation, after obtaining the inspection results, the inspection management equipment acquires the online monitoring data corresponding to the equipment identifier, and determines whether there is a corresponding target monitoring item in the online monitoring data based on the result type of the inspection results; if there is a target monitoring item, the corresponding matching rule is selected to determine the matching degree corresponding to the inspection results; if there is no target monitoring item, a neutral matching degree is used to indicate that the online monitoring data neither supports nor denies the inspection results; the inspection management equipment verifies the validity of the inspection results based on the matching degree.

[0016] In one possible implementation, the inspection management equipment determines the overall confidence level based on the identification confidence level and matching degree corresponding to the inspection results; when the overall confidence level does not reach the preset confidence threshold, a secondary review shooting instruction is generated based on the reason for failing the validity verification, and the image acquisition device is controlled to adjust the shooting direction, zoom level, exposure parameters, or continuously acquire the number of frames and re-acquire the inspection image.

[0017] Secondly, this application provides an intelligent substation integrated monitoring system, including a data acquisition module, an equipment association module, a task generation module, a strategy determination module, and an inspection processing module. These modules cooperate to implement the method described in the first aspect.

[0018] Compared with existing technologies, this application determines equipment components, their categories, and area information based on scanned image data, and combines this with substation equipment information to determine equipment identifiers. This allows for the automatic generation of preset position configurations and identification rules corresponding to equipment identifiers based on the substation's on-site equipment layout. This reduces the configuration work required for setting preset positions one by one and manually binding equipment identifiers and identification rules, and improves the adaptability of inspection task information to different on-site equipment layouts. Furthermore, by determining the execution strategy for inspection task information based on equipment status information, the execution schedule of inspection tasks can be adjusted according to changes in the actual equipment status, improving the targeting of inspection resources to equipment in abnormal condition. Attached Figure Description

[0019] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an undue limitation of this application.

[0020] Figure 1 This application provides a schematic diagram of an application scenario for an integrated monitoring system for intelligent substations.

[0021] Figure 2 A flowchart illustrating an inspection data processing method provided in this application embodiment;

[0022] Figure 3 A flowchart of multi-location scanning and panoramic image construction provided in this application embodiment;

[0023] Figure 4 A flowchart for identifying device components and determining region information is provided in an embodiment of this application;

[0024] Figure 5 A flowchart illustrating the association between equipment components and substation equipment ledgers is provided for embodiments of this application.

[0025] Figure 6 A flowchart for establishing parameter mapping relationships and generating preset bit configurations is provided in an embodiment of this application;

[0026] Figure 7 A schematic diagram illustrating the association between equipment component categories, inspection content, and identification rules, provided in an embodiment of this application;

[0027] Figure 8 A flowchart for standardizing device status information and determining abnormal status levels is provided in this application embodiment;

[0028] Figure 9 A flowchart for determining the execution strategy of a patrol task is provided in an embodiment of this application;

[0029] Figure 10A flowchart illustrating the matching of inspection results and online monitoring data is provided in this application embodiment.

[0030] Figure 11 A flowchart of a comprehensive confidence level verification and secondary review shooting process is provided for embodiments of this application;

[0031] Figure 12 This is a schematic diagram of the module structure of an intelligent substation integrated monitoring system provided in an embodiment of this application. Detailed Implementation

[0032] The present application will be further described below with reference to embodiments. It should be understood that the following embodiments are used to illustrate the technical solutions of the present application, and not to limit the scope of protection. Where there is no contradiction, technical features in different embodiments can be combined with each other.

[0033] Throughout this application, the term "image acquisition device" is used to refer to any device capable of acquiring images and changing the shooting direction or imaging range according to control commands. An image acquisition device may include an image sensor, a lens, a rotation mechanism, and a communication interface; a camera, a dome camera, or a pan-tilt zoom camera can all serve as specific implementations of an image acquisition device, but an image sensor itself is not equivalent to a complete image acquisition device.

[0034] "Inspection management equipment" refers to the logical processing entity that executes all processing steps of the inspection data processing method of this application. The logical processing entity can be composed of a single processing device, or it can be composed of a station-side edge computing device and a central control server that communicate with each other. When it is composed of multiple physical devices, each physical device, as a component of the inspection management equipment, collaboratively completes the method steps, and the intermediate results transmitted between the physical devices are regarded as the processing data within the inspection management equipment.

[0035] "Area Information" describes the location and extent of equipment components within scanned or panoramic image data. This can include at least one of the following: target image area, area boundary location, area center location, area width, area height, area size, and area size ratio. "Equipment Component Category" indicates the type of component to which the equipment component belongs, and can be a pointer meter, digital meter, disconnector switch, circuit breaker status indicator, oil level gauge, insulator, bushing, or equipment housing, etc. When inspecting and managing equipment, both the equipment component category and area information are obtained simultaneously based on the scanned image data.

[0036] "Substation Equipment Information" refers to a set of data used to identify equipment, which may include at least one of equipment ledgers, equipment layout data, equipment topology data, and equipment coding data. "Equipment Identifier" is used to uniquely identify the target equipment in the substation equipment information, and may be a combination of equipment unique code, bay number and equipment type, or other data that can uniquely identify the target equipment. When the same target equipment has multiple equipment components to be inspected, the inspection task identifier or component serial number is used to distinguish the inspection tasks corresponding to different equipment components.

[0037] "Inspection Task Information" refers to the set of information required for the control image acquisition device to acquire and process inspection images of target equipment components. It includes at least preset position configurations and recognition rules, and may also include equipment identification, inspection content, target imaging requirements, result output format, and verification rules. After the inspection task information is generated, it can be further associated with the execution strategy determined based on equipment status information.

[0038] "Identification rules" refer to the processing rules for determining inspection results based on inspection images. These rules can include at least one of the following: identification model, image preprocessing method, input image size, target region cropping method, inspection result identification confidence threshold, result post-processing rules, and multi-frame confirmation rules. The equipment component detection confidence threshold is used to filter equipment components in the scanned image; the inspection result identification confidence threshold is used to determine whether a single inspection identification result meets the identification rule requirements; and the comprehensive confidence threshold is used to determine whether the inspection result passes the final validity check. These three rules operate at different processing stages. The identification rules are not limited to a specific neural network model.

[0039] "Execution strategy" refers to the execution arrangement of patrol task information, which may include at least one of the following: execution priority, execution frequency, image acquisition parameters, and processing parameters of recognition rules. Among them, execution frequency indicates the number of times patrol task information is executed per unit time; the time length between two adjacent executions is called the execution interval, and an increase in execution frequency corresponds to a shorter execution interval.

[0040] like Figure 1As shown, the intelligent substation integrated monitoring system provided in this embodiment includes inspection management equipment, at least one image acquisition device, equipment status monitoring device, and substation equipment to be inspected. The image acquisition device is installed in a location capable of covering one or more substation inspection areas. It is used to change the shooting direction and imaging range according to control commands issued by the inspection management equipment, and to upload scanned image data or inspection images to the inspection management equipment. The equipment status monitoring device is used to collect online monitoring data and equipment alarm information corresponding to the substation equipment to be inspected, and upload the collected data to the inspection management equipment. The inspection management equipment is used to generate inspection task information based on the scanned image data and substation equipment information, determine the execution strategy of the inspection task information based on the equipment status information, process the inspection images collected according to the execution strategy, generate and store inspection results, and output corresponding reports.

[0041] like Figure 2 As shown, the inspection data processing method of the intelligent substation integrated monitoring system provided in this embodiment includes steps S101 to S105.

[0042] Step S101: The inspection and management equipment acquires scanned image data and substation equipment information obtained by the image acquisition device scanning the substation inspection area.

[0043] In step S102, the inspection and management equipment determines the equipment component, the category of the equipment component, and the area information of the equipment component in the scanned image data based on the scanned image data, and determines the equipment identifier corresponding to the equipment component based on the category of the equipment component, the area information, and the substation equipment information.

[0044] In step S103, the inspection management equipment generates inspection task information corresponding to the equipment identifier based on the area information and the category of equipment components. The inspection task information includes the preset position configuration of the image acquisition device and the identification rules for the equipment components.

[0045] Step S104: The inspection and management equipment obtains the equipment status information corresponding to the equipment identifier, and determines the execution strategy of the inspection task information based on the equipment status information.

[0046] In step S105, the inspection management equipment controls the image acquisition device to acquire inspection images according to the preset position configuration in accordance with the execution strategy, and processes the inspection images using recognition rules to obtain the inspection results corresponding to the equipment identifier.

[0047] The acquisition of scanned image data and the construction of panoramic image data in step S101 can be found in [reference needed]. Figure 3 The determination of equipment components, area information, and equipment identification in step S102 can be found in [reference needed]. Figure 4 and Figure 5The determination of preset bit configuration and recognition rules in step S103 can be found in [reference needed]. Figure 6 and Figure 7 The equipment status information processing and execution strategy determination in step S104 can be found in [reference needed]. Figure 8 and Figure 9 The execution of the inspection task in step S105, as well as the subsequent validity verification and secondary review of the inspection results, can be found in [reference needed]. Figure 10 and Figure 11 .

[0048] like Figure 3 As shown, when the image acquisition device is first connected, its installation location changes, or a reconfiguration command is received, the inspection and management equipment generates a scanning position sequence. The scanning position sequence includes multiple horizontal scanning positions and multiple vertical scanning positions, each represented by horizontal rotation parameters, vertical rotation parameters, and initial zoom parameters.

[0049] The inspection and management equipment determines adjacent scanning positions based on the horizontal and vertical field of view of the image acquisition device and the inspection area, so that an overlapping area for registration is retained between adjacent scanned images. For example, the overlap ratio of horizontally adjacent images or vertically adjacent images can be between 20% and 50%, but this application is not limited to this range.

[0050] The image acquisition device sequentially acquires multiple scan images according to the scan position sequence, and associates each scan image with the horizontal rotation parameters, vertical rotation parameters, zoom parameters, and acquisition time. The multiple scan images and their acquisition parameters constitute the scan image data.

[0051] The inspection and management equipment performs image registration on adjacent scanned images. As one implementation, feature points with stable grayscale or texture characteristics are first extracted from adjacent scanned images. Then, candidate matching pairs between the source and target scanned images are established based on feature description information. The inspection and management equipment uses geometric consistency judgment and duplicate sampling verification to eliminate erroneous matches, and selects at least four valid matching pairs that are not located on the same straight line from the remaining matching pairs to determine the perspective transformation relationship from the source to the target scanned image. For any pixel position in the source scanned image, the inspection and management equipment performs coordinate transformation and scaling normalization according to the perspective transformation relationship to obtain the corresponding position of that pixel in the target scanned image.

[0052] The perspective transformation relationship can be represented by a three-row, three-column set of transformation parameters, each obtained based on valid matching pairs. After determining the transformation parameters, the inspection and management equipment transforms the source image positions in the valid matching pairs to the target image coordinate system and calculates the pixel difference between the transformed position and the actual matching position. When the pixel difference exceeds the registration error condition, the corresponding matching pair is deleted and the transformation parameters are re-determined. The resulting perspective transformation relationship is used to describe the rotation, translation, scaling, and perspective changes between adjacent scanned images.

[0053] The inspection and management equipment transforms adjacent scanned images to a unified image coordinate system based on the perspective transformation relationship, performs seam fusion and brightness compensation on overlapping areas, and obtains panoramic image data corresponding to the substation inspection area. The panoramic image data can be a single stitched image or a logical panoramic image composed of multiple scanned images and the transformation relationship of each scanned image to a unified coordinate system.

[0054] When substation equipment is distributed across multiple discontinuous inspection areas, multiple panoramic image data can be generated, and these can be distinguished by inspection area markers. The panoramic image data described in this application includes the aforementioned physically stitched images and logical panoramic images.

[0055] like Figure 4 As shown, the patrol management equipment identifies equipment components from panoramic image data and outputs one or more identified targets. Each identified target includes at least the equipment component category, the target image region, and the identification confidence level. The equipment component category is determined by the equipment component identification model or a classification rule based on image features, and the target image region is used to determine the regional information of the equipment component.

[0056] The equipment component identification model can employ a target detection model pre-trained using multiple substation equipment samples. The model input is panoramic image data or image patches segmented from panoramic image data, and the model output is the category, target image region, and recognition confidence score. Retraining using samples from the target substation is not required when the target substation is first put into operation; inspection samples confirmed manually or through multi-source data can be used for subsequent incremental model updates, but incremental updates are not a necessary condition for initial identification.

[0057] The inspection and management equipment identifies equipment components from the target pool based on a component detection confidence threshold. This threshold can be a uniform threshold or set separately for different component categories. For example, a higher threshold can be used for components with stable shapes and sufficient samples, while a lower threshold can be used for components that are easily obscured, with subsequent verification. The threshold can be determined based on the confidence distribution of the identified targets in the verification samples, the allowable false negative rate, and the allowable false positive rate.

[0058] When the same equipment component corresponds to multiple overlapping targets, the inspection and management equipment uses non-maximum suppression to remove duplicate targets. Specifically, the overlapping area of ​​two candidate target image regions of the same category and the merged area covered by both candidate target image regions are determined separately. The ratio of the overlapping area to the merged area is used as the intersection-union ratio (IUR). The closer the IUR is to one, the higher the degree of overlap between the two candidate target image regions.

[0059] The Cross-Union Ratio (CUI) ranges from zero to one. When the CUI reaches a preset overlap threshold, the recognition confidence of two candidate targets is compared. The candidate target with the higher recognition confidence is retained, and the other candidate target is deleted. When the CUI does not reach the preset overlap threshold, both candidate targets are retained. The preset overlap threshold can be determined based on the CUI distribution of duplicate detection boxes of the same device component and detection boxes of adjacent different device components, so that duplicate detection boxes usually reach this threshold while adjacent different device components usually do not. In this application, the non-maximum suppression comparison object is the candidate prediction region, not the non-predicted region and the ground truth labeled region.

[0060] When the target image area of ​​a device component is represented by the upper left and lower right corners, the inspection and management equipment takes the arithmetic mean of the two horizontal positions and the two vertical positions to obtain the center position of the area. The horizontal position of the lower right corner is subtracted from the horizontal position of the upper left corner to obtain the area width, and the vertical position of the lower right corner is subtracted from the vertical position of the upper left corner to obtain the area height. The inspection and management equipment then compares the area width with the image width and the area height with the image height to obtain the width ratio and height ratio, and uses the larger of the two ratios as the area size ratio.

[0061] The bottom right corner of the target image region is wider than the top left corner horizontally, and the bottom right corner is wider than the top left corner vertically. Both the image width and height are positive values. The region center position is represented by pixel position, and the region width and height are represented by the number of pixels. The width ratio, height ratio, and region size ratio are all dimensionless values ​​between zero and one. The aforementioned positions, region sizes, and their derived center position and ratios all constitute region information.

[0062] like Figure 5 As shown, the inspection and management equipment obtains substation equipment information. One implementation method includes a substation equipment ledger, which records at least the equipment identifier, equipment category, and physical location. It may also record the associated bay, voltage level, adjacent equipment identifiers, equipment orientation, and equipment nameplate information.

[0063] To establish a positional mapping between image locations in panoramic image data and the physical locations of equipment, the inspection and management equipment acquires the installation location, installation posture, lens calibration information of the image acquisition device, as well as the horizontal rotation parameters, vertical rotation parameters, and zoom parameters corresponding to each scanned image. It also selects multiple spatial reference objects with known physical locations. These spatial reference objects can be known equipment, building structure markers, or specially set calibration marks. The inspection and management equipment determines the reference image location of each spatial reference object in the scanned images and panoramic image data, and obtains the reference physical location of each spatial reference object in the substation coordinate system.

[0064] The inspection and management equipment establishes a forward projection relationship from the physical location of the equipment to the panoramic image location based on the installation information of the image acquisition device, lens calibration information, scanning position parameters, and the image and physical locations of spatial reference objects. After establishment, a spatial reference object not involved in the establishment process is used for verification. Its physical location is converted into a predicted image location through the forward projection relationship, and the predicted image location is compared with the actual reference image location. When the pixel distance between the two exceeds a preset calibration error threshold, the installation posture parameters, lens parameters, or local partition mapping parameters are readjusted. The preset calibration error threshold can be determined based on the image resolution, the annotation error of the spatial reference object, and the minimum image spacing between adjacent similar devices.

[0065] When determining the candidate physical locations of equipment components, the inspection and management equipment first filters candidate equipment from the equipment ledger that matches the equipment category in the ledger, based on a preset category correspondence. This category matching does not require identical names, but rather a pre-defined attribution or correspondence between the equipment component category and the equipment category in the ledger. For example, a circuit breaker status indicator corresponds to a circuit breaker, and an oil temperature gauge corresponds to a transformer or oil temperature measuring device.

[0066] For each candidate device, the inspection and management equipment converts its physical location into a predicted image location in the panoramic image data according to the aforementioned forward projection relationship, and compares the image location difference between the predicted image location and the center location of the device component area. The physical locations corresponding to candidate devices whose image location differences are within the location matching range are grouped into a candidate physical location set, thereby realizing the determination of candidate physical locations corresponding to area information based on the location mapping relationship.

[0067] The location matching range is determined based on the maximum projection error of the spatial reference object, the location error of the device component identification, and a preset safety margin. It can be further limited to less than half the distance between the predicted image locations of adjacent similar devices to reduce the possibility of adjacent devices being simultaneously identified as location matches. Thus, the matching has an executable and objective basis for determination.

[0068] When only one candidate device exists that matches both category and location, the device identifier of that candidate device is determined as the device identifier corresponding to the device component. When multiple candidate devices exist, further filtering is performed based on the interval to which the device belongs, the spatial arrangement relationship of adjacent devices, the device nameplate recognition result, or the device topology relationship. When no candidate device exists or it still cannot be uniquely determined, the device component is marked as a target to be confirmed, and formal inspection task information is not automatically generated.

[0069] For example, when three adjacent pointer meters are identified in the same interval, the inspection and management equipment projects the physical locations of the three meters in the ledger onto the panoramic image data, compares the left-right arrangement and positional differences between the three predicted image locations and the three identified targets, and then determines the corresponding equipment identification by combining the measurement object or nameplate content of each meter.

[0070] like Figure 6 As shown, the preset position configuration of the image acquisition device includes horizontal rotation parameters, vertical rotation parameters, and zoom parameters. The horizontal rotation parameters are used to control the image acquisition device to rotate in the horizontal direction; the vertical rotation parameters are used to control the image acquisition device to rotate in the vertical direction; the zoom parameters can be optical zoom magnification, equivalent focal length, or lens drive position.

[0071] During the installation or recalibration of the image acquisition device, the patrol management equipment establishes a parameter mapping relationship between the image position and the rotation and zoom parameters. Specifically, the image acquisition device is controlled to be in multiple known horizontal rotation positions, vertical rotation positions, and zoom positions. At each calibration position, a calibration image containing the reference target is acquired, and the image center position, area size ratio, and corresponding horizontal rotation parameters, vertical rotation parameters, and zoom parameters of the reference target are recorded. The above corresponding data form a calibration sample.

[0072] The inspection and management equipment can generate a parameter mapping table from calibration samples. Once a device component is identified in the panoramic image data, the target scan image containing that component is first determined based on the registration relationship between the panoramic image data and each scanned image. The center position and boundary of the component's region are then transformed to the coordinate system of the target scan image. Subsequently, based on the center position and size ratio of the region in the target scan image, adjacent calibration samples are searched in the parameter mapping table and interpolated to obtain the initial horizontal rotation parameters, initial vertical rotation parameters, and initial zoom parameters.

[0073] As an alternative initial estimation method, the inspection and management equipment obtains the horizontal rotation parameters, vertical rotation parameters, and zoom parameters during the acquisition of the target scan image, and obtains the angular change corresponding to each pixel horizontally and vertically at the current zoom position. Based on the horizontal pixel offset of the equipment component center relative to the target scan image center and its corresponding horizontal angular change per pixel, the horizontal angle offset is determined; based on the vertical pixel offset and its corresponding vertical angular change per pixel, the vertical angle offset is determined. The horizontal angle offset is superimposed on the horizontal rotation parameters during target scan image acquisition according to the convention of the pan-tilt-zoom (PTZ) horizontal direction, and the vertical angle offset is superimposed or subtracted from the vertical rotation parameters during target scan image acquisition according to the convention between the image vertical direction and the PTZ vertical direction, to obtain the initial rotation parameters.

[0074] The change in angle per pixel can be determined by the field of view and image pixel size of the image acquisition device at different zoom positions, or it can be directly obtained from the aforementioned calibration samples. When using this initial estimation method, the center position of the device components, the center position of the image, the image pixel size, and the current zoom parameters are all taken from the same target scan image to avoid mixing panoramic image coordinates with single-frame scan image coordinates.

[0075] The inspection and management equipment determines the target imaging requirements based on the category of equipment components. Target imaging requirements may include at least one of the following: the center of the equipment component's area is located within a preset imaging area; the area size ratio is within the target ratio range; the key structure to be identified is fully displayed; and the image sharpness reaches a sharpness threshold. Image sharpness can be determined based on edge strength, local contrast, or the image quality evaluation results output by the recognition model within the equipment component's area; the image sharpness threshold can be determined based on the distribution of evaluation results between sharp and blurry samples.

[0076] The target scale range can be determined based on the minimum pixel size required by the recognition rules and the image size of the image acquisition device. For example, the target imaging requirements for pointer meters may include fully displaying the dial and pointers and ensuring that the dial area width reaches a preset pixel size; the target imaging requirements for disconnect switches may include fully displaying the moving contact, stationary contact, and transmission structure.

[0077] The inspection and management equipment determines the current area size ratio of the equipment component based on its area width and height in the target scan image, as well as the width and height of the target scan image. It then compares the current area size ratio with the target area size ratio required for target imaging. When the current area size ratio is smaller than the target area size ratio, the zoom level is increased; when the current area size ratio is larger than the target area size ratio, the zoom level is decreased. The initial adjustment range can be determined based on the relative difference between the target area size ratio and the current area size ratio, and the adjustment result is limited to the zoom range allowed by the image acquisition device.

[0078] The current area size ratio is positive, and the minimum zoom parameter allowed by the image acquisition device is no greater than the maximum zoom parameter. For lenses where the zoom position and image size are not approximately linearly related, the inspection and management equipment uses pre-recorded target size change data at different zoom positions for lookup or interpolation, rather than directly adjusting according to the ratio. The above processing is used to obtain the initial zoom parameters, and the final preset position configuration is determined through subsequent verification and correction.

[0079] The inspection and management equipment controls the image acquisition device to acquire verification images according to the initial preset position configuration, and then identifies equipment components again in the verification images. The inspection and management equipment determines the pixel distance between the center of the equipment component and the center of the verification image, and normalizes this distance using the diagonal pixel length of the verification image to obtain the center deviation; at the same time, it determines the absolute difference between the area size ratio of the equipment component in the verification image and the target area size ratio to obtain the size deviation.

[0080] When the center deviation exceeds the preset center deviation threshold, the horizontal rotation parameters are corrected based on the lateral offset direction and degree of the device component relative to the center of the verification image, and the vertical rotation parameters are corrected based on the longitudinal offset direction and degree of the longitudinal offset. When the size deviation exceeds the preset size deviation threshold, the zoom level is adjusted accordingly based on whether the device component's image is too small or too large. When the center deviation and size deviation are not greater than the preset center deviation threshold, and the device component meets other target imaging requirements, the preset position configuration is deemed to have passed verification. The inspection and management equipment repeatedly acquires verification images and corrects parameters until the preset position configuration passes verification or the preset maximum number of corrections is reached.

[0081] The preset center deviation threshold limits the maximum permissible deviation of the device component center from the center of the verification image. It can be determined based on the target eccentricity range allowed by the recognition rules and the rotation control error of the image acquisition device. The preset size deviation threshold limits the maximum permissible difference between the current area size ratio and the target area size ratio. It can be determined based on the minimum target pixel size required by the recognition rules and the permissible imaging ratio range. The preset maximum number of corrections can be determined based on the average number of convergences after a single parameter correction and the permissible configuration time. If the verification fails even after reaching the preset maximum number of corrections, the preset position is marked as a pending confirmation preset position, and the configuration with the highest degree of satisfaction is retained for subsequent confirmation.

[0082] like Figure 7As shown, the patrol management equipment pre-sets the correspondence between equipment component categories, patrol content, and identification rules. This correspondence can be implemented using rule tables, configuration files, or database records. Each record includes at least the equipment component category, patrol content identifier, and identification rule identifier, and may also include target imaging requirements, result data type, and online monitoring project type.

[0083] Table 1. Examples of the correspondence between equipment component categories, inspection contents, and identification rules.

[0084] pointer meter Meter reading identification Dial positioning, hand angle, and scale mapping rules Full display of dial and hands Continuous values Digital Meters Meter reading identification Character region detection and character recognition rules The digital display area is clear Continuous values disconnect switch Device status identification Rules for identifying the positional relationship between moving and stationary contacts Full display of key mechanical structure Discrete state Equipment casing Appearance Defect Identification Rules for identifying damage, dirt, and foreign objects The surface to be tested covers the target area Defect Category Circuit breaker status indicator component Device status identification Status indicator location and color recognition rules Full display status indicator area Discrete state Oil level gauge Meter reading identification Liquid level boundary and scale mapping rules Fully display liquid level boundaries and scale Continuous values casing Appearance Defect Identification Crack, discharge marks and dirt identification rules The surface to be tested is clear and unobstructed. Defect Category

[0085] The inspection management equipment queries the above correspondence based on the equipment component category to obtain the inspection content and identification rules, and associates the equipment identifier, preset position configuration, inspection content, and identification rules as inspection task information. The same equipment identifier can correspond to multiple inspection task information. For example, the same transformer can correspond to oil temperature gauge reading task, oil level identification task, and appearance defect task, respectively.

[0086] When multiple recognition rules exist for the same equipment component category, the recognition rule can be selected based on the equipment model, the resolution of the image acquisition device, the target imaging requirements, or historical recognition results. The selection result of the recognition rule is written into the patrol task information, and it does not need to be specified again by personnel during execution.

[0087] like Figure 8 As shown, the inspection and management equipment acquires equipment status information corresponding to the equipment identifier. Equipment status information includes at least one of online monitoring data, equipment alarm information, and historical defect information. Online monitoring data may include temperature, current, voltage, pressure, oil level, vibration, or partial discharge; equipment alarm information may include alarm type, alarm level, and duration; historical defect information may include defect type, defect severity, and defect handling status.

[0088] For online monitoring data, the inspection and management equipment determines the degree of data exceeding the limit based on the difference between the online monitoring value and the corresponding normal range. When the online monitoring value is within the normal range, the degree of data exceeding the limit is set to zero; when the online monitoring value is higher than the upper limit of the normal range, the difference between the online monitoring value and the upper limit of the normal range is first determined, and then the difference is divided by the pre-set allowable limit exceeding scale. If the result is less than zero, it is treated as zero; if it is higher than one, it is treated as one. When the online monitoring value is lower than the lower limit of the normal range, the difference between the lower limit of the normal range and the online monitoring value is processed in the same way.

[0089] The permissible limit-breaking scale is a positive value used to convert monitoring items with different dimensions and normal ranges into a limit-breaking level between zero and one. The permissible limit-breaking scale can be determined based on the abnormal range specified in the equipment operation specifications, the limit-breaking distribution of historical abnormal samples, or the alarm sensitivity allowed by maintenance personnel. Before conversion, it should be ensured that the online monitoring values, upper normal limit, lower normal limit, and permissible limit-breaking scale use consistent physical quantity units.

[0090] For equipment alarm information, the alarm event level is determined based on a preset correspondence between alarm level and normalized score. For example, no alarm, alert alarm, general alarm, and critical alarm are mapped to normalized values ​​from low to high. For historical defect information, the historical defect level is determined based on the defect severity, whether it has been processed, and the time interval from the current time.

[0091] Since equipment status information includes at least one of the aforementioned data types, when a certain data type is not available, the inspection and management equipment generates a default status record for that data type, sets the corresponding available identifier to zero, and excludes it from the comprehensive status calculation; when a certain data type is available, the corresponding available identifier is set to one. Therefore, even if only one of online monitoring data, equipment alarm information, or historical defect information is obtained, status records for the degree of monitoring data exceeding limits, alarm event level, and historical defect level can be formed respectively. The status record corresponding to the unobtained data does not affect the processing results of valid data.

[0092] When one or more types of available status data exist, the inspection and management equipment performs comprehensive processing on the available status data. Specifically, the normalized score for each type of available status data is multiplied by the basic weight corresponding to that data type. The products are summed to obtain the total weighted score. Then, the total weighted score is divided by the sum of the basic weights corresponding to all available status data to obtain the comprehensive status score. Data types that are not available are not included in the total weighted score, nor are they included in the sum of basic weights.

[0093] The base weight for each type of available status data is positive. Therefore, as long as at least one of the online monitoring data, equipment alarm information, and historical defect information is available, the sum of the base weights will be greater than zero. The values ​​for each normalized score and the overall status score range from zero to one. When only one type of status data is available, the overall status score equals the normalized score of that status data. Each base weight can be determined based on the equipment type, the historical reliability of various status data types, and historically confirmed abnormal samples.

[0094] The inspection and management equipment is equipped with progressively increasing thresholds for the first, second, and third levels, all ranging from zero to one. A comprehensive status score below the first level threshold is classified as normal; a score reaching the first level threshold but below the second level threshold is classified as attentive; a score reaching the second level threshold but below the third level threshold is classified as abnormal; and a score reaching or exceeding the third level threshold is classified as severe.

[0095] The level threshold can be determined based on the equipment operation specifications, the score distribution of various states in historical samples, the allowable false alarm rate, and the allowable false alarm rate. The specific value may vary depending on the type of equipment, and a fixed threshold is not a necessary condition for all implementation methods in this application.

[0096] As an alternative implementation method that does not employ weighted averaging, the inspection and management equipment determines candidate anomaly levels based on the degree of exceedance of monitoring data, alarm event levels, and historical defect levels. The highest level among the candidate anomaly levels corresponding to available data is then selected as the final anomaly status level. This method is suitable for scenarios where any severe condition requires an immediate increase in inspection resources, and it also falls under the category of determining anomaly status levels based on standardized processing results.

[0097] like Figure 9 As shown, the patrol management equipment pre-stores the correspondence between abnormal status levels and execution strategies. The execution strategy includes at least one of the following: execution priority, execution frequency, image acquisition parameters, and processing parameters for recognition rules. When the execution strategy includes execution priority, the correspondence is used to determine the task ordering for different abnormal status levels; when the execution strategy includes execution frequency, the correspondence is used to determine the number of times a task is executed per unit time.

[0098] Table 2. Example of the correspondence between abnormal status levels and patrol mission execution strategies

[0099] normal conventional Execution frequency per unit time Standard resolution, single frame capture Standard confidence conditions Notice improve Increase the number of executions per unit time and shorten the execution interval. Increase resolution or increase the number of capture frames Add multi-frame consistency confirmation abnormal high Further increase the number of executions per unit time and further shorten the execution interval. Add close-up capture High-precision processing configuration serious Highest High-frequency execution or tracking acquisition Continuously acquire and retain original images Strict review and alarm confirmation

[0100] When the execution strategy includes execution priority, the higher the abnormal status level, the higher the execution priority of the patrol task information; when the execution strategy includes execution frequency, the higher the abnormal status level, the more times it is executed per unit time, and the execution interval between two adjacent executions is correspondingly shortened. The specific number of executions and execution intervals can be determined based on the equipment type, the available time of the image acquisition device, and the station patrol plan.

[0101] Image acquisition parameters may include image resolution, exposure parameters, focus parameters, zoom parameters, and number of consecutive acquisition frames; the processing parameters for recognition rules may include input image size, recognition confidence conditions, result post-processing parameters, and multi-frame consistency conditions.

[0102] When multiple inspection tasks use the same image acquisition device, the inspection management equipment generates a task queue based on execution priority and planned execution time. Inspection tasks with increased abnormal status levels can be moved to an earlier position in the task queue, and new task instances can be generated based on execution frequency.

[0103] When equipment status information changes, the inspection and management equipment recalculates the abnormal status level and updates the execution strategy. When the abnormal status recovers, the execution priority and frequency can be reduced; when the abnormal status persists or worsens, the execution priority can be increased, close-up preset positions can be added, or continuous acquisition can be enabled.

[0104] The inspection and management equipment selects the inspection task information to be executed according to the execution strategy and sends a preset position control command to the image acquisition device. The image acquisition device reaches the corresponding preset position based on the horizontal rotation parameters, vertical rotation parameters, and zoom parameters, and acquires the inspection image after stabilization.

[0105] The inspection management equipment processes the inspection images according to the identification rules in the inspection task information. For meter reading identification, the inspection results may include the numerical value and identification confidence level; for equipment status identification, the inspection results may include the status of the equipment being in a split position, a closed position, or an intermediate state, as well as the identification confidence level; for appearance defect identification, the inspection results may include the defect category, the defect image area, and the identification confidence level.

[0106] The inspection management equipment associates and stores the inspection results with the equipment identification, and records the preset location configuration, acquisition time, image acquisition parameters and recognition rule version, so as to facilitate subsequent validity verification and traceability.

[0107] like Figure 10 As shown, after obtaining the inspection results corresponding to the equipment identifier, the inspection management equipment acquires the online monitoring data corresponding to that equipment identifier and determines whether a corresponding target monitoring item exists in the online monitoring data based on the result type of the inspection results. If a target monitoring item exists, the inspection management equipment determines the target monitoring item and matching rules, and obtains the matching degree according to the method described in the following paragraphs; if no target monitoring item exists, the inspection management equipment adopts a preset neutral matching degree, indicating that the online monitoring data neither supports nor negates the inspection result.

[0108] When the inspection results are continuous values ​​such as meter readings, the target online monitoring item is the online monitoring value that is the same as the object being measured by the meter. The inspection management equipment first aligns the inspection image recognition values ​​and the online monitoring values ​​in terms of time and units, and then determines the absolute difference between the two. The absolute difference is divided by a pre-set allowable difference scale to obtain the difference ratio; the initial numerical matching degree is obtained by subtracting this difference ratio from one. If the initial numerical matching degree is less than zero, it is treated as zero. Therefore, when two values ​​are completely identical, the numerical matching degree is one; when the difference between two values ​​reaches or exceeds the allowable difference scale, the numerical matching degree is zero.

[0109] The permissible difference scale is a positive value and can be determined jointly based on the permissible error of the measuring equipment, the permissible error of the online monitoring device, the permissible variation caused by asynchronous data acquisition time, and the normal fluctuation range of the monitored item. The inspection image recognition values, online monitoring values, and permissible difference scales use the same unit. The smaller the difference between two values, the higher the degree of numerical matching.

[0110] When the inspection results show discrete states such as switch open / closed status, the target online monitoring item is the remote signaling status of the corresponding equipment. If the inspection results are consistent with the remote signaling status, the status matching degree is set to one; if the two clearly conflict, the status matching degree is set to zero; if the remote signaling status is invalid or in an unknown state, a neutral matching degree is used and it is marked as incomplete data.

[0111] When the inspection results indicate an external defect in the equipment, the inspection management system selects one or more relevant online monitoring items based on the defect category. For example, overheating or discoloration defects are associated with temperature monitoring data, leakage defects with oil level or pressure monitoring data, and discharge traces with partial discharge monitoring data. For each relevant online monitoring item, the degree of correlation between zero and one is determined based on the direction and degree of change of that monitoring item relative to the normal range. Then, positive correlation weights are assigned according to the strength of the correlation between each online monitoring item and the corresponding defect category, and a weighted average of the degree of correlation for each item is calculated to obtain the defect matching degree.

[0112] When a certain inspection result has no directly related online monitoring item, irrelevant online monitoring data is not used to negate the inspection result. The neutral matching degree indicates that the online monitoring data neither supports nor negates the inspection result; its value lies between the lowest and highest matching degree. As an example, a value between zero and one can be used, or it can be determined based on the misclassification rate corresponding to missing samples in historical data. When using the neutral matching degree, an incomplete data flag is also written, and a second verification shot can be triggered based on recognition confidence, image clarity, or multi-frame consistency conditions.

[0113] Therefore, the matching degree is not a comparison of arbitrary irrelevant data. Instead, when there is a target monitoring item, it is a normalized degree value obtained by comparing data of the same equipment, the same measurement object, or data with a clear defect correlation according to the matching rules corresponding to the inspection result type. When there is no target monitoring item, a neutral matching degree is used as a verification input that does not favor support or negation.

[0114] like Figure 11 As shown, the inspection management equipment obtains the identification confidence level and the aforementioned matching degree corresponding to the inspection results, and converts both to the same value range of zero to one. Then, a fusion coefficient between zero and one is selected, the identification confidence level is multiplied by the fusion coefficient, the matching degree is multiplied by the remaining proportion obtained by subtracting the fusion coefficient from one, and the two products are added together to obtain the comprehensive confidence level.

[0115] The fusion coefficient is used to determine the relative contribution of identification confidence and matching degree to the overall confidence. The larger the fusion coefficient, the greater the contribution of identification confidence; the smaller the fusion coefficient, the greater the contribution of matching degree. Since identification confidence, matching degree, and both contribution ratios are all between zero and one, and the sum of the two contribution ratios is one, the overall confidence also falls between zero and one.

[0116] The fusion coefficient can be determined based on historically confirmed inspection samples. For example, in scenarios where the image recognition model is stable and the online monitoring data is updated infrequently, the fusion coefficient corresponding to the recognition confidence level can be increased; in scenarios where the online monitoring data is reliable and synchronized with the inspection results in time, the fusion coefficient corresponding to the matching degree can be increased.

[0117] The inspection and management equipment compares the overall confidence level with the overall confidence level threshold. The overall confidence level threshold can be determined based on the false positive rate, false negative rate, and secondary review resource consumption of historical samples, and can be set separately for different equipment component categories or inspection contents.

[0118] When the overall confidence level reaches the preset confidence threshold, the inspection results are determined to have passed the validity check; when the overall confidence level does not reach the preset confidence threshold, the inspection results are determined to have failed the validity check and a second review shooting instruction is generated.

[0119] The inspection and management equipment determines the reasons for failure to pass the validity check based on the center position of the equipment component in the inspection image, the area size ratio, the image sharpness evaluation result, and the continuous frame recognition result. If the center of the equipment component exceeds the preset imaging area, it is determined that the shooting direction is incompatible; if the area size ratio is lower than the target ratio range, it is determined that the imaging size is insufficient; if the image sharpness does not reach the image sharpness threshold, it is determined that the image sharpness is insufficient; if the recognition results of different frames do not meet the multi-frame consistency condition, it is determined that the single-frame recognition result is unstable. A secondary verification shooting command is generated based on the determined reasons: if the shooting direction is incompatible, the command includes corrected horizontal rotation parameters and / or vertical rotation parameters; if the imaging size is insufficient, the command includes improved zoom parameters; if the image sharpness is insufficient, the command includes adjusted focus parameters or exposure parameters; if the single-frame recognition result is unstable, the command includes an increased number of consecutive acquisition frames.

[0120] The patrol management equipment controls the image acquisition device to re-acquire patrol images according to the secondary review shooting instruction, and processes the re-acquired patrol images using the same recognition rule or the recognition rule corresponding to the secondary review. The validity can be re-determined based on the consistency of multi-frame recognition results, the confidence level of secondary recognition, and online monitoring data.

[0121] When the results of the second review inspection still fail the validity check, a manual confirmation mark can be output, the initial inspection image and the second inspection image can be saved, and a review prompt can be generated, instead of directly taking the low-confidence result as the final conclusion of equipment abnormality.

[0122] In one optional implementation, taking a smart substation equipped with transformer oil temperature gauges, disconnect switches, and bushings as an example, the image acquisition device generates panoramic image data after scanning at multiple horizontal and vertical positions. The inspection and management equipment identifies the oil temperature gauge target, obtains the oil temperature gauge type, target image area, and recognition confidence level, and determines the center position and size ratio of the area based on the target image area.

[0123] The inspection and management equipment determines candidate physical locations based on the location mapping relationship between panoramic image data and equipment ledgers. It then selects equipment from the equipment ledger that is classified as oil temperature gauge and whose physical location matches the candidate physical location, and uses the unique code of that equipment as the equipment identifier.

[0124] The inspection and management equipment determines the initial horizontal rotation parameters, vertical rotation parameters, and zoom parameters based on the parameter mapping relationship, collects and verifies images, and corrects them according to the center deviation and size deviation to obtain the preset position configuration corresponding to the oil temperature gauge; then, based on the oil temperature gauge type, it determines the meter reading inspection content and pointer meter identification rules to generate inspection task information.

[0125] The equipment condition monitoring device provides online temperature data and alarm information for the corresponding transformer. The inspection and management system standardizes the processing of available condition data, determines the level of abnormal conditions, and increases the priority and frequency of oil temperature gauge inspection tasks based on the level.

[0126] After completing the inspection task, the system identifies the oil temperature reading and the recognition confidence level. The inspection management equipment then performs numerical matching between the oil temperature reading and the online temperature data of the same measurement object to determine the degree of matching and the overall confidence level. If the overall confidence level is insufficient, the system increases the zoom parameter and refocuses before performing a second verification shot.

[0127] The above example illustrates the continuous data transmission relationship between scanned image data, device identification, inspection task information, device status information, execution strategy, and inspection results. Those skilled in the art can implement the complete processing flow described in this specification.

[0128] In a disconnector switch inspection task, the equipment component identification results showed that the moving and stationary contacts of the disconnector switch were in the engaged position. The identification rules output the engaged state and the corresponding identification confidence level. The inspection management equipment obtains the remote signaling status of the same disconnector switch based on the equipment identification and aligns the inspection image acquisition time with the remote signaling data time.

[0129] When the remote signaling status also indicates a combined position, the status matching degree is determined to be the highest matching degree; when the remote signaling status indicates a separated position, the status matching degree is determined to be the lowest matching degree. If the remote signaling data is in a maintenance shielded, communication interrupted, or unknown state, a neutral matching degree is used and an incomplete data flag is written.

[0130] For example, the initial inspection image is affected by the obstruction of the connecting rod, and the recognition rule outputs a "closed" position, while the effective remote signaling status is "separated," creating a clear conflict. Therefore, the overall confidence level does not reach the overall confidence threshold. Based on the reason for the incomplete display of the key mechanical structure, the inspection management equipment adjusts the horizontal rotation parameters to the adjacent observation direction and appropriately reduces the zoom level to ensure that the moving contact, stationary contact, and transmission structure simultaneously enter the inspection image.

[0131] If the identification result in the secondary review image changes to a quantile and is consistent with the remote signaling status, the secondary inspection result is determined to have passed the validity check; if the secondary identification still conflicts with the remote signaling status, a manual confirmation mark is output, and the two inspection images, remote signaling data and corresponding time information are retained.

[0132] During a transformer equipment casing inspection, identification rules identified color anomalies and localized discoloration defects in the bushing connection area. The inspection management equipment selected temperature monitoring items relevant to the area from online monitoring data based on the defect category and compared temperature data near the defect identification time with the corresponding normal range.

[0133] When temperature data shows a significant increase in the same direction as the normal range, a higher degree of single-item correlation is determined; when temperature data shows only slight fluctuations, a lower degree of single-item correlation is determined. If load current data is also present, the rationality of the temperature rise under operating conditions can be determined based on the current load level, and the temperature correlation degree and load correlation degree can be combined through pre-set positive correlation weights.

[0134] For example, if a discolored area is identified in the inspection image and the temperature monitoring value is consistently higher than the recent stable range of similar equipment, the defect matching degree is improved. If the identification confidence is also high, the overall confidence reaches the overall confidence threshold, and the inspection result of the overheating-related defect is output. If the temperature data is normal and there is strong reflection in the image, the overall confidence is reduced, and a second verification is performed by changing the exposure parameters and shooting direction.

[0135] This example illustrates that visual defect verification does not require online monitoring data and image defects to have the same data format. Instead, it selects monitoring items with a clear correlation based on the defect category, and then determines the degree of matching based on the direction and degree of change of the monitoring items.

[0136] During an insulator appearance inspection, the identification rules detected a notch at the edge of the insulator skirt, but the equipment condition monitoring device did not have an online monitoring item that could directly reflect this notch. The inspection management equipment determined that there was no target monitoring item corresponding to this result type in the online monitoring data and did not use temperature, current, or other irrelevant data to reject the inspection result.

[0137] The inspection management equipment uses the neutral matching degree as the matching degree of the inspection result. As an example, the neutral matching degree can be the middle position of the range of zero to one; this value only indicates the lack of external supporting information, not that the defect does not exist. The inspection management equipment also writes an incomplete data flag and improves the role of identification confidence, image clarity, and multi-frame consistency in validity verification.

[0138] If the pixel size of the gap area is small or appears only in a single frame in the initial inspection image, the overall confidence level may not reach the threshold. The inspection management equipment increases the zoom level and continuously acquires multiple inspection images; when the position and shape of the gap remain consistent across multiple frames, the recognition confidence level is stable, and the image clarity reaches the threshold, the results of the second inspection can be confirmed.

[0139] This process avoids incorrectly rejecting genuine appearance defects due to the absence of corresponding online monitoring items, while reducing misidentification caused by stains, shadows, or short-term occlusions through secondary verification and multi-frame consistency.

[0140] During a surge arrester inspection, the inspection management equipment only acquired equipment alarm information, but not real-time online monitoring data or historical defect information. The inspection management equipment set the available flags corresponding to the equipment alarm information to valid, and set the available flags for the other two types of data to invalid, preventing the unacquired data from participating in the overall status processing.

[0141] For example, when the alarm information is a general alarm, the alarm event level is determined according to the correspondence between the alarm level and the normalized score; since only this status data is available, the comprehensive status score is equal to the normalized score of the alarm event level, and then the abnormal status level is determined according to the level threshold.

[0142] If the abnormal state level is a warning level, the execution strategy may only include increasing the execution frequency without changing the execution priority; if the execution strategy also includes execution priority, the inspection task can be moved forward. Therefore, the components of the execution strategy in the claims can be selected based on available resources and business configuration, without requiring simultaneous adjustment of all components each time.

[0143] When online monitoring data is subsequently restored or historical defect information is obtained, the inspection and management equipment re-determines the standardized status data and updates the abnormal status level and execution strategy, so that the absence of status information will not interrupt the execution of the inspection task.

[0144] When three identical pointer meters are placed side-by-side in the same switch bay, the equipment identification cannot be uniquely determined solely by the equipment component category. The inspection and management equipment projects the physical locations of the three meters from the logbook onto the panoramic image data and compares the predicted image locations with the center locations of the three identified targets.

[0145] When the predicted location differences are all within the location matching range, further compare the left-right arrangement order, dial range, nameplate text, or measured object of the three meters. If the arrangement order of the identified targets is consistent with the arrangement order in the ledger, the equipment identification is determined according to the corresponding order; if the nameplate identification result conflicts with the location result, the corresponding target is marked as a target to be confirmed, and a formal inspection task is not automatically established.

[0146] When the image acquisition device is moved due to maintenance, the inspection and management equipment does not use the original parameter mapping relationship. Instead, it reacquires the image position of the spatial reference object, verifies the forward projection relationship from the physical position of the device to the panoramic image position, and re-establishes the parameter mapping table between the image position and the rotation and zoom parameters.

[0147] After rescanning, the associations between equipment component categories, area information, and equipment identifiers are redefined. The original inspection task information can inherit the inspection content and identification rules based on the equipment identifiers, but the preset position configurations are regenerated by the new parameter mapping relationships. This example illustrates that the business attributes in an inspection task can be reused, while the preset position configurations related to the installation location of the image acquisition device need to be updated adaptively.

[0148] After reconfiguration, the inspection and management equipment uses verification images to verify each preset position; only preset positions that meet the corresponding thresholds for center deviation, size deviation and image clarity are set to the executable state, while the remaining preset positions are left in the pending confirmation state.

[0149] like Figure 12 As shown, the intelligent substation integrated monitoring system includes a data acquisition module, an equipment association module, a task generation module, a strategy determination module, and an inspection and processing module.

[0150] The data acquisition module is used to acquire scanned image data obtained by the image acquisition device from scanning the substation inspection area, as well as substation equipment information. The data acquisition module can also acquire the acquisition parameters of the image acquisition device, equipment status information, and online monitoring data.

[0151] The equipment association module is used to determine equipment components, their categories, and their regional information within the scanned image data based on the scanned image data. It then determines the corresponding equipment identifier based on the category, region information, and substation equipment information. The equipment association module may include a panoramic image construction submodule, an equipment component identification submodule, and an equipment identity matching submodule.

[0152] The task generation module is used to generate inspection task information corresponding to equipment identifiers based on area information and category. The task generation module may include a parameter mapping submodule, a preset bit configuration submodule, and an identification rule matching submodule.

[0153] The strategy determination module is used to obtain the equipment status information corresponding to the equipment identifier, and determine the execution strategy for the inspection task information based on the equipment status information. The strategy determination module may include a status standardization submodule, an abnormal status level determination submodule, and a strategy mapping submodule.

[0154] The inspection processing module controls the image acquisition device to acquire inspection images according to preset position configurations based on the execution strategy, and processes the inspection images using recognition rules to obtain the inspection results corresponding to the equipment identifiers. The inspection processing module may also include a matching degree determination submodule, a validity verification submodule, and a secondary review control submodule.

[0155] The modules described above can be implemented in software, hardware, or a combination of both. Multiple modules can be deployed on the same processing device or on interconnected edge computing devices and servers. The deployment location of the modules does not change the data processing relationships between them.

[0156] The inspection and management equipment may include at least one processor, a memory, and a communication interface. The memory stores program instructions, which the processor executes to implement the method steps described in this specification. The communication interface is used to communicate with image acquisition devices, equipment status monitoring devices, and centralized control systems.

[0157] The image registration method in this application is not limited to feature point homography registration. Direct registration, depth information-based registration, or initial registration based on the rotation parameters of the image acquisition device followed by image content correction can also be used.

[0158] The association between equipment components and equipment identifiers can also be achieved by combining equipment nameplates, equipment QR codes, equipment topology relationships, or the arrangement of multiple adjacent equipment components. This information can be used individually or in conjunction with spatial location matching results; when the results of multiple association information conflict, the corresponding equipment component is marked as a target to be confirmed.

[0159] In addition to horizontal rotation parameters, vertical rotation parameters, and zoom parameters, preset position configurations may also include focus parameters, exposure parameters, aperture parameters, and image stabilization waiting time. Additional parameters are used to further improve the imaging quality of the survey images.

[0160] Standardization methods for equipment status information are not limited to linear normalization; interval mapping, level mapping, or quantile mapping based on historical distribution can also be used. The comprehensive status score can also be replaced by logic gating or a method that determines the highest level.

[0161] This embodiment preferably adopts the aforementioned method of determining the overall confidence level based on the fusion coefficient. This method has clearly defined the input data, processing order, and parameter determination basis. For different inspection result types, fusion coefficients and preset confidence thresholds can be set separately, but it should be ensured that the identification confidence level and matching degree are within the same value range, and that the processing rules adopted can output an overall confidence level that increases with the improvement of identification confidence level and data matching degree.

[0162] Without contradicting each other, the technical features in the above embodiments can be combined to adapt to different types of substation equipment, monitoring data conditions, and image acquisition device configurations.

Claims

1. A method for processing inspection data in an intelligent substation integrated monitoring system, characterized in that, The method, applied to inspection and management equipment, includes: Acquire scanned image data and substation equipment information obtained by the image acquisition device scanning the substation inspection area; Based on the scanned image data, determine the equipment component, the category of the equipment component, and the area information of the equipment component in the scanned image data, and determine the equipment identifier corresponding to the equipment component based on the category, the area information, and the substation equipment information; Based on the area information and the category, patrol task information corresponding to the device identifier is generated. The patrol task information includes the preset position configuration of the image acquisition device and the identification rules for the device components. Obtain the device status information corresponding to the device identifier, and determine the execution strategy of the patrol task information based on the device status information; According to the execution strategy, the image acquisition device is controlled to acquire inspection images according to the preset position configuration, and the inspection images are processed using the recognition rules to obtain the inspection results corresponding to the device identifier; The method further includes, after obtaining the inspection result corresponding to the equipment identifier, acquiring online monitoring data corresponding to the equipment identifier; determining whether a target monitoring item corresponding to the result type exists in the online monitoring data based on the result type of the inspection result; if the target monitoring item exists, identifying the target monitoring item from the online monitoring data and determining the matching rule corresponding to the result type; comparing the inspection result with the online monitoring data corresponding to the target monitoring item according to the matching rule to obtain the matching degree corresponding to the inspection result; if the target monitoring item does not exist, determining a preset neutral matching degree as the matching degree corresponding to the inspection result; validating the inspection result based on the matching degree; generating a secondary verification shooting instruction corresponding to the equipment identifier when the inspection result fails the validity verification; controlling the image acquisition device to re-acquire the inspection image corresponding to the equipment component according to the secondary verification shooting instruction, and processing the re-acquired inspection image using the recognition rule; The step of validating the inspection results based on the matching degree includes: obtaining the identification confidence level corresponding to the inspection results; determining the comprehensive confidence level corresponding to the inspection results based on the identification confidence level and the matching degree; comparing the comprehensive confidence level with a preset confidence threshold; determining that the inspection results pass the validity verification when the comprehensive confidence level reaches the preset confidence threshold; and determining that the inspection results fail the validity verification when the comprehensive confidence level does not reach the preset confidence threshold, and generating the secondary review shooting instruction.

2. The method as described in claim 1, characterized in that, The scanned image data includes multiple scanned images acquired by the image acquisition device at multiple horizontal and multiple vertical scanning positions; The step of determining the device component, the category of the device component, and the region information of the device component in the scanned image data based on the scanned image data includes: Image registration and image stitching are performed on adjacent scanned images from the multiple scanned images to obtain panoramic image data corresponding to the substation inspection area; The panoramic image data is used to identify device components to obtain the category, target image region, and recognition confidence level of the identified target. The device component is determined from the target based on a preset confidence condition, and the position and size of the device component in the panoramic image data are determined based on the target image region corresponding to the device component. The position and size of the region are then used as the region information.

3. The method as described in claim 2, characterized in that, The substation equipment information includes a substation equipment ledger, which records equipment identification, equipment category, and physical location. The step of determining the equipment identifier corresponding to the equipment component based on the category of the equipment component, the area information, and the substation equipment information includes: Spatially align the panoramic image data with the substation equipment ledger to establish a positional mapping relationship between the image position in the panoramic image data and the physical position of the equipment. The candidate physical locations corresponding to the region information are determined based on the location mapping relationship; The system retrieves a target device from the substation equipment ledger whose equipment category and equipment component category satisfy a preset category correspondence and whose physical location matches the candidate physical location. The device identifier of the target device is then determined as the device identifier corresponding to the equipment component.

4. The method as described in claim 3, characterized in that, The patrol management equipment is pre-set with a correspondence between the categories of the equipment components, the patrol content, and the identification rules; The generation of patrol task information corresponding to the device identifier also includes: Based on the category of the equipment component, determine the inspection content and identification rules corresponding to the equipment component from the correspondence; The equipment identifier, the preset position configuration, the inspection content, and the identification rules are associated to generate the inspection task information; The inspection content is used to instruct the equipment components to perform at least one of meter reading identification, equipment status identification, or equipment appearance defect identification.

5. The method as described in claim 4, characterized in that, The preset position configuration includes horizontal rotation parameters, vertical rotation parameters, and zoom parameters; the preset position configuration in the patrol mission information is determined in the following way: Obtain the parameter mapping relationship between the image position and the rotation zoom parameters of the image acquisition device; Based on the area information, determine the center position and size ratio of the area corresponding to the device component, and determine the target imaging requirements corresponding to the device component based on the category of the device component; Based on the parameter mapping relationship, the center position of the region, the size ratio of the region, and the target imaging requirements, the horizontal rotation parameters, the vertical rotation parameters, and the zoom parameters are determined to obtain the preset position configuration.

6. The method according to any one of claims 1 to 5, characterized in that, The device status information includes at least one of online monitoring data, device alarm information, and historical defect information; The execution strategy for determining the inspection task information based on the equipment status information includes: When the device status information includes the online monitoring data, the degree of exceeding the limit of the monitoring data corresponding to the device identifier is determined based on the online monitoring data. When the device status information includes the device alarm information, the alarm event level corresponding to the device identifier is determined based on the device alarm information; When the device status information includes the historical defect information, the historical defect level corresponding to the device identifier is determined based on the historical defect information; Standardize at least one of the determined monitoring data exceedance level, alarm event level, and historical defect level, and determine the abnormal state level corresponding to the device identifier based on the standardization result; The execution strategy for the patrol task information is determined based on the level of the abnormal state.

7. The method as described in claim 6, characterized in that, The execution strategy includes at least one of execution priority, execution frequency, image acquisition parameters, and processing parameters of the recognition rules; The execution strategy for determining the patrol task information based on the abnormal state level includes: Based on the preset policy correspondence between the abnormal state level and the execution policy, at least one of the following is determined: execution priority, execution frequency, image acquisition parameters, and processing parameters of the recognition rule corresponding to the abnormal state level. Specifically, when the execution strategy includes the execution priority, the higher the abnormal state level, the higher the execution priority of the patrol task information; when the execution strategy includes the execution frequency, the higher the abnormal state level, the higher the execution frequency of the patrol task information.

8. A smart substation integrated monitoring system, used to implement the method described in claim 1, characterized in that, include: The data acquisition module is used to acquire scanned image data and substation equipment information obtained by the image acquisition device scanning the substation inspection area. The device association module is used to determine the device component, the category of the device component, and the area information of the device component in the scanned image data based on the scanned image data, and to determine the device identifier corresponding to the device component based on the category, the area information, and the substation equipment information. The task generation module is used to generate patrol task information corresponding to the device identifier based on the area information and the category. The patrol task information includes the preset position configuration of the image acquisition device and the identification rules for the device components. The strategy determination module is used to obtain the device status information corresponding to the device identifier and determine the execution strategy of the patrol task information based on the device status information. The inspection processing module is used to control the image acquisition device to acquire inspection images according to the preset position configuration in accordance with the execution strategy, and to process the inspection images using the recognition rules to obtain the inspection result corresponding to the device identifier.

Images