A substation video patrol point automatic planning method, system, device and medium based on multi-source data fusion

Abstract

This invention relates to the field of substation inspection technology, and discloses a method, system, equipment, and medium for automatic planning of substation video inspection points based on multi-source data fusion. The method includes: acquiring substation equipment status data and environmental monitoring data, fusing them to generate a target inspection influencing factor set; generating multiple initial inspection candidate points based on this set, and simultaneously performing multiple rounds of point optimization processing in conjunction with equipment status and environmental data to obtain the target inspection points and their coverage evaluation results; finally, integrating and generating inspection point planning data for configuring camera deployment parameters. This invention achieves dynamic planning of inspection points by setting up a multi-source data-driven and iterative optimization mechanism, enabling video surveillance to fully cover the entire substation area.

Description

Technical Field

[0001] This invention relates to the field of substation inspection technology, and in particular to a method, system, equipment and medium for automatic planning of substation video inspection points based on multi-source data fusion. Background Technology

[0002] Currently, the planning of video inspection points in substations still relies heavily on manual experience. There is a lack of automatic planning methods for inspection points driven by multi-source data fusion. In this situation, if potential hazards such as abnormal temperature or damage to equipment in the substation occur, maintenance personnel can only determine the camera placement by manually reviewing historical inspection records and conducting on-site surveys. This operating mode is not only inefficient but also difficult to cover all critical equipment areas. Especially in high-voltage substations with dense equipment and complex structures, or in situations where extreme weather causes drastic changes in environmental parameters, it is difficult to achieve accurate visual coverage and dynamic adjustment of high-risk equipment if existing technologies are still used. Summary of the Invention

[0003] In view of the aforementioned existing problems, the present invention is proposed.

[0004] Therefore, this invention provides an automatic planning method, system, equipment, and medium for substation video inspection points based on multi-source data fusion, which can solve the problems of high missed detection rate of key equipment in the prior art.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides an automatic planning method for substation video inspection points based on multi-source data fusion, comprising: Acquire substation equipment status data and environmental monitoring data to be processed. The substation equipment status data includes equipment temperature anomaly information and appearance anomaly information obtained based on video image recognition. The environmental monitoring data includes humidity, temperature and wind speed. The substation equipment status data and environmental monitoring data are fused and processed to obtain a set of target inspection influencing factors; Based on the set of target inspection impact factors, multiple initial inspection candidate locations are generated; Based on the multiple initial inspection candidate points, the substation equipment status data, and the environmental monitoring data, multiple point optimization processes are performed to obtain the target inspection point and the coverage assessment result corresponding to the target inspection point. The target inspection point includes multiple historical candidate points before the target inspection point and the historical coverage assessment result corresponding to each historical candidate point. The target inspection impact factor set, the target inspection points, and the coverage assessment results are integrated and processed to obtain the inspection point planning data corresponding to the substation. The inspection point planning data is used to configure the camera deployment parameters of the video inspection system.

[0006] As a preferred embodiment of the automatic planning method for substation video inspection points based on multi-source data fusion described in this invention, the step of performing multiple point optimization processes based on the multiple initial candidate inspection points, the substation equipment status data, and the environmental monitoring data to obtain the target inspection point and the coverage evaluation result corresponding to the target inspection point includes: Determine the optimization round M corresponding to the multiple point optimization processes, where M is an integer greater than 1; Based on the multiple initial inspection candidate points, the substation equipment status data, and the environmental monitoring data, M-1 point optimization processes are performed to determine M-1 historical candidate points and the historical coverage assessment results corresponding to each historical candidate point; The target inspection point is generated based on the M-1 historical candidate points, the historical coverage assessment results corresponding to each historical candidate point, and the multiple initial inspection candidate points. The coverage assessment results corresponding to the target inspection points are generated using the point assessment model.

[0007] As a preferred embodiment of the automatic planning method for substation video inspection points based on multi-source data fusion described in this invention, the step of generating the target inspection point based on the M-1 historical candidate points, the historical coverage evaluation results corresponding to each historical candidate point, and the plurality of initial inspection candidate points includes: Obtain preset multi-round point optimization templates; Determine the current candidate point from the plurality of initial inspection candidate points. The current candidate point is the candidate point other than the M-1 historical candidate points from the plurality of initial inspection candidate points. Based on the multi-round point optimization template, the M-1 historical candidate points, the historical coverage evaluation results corresponding to each historical candidate point, and the current candidate point are integrated and processed to obtain the target inspection point.

[0008] As a preferred embodiment of the automatic planning method for substation video inspection points based on multi-source data fusion described in this invention, the step of performing M-1 point optimization processes based on the multiple initial candidate inspection points, the substation equipment status data, and the environmental monitoring data to determine M-1 historical candidate points and the historical coverage evaluation results corresponding to each historical candidate point includes: Based on the target inspection impact factor set, generate the first historical coverage assessment result corresponding to the first historical candidate point, wherein the first historical candidate point is any one of the plurality of initial inspection candidate points; Based on the j-th historical candidate point, the j-th historical coverage assessment result, and the multiple initial inspection candidate points, the (j+1)-th historical candidate point is generated, where j is an integer greater than or equal to 1. Based on the set of target inspection impact factors, generate the (j+1)th historical coverage assessment result corresponding to the (j+1)th historical candidate point, until j is M-2, and obtain the (M-1)th historical coverage assessment result.

[0009] As a preferred embodiment of the automatic planning method for substation video inspection points based on multi-source data fusion described in this invention, the coverage assessment result includes a coverage score and a coverage analysis text; the target inspection influencing factor set, the target inspection points, and the coverage assessment result are integrated and processed to obtain the inspection point planning data corresponding to the substation, including: Based on the target inspection point, the coverage score, and the coverage analysis text, the target point object is determined, and the coverage analysis text is filtered text. In the target inspection point, multiple historical point objects are obtained, including historical candidate points, historical coverage score values, and historical coverage analysis text; The target inspection impact factor set, the target location object, and the multiple historical location objects are integrated and processed to obtain the inspection location planning data.

[0010] As a preferred embodiment of the automatic planning method for substation video inspection points based on multi-source data fusion described in this invention, wherein: the fusion processing of the substation equipment status data and environmental monitoring data to obtain a set of target inspection influencing factors includes: The substation equipment status data is subjected to image enhancement processing to obtain enhanced equipment status image data. The image enhancement processing adopts a nonlinear adaptive enhancement algorithm based on a second Taylor series. Feature extraction is performed on the enhanced device status image data to obtain device feature data; The equipment feature data and the environmental monitoring data are normalized and fused to obtain the set of target inspection influencing factors.

[0011] As a preferred embodiment of the automatic planning method for substation video inspection points based on multi-source data fusion described in this invention, the step of normalizing and fusing the equipment feature data and the environmental monitoring data to obtain the target inspection influencing factor set includes: Obtain preset weight configuration information, which includes device status weight coefficient and environmental factor weight coefficient; Based on the weight configuration information, the equipment feature data and the environmental monitoring data are weighted and fused to obtain the target inspection impact factor set.

[0012] Secondly, the present invention provides an automatic planning system for substation video inspection points based on multi-source data fusion, comprising: The data acquisition module is used to acquire substation equipment status data and environmental monitoring data to be processed. The substation equipment status data includes equipment temperature anomaly information and appearance anomaly information obtained based on video image recognition. The environmental monitoring data includes humidity, temperature and wind speed. The data fusion module is used to fuse the substation equipment status data and environmental monitoring data to obtain a set of target inspection influencing factors. The candidate point generation module is used to generate multiple initial inspection candidate points based on the target inspection impact factor set; The site optimization module is used to perform multiple site optimization processes based on the multiple initial inspection candidate sites, the substation equipment status data, and the environmental monitoring data to obtain the target inspection site and the coverage assessment result corresponding to the target inspection site. The target inspection site includes multiple historical candidate sites before the target inspection site and the historical coverage assessment result corresponding to each historical candidate site. The planning data generation module is used to integrate and process the target inspection impact factor set, the target inspection points, and the coverage assessment results to obtain the inspection point planning data corresponding to the substation. The inspection point planning data is used to configure the camera deployment parameters of the video inspection system.

[0013] Thirdly, the present invention provides an electronic device including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0014] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0015] Compared with existing technologies, the beneficial effect of this invention is that it proposes an automatic planning method for substation video inspection points based on multi-source data fusion. By acquiring substation equipment status data and environmental monitoring data, it fuses and generates a set of target inspection influencing factors; based on this set, it generates multiple initial candidate inspection points; and simultaneously performs multiple rounds of point optimization processing combined with equipment status and environmental data to obtain the target inspection points and their coverage evaluation results; finally, it integrates and generates inspection point planning data for configuring camera deployment parameters. This invention achieves dynamic planning of inspection points by setting up a multi-source data-driven and iterative optimization mechanism, enabling video surveillance to fully cover the entire substation area. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 The present invention provides a flowchart of an automatic planning method for substation video inspection points based on multi-source data fusion, which is an embodiment of the present invention.

[0018] Figure 2 This is an internal structure diagram of an electronic device for an automatic planning method of substation video inspection points based on multi-source data fusion, provided as an embodiment of the present invention. Detailed Implementation

[0019] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0020] It should be noted in advance that the system mentioned in the embodiments as the subject of real-time operation refers to any system configured with this method.

[0021] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides an automatic planning method for substation video inspection points based on multi-source data fusion, including: This invention provides a method that can effectively solve the problems mentioned above. The following will describe in detail how to implement the automatic planning method for substation video inspection points based on multi-source data fusion using multiple embodiments. Figure 1 A flowchart illustrating an automatic planning method for substation video inspection points based on multi-source data fusion is shown, including: S1, acquire the substation equipment status data and environmental monitoring data to be processed. The substation equipment status data includes abnormal equipment temperature and abnormal appearance information obtained based on video image recognition. The environmental monitoring data includes humidity, temperature and wind speed. It should be noted that existing technologies primarily rely on historical inspection records and on-site survey experience to manually deploy camera locations. This method typically uses fixed-point presets or periodic manual re-surveys to achieve monitoring coverage, and cannot dynamically respond to anomalies caused by sudden equipment failures or weather changes.

[0022] For example, in a scenario where localized overheating occurs in a high-voltage switchgear area but does not trigger an alarm threshold, existing technology may fail to detect the potential hazard in this scenario during multiple rounds of inspections because the location is fixed and no high-priority viewpoint is assigned to this area.

[0023] Alternatively, in the event of strong winds and heavy rain, if the video surveillance points are not adjusted to take into account the impact of the current wind speed and humidity on the external insulation of the equipment, the camera may overlook the bushings or terminals that are most severely affected by wind and rain erosion.

[0024] The aforementioned abnormal equipment temperature information refers to the phenomenon that the surface temperature rise of the equipment deviates from the normal operating range, which is identified through infrared thermal imaging or visible light image analysis. Abnormal appearance information refers to visible defects such as damage to the equipment casing, rust, dirt accumulation, or missing parts.

[0025] In addition, some existing technical solutions attempt to introduce sensor data to assist video inspection, but these solutions only use equipment status data and environmental monitoring data separately for alarm judgment or weather recording, without establishing a fusion and correlation mechanism between the two.

[0026] Therefore, in practice, the existing technical solutions in this part not only fail to generate comprehensive impact factors for video location optimization, but may also lead to misjudgment due to data fragmentation.

[0027] In this embodiment of the invention, the solution simultaneously collects equipment temperature anomaly information and appearance anomaly information based on video image recognition output, and accesses humidity, temperature and wind speed data fed back by environmental sensors deployed in the substation in real time. Then, the two types of heterogeneous data are uniformly input into the subsequent fusion processing flow, which can ultimately provide a set of target inspection influencing factors with physical meaning and risk weight for automatic planning of inspection points.

[0028] The equipment status data obtained by this solution comes from the visible light and infrared dual-mode cameras already deployed in the station. The visible light and infrared dual-mode cameras are used to perform pixel-level anomaly detection on the appearance and thermal distribution of the equipment.

[0029] Environmental monitoring data is provided by multi-point integrated temperature, humidity and wind speed sensors installed in the main control room, outdoor power distribution area and near transformers. This setup ensures that the acquired environmental monitoring data covers all key areas of the station.

[0030] It should be noted that the above data acquisition process is completed entirely on local edge computing nodes, without the need to upload the original video or sensor data to a remote platform, thus avoiding data privacy leaks.

[0031] S2, integrates substation equipment status data and environmental monitoring data to obtain a set of target inspection influencing factors; In some embodiments, the fusion processing of substation equipment status data and environmental monitoring data mentioned in step S2 to obtain a set of target inspection influencing factors includes: S21, perform image enhancement processing on the substation equipment status data to obtain enhanced equipment status image data; In some embodiments, the image enhancement processing employs a nonlinear adaptive enhancement algorithm based on a second-order Taylor series, which may specifically include the following steps: Step 2.1: Extract visible light and infrared fused image frames containing abnormal device temperature or appearance information from the original video stream, and perform noise reduction preprocessing on the images to remove high-frequency artifacts caused by camera sensor thermal noise, transmission interference, or rain and fog weather. It should be noted that the noise reduction intensity threshold here can be set according to the actual needs of relevant technical personnel based on the lighting conditions at the substation site and the signal-to-noise ratio index of the imaging equipment; this invention does not impose any limitations on it. Step 2.2: Divide the preprocessed image into multiple local regions, calculate the mean gray level and gradient magnitude of each region, and construct a local brightness response function based on the second Taylor series expansion. This function takes the first and second derivatives of the gray level of the pixel neighborhood as input variables to approximate the local contrast change trend. It should be noted that the size of the neighborhood window here can be set according to the actual needs of relevant technicians by matching the complexity of the device surface texture with the image resolution; this invention does not limit this. Step 2.3: Dynamically adjust the gain coefficient of each region according to the local brightness response function. Apply high gain to low-light areas to improve detail visibility, apply compression gain to bright areas to suppress overexposure, and at the same time preserve the sharpness of edge structures to avoid false edges or halo effects caused by global enhancement. It should be noted that the upper and lower limits of gain here can be set according to the actual needs of relevant technical personnel based on the reflective characteristics of the equipment nameplate and the intensity distribution of background stray light; this invention does not impose any limitations on them. Step 2.4 involves seamlessly stitching together the images of each region after nonlinear gain adjustment and performing color consistency correction to finally output enhanced equipment status image data. The equipment status image data output at this location will serve as the input source for subsequent anomaly localization and risk assessment models.

[0032] S22, extract features from the enhanced equipment status image data to obtain equipment feature data; It should be pointed out that feature extraction methods based on a single scale or fixed template are difficult to simultaneously address the identification needs of both local minor defects and overall structural anomalies. Moreover, key anomaly information such as temperature rise areas, crack directions, and rust patches on the equipment surface often appear in images as low-contrast, small areas, or irregular shapes. If traditional edge detection or color histogram methods are still used for feature extraction, a large number of early fault features will be ignored, which may lead to a lack of input information for subsequent risk assessment models.

[0033] Therefore, this invention proposes a feature extraction mechanism that combines multi-scale fusion with semantic guidance.

[0034] Specifically, the enhanced equipment status image data can be processed in three stages: local texture gradient calculation, thermal distribution contour extraction, and component geometric topology modeling. First, the improved local binary mode operator is used to extract the micron-level texture changes on the surface of the equipment shell. This improved local binary mode operator can be used to identify rust, dirt, or coating peeling.

[0035] Furthermore, thermal contour maps are generated based on the temperature gradient field of the infrared channel.

[0036] Furthermore, by combining the pre-set three-dimensional structural prior model of the equipment, the boundaries of physical components in the visible light image are spatially aligned with the thermal anomaly area to construct a multi-dimensional feature vector with component attribution labels.

[0037] If the same device has multiple abnormal regions with similar characteristic response strengths, then feature weighting needs to be performed based on the electrical importance level of the device in the power grid topology.

[0038] For example, the main transformer bushing and the station service transformer terminal may show similar temperature rise characteristics at the same time, but because the main transformer undertakes the core power transmission and transformation function of the entire station, its abnormal characteristic weight will be set to 1.8 times that of the station service transformer.

[0039] In some embodiments, the device feature data mentioned herein may include five dimensions: anomaly type, spatial location, morphological parameters, thermal intensity, and device importance weight.

[0040] Among them, the aforementioned local texture gradient refers to the high-frequency image derivative response that reflects the microscopic morphological changes on the surface of the device; the thermal distribution profile refers to the closed region of isotherms generated based on the infrared image temperature gradient; and the component geometric topology modeling refers to the spatial alignment process of mapping image pixel coordinates to the structural framework of standard components of the device.

[0041] In some embodiments, the step of extracting micron-level texture variations on the surface of the device housing using an improved local binary mode operator can be implemented using a multi-level objective function modeling method.

[0042] Specifically, a multi-layer objective function system can be constructed, with each layer's objective function corresponding to a different optimization objective, and each optimization objective being equipped with matching constraints. The resulting multi-layer objective function system can effectively extract texture features.

[0043] In practice, the hierarchical structure of the objective function in the multi-layer objective function system can be determined based on the illumination uniformity and noise level of the local area in the enhanced device status image.

[0044] For example, when the signal-to-noise ratio in a local area of ​​the image is higher than a preset threshold and the illumination gradient is less than a set slope, a two-layer objective function structure can be used, where: The upper-level objective function is to maximize the local contrast of the texture response; The lower-level objective function is to minimize the pseudo-texture response caused by rain, fog, reflection, or sensor thermal drift.

[0045] More specifically, the concrete expression of the aforementioned upper-level objective function can be represented as follows: in, This represents the local region of the image currently being processed. The center pixel within the region. For The set of all sampling points within the annular neighborhood centered on the sample. Total number of neighboring points and These are the grayscale values ​​of pixels p and q, respectively. The Euclidean distance between the two points is... This is a distance weighting function derived based on a second-order Taylor expansion.

[0046] It should be noted that the upper objective function amplifies the grayscale perturbation caused by micron-level surface undulations through weighted difference operations, thereby improving the detectability of early defects such as cracks or spalling.

[0047] Furthermore, the specific expression for the lower-level objective function mentioned above can be as follows: in, This is the second Laplacian derivative of the image at pixel p, used to reflect local curvature changes; This is an ideal, defect-free surface texture template generated based on prior knowledge of the equipment's material properties. For the weight function Smoothing regularization terms; This is the regularization coefficient.

[0048] It should be noted that the lower-level objective function effectively eliminates false texture responses caused by environmental interference by constraining the consistency between the extracted results and the ideal surface model.

[0049] Furthermore, when the signal-to-noise ratio of a local area of ​​the image is lower than a preset threshold, or the illumination gradient exceeds a set slope, only the optimization process of the lower-level objective function needs to be executed.

[0050] It should be noted that the preset thresholds, set slopes, and corresponding constraints mentioned in this section can be set according to the actual needs of relevant technical personnel, combined with the substation field camera model, installation angle, and imaging quality statistics under typical weather conditions. This invention does not impose any limitations on these settings.

[0051] S23, normalize and merge equipment characteristic data with environmental monitoring data to obtain a set of target inspection influencing factors.

[0052] In some embodiments, step S23 involves normalizing and fusing equipment feature data with environmental monitoring data to obtain a set of target inspection influencing factors, including: S231, Obtain preset weight configuration information, including device status weight coefficient and environmental factor weight coefficient; The equipment status weight coefficient mentioned in step S231 refers to a numerical parameter used to quantify the decision-making weight of abnormal equipment temperature information and abnormal appearance information in the comprehensive risk assessment.

[0053] In this invention, the equipment status weighting coefficient is used to reflect the degree to which the health status of the primary equipment dominates the overall failure probability.

[0054] The environmental factor weighting coefficient mentioned in step S231 refers to a numerical parameter used to quantify the impact of three types of environmental monitoring data—humidity, temperature, and wind speed—on the operational risks of equipment.

[0055] In this invention, the environmental factor weighting coefficient is used to characterize the additional risk contribution of external meteorological conditions to the insulation performance, heat dissipation efficiency and mechanical stability of equipment.

[0056] In some embodiments, the equipment status weight coefficient and environmental factor weight coefficient can be directly read from the weight configuration table pre-stored in the local edge computing node of the substation.

[0057] In some embodiments, a secure communication link can be established with the main control room operation and maintenance management system to dynamically download the latest version of weight configuration information from the centralized policy server.

[0058] Furthermore, in certain scenarios, the numerical allocation of the two types of weighting coefficients can be automatically adjusted by receiving risk control instructions from the dispatch center and combining them with the current power grid operation mode. The aforementioned current power grid operation mode can include heavy load operation, maintenance isolation, or emergency power supply modes.

[0059] The aforementioned weight configuration information is stored in the form of structured key-value pairs and managed according to a unified version number.

[0060] S232, based on the weight configuration information, the equipment feature data and environmental monitoring data are weighted and fused to obtain the set of target inspection impact factors.

[0061] It is worth noting that the operational risk of substation equipment is determined by both the equipment's own deterioration status and external environmental stress. If the two types of data are fused using only equal weighting or a fixed ratio set by human experience, it will be impossible to dynamically reflect the changes in the dominant factors of equipment failure mechanisms under different meteorological conditions. This will lead to underestimating the risk of insulation flashover caused by condensation in high temperature and high humidity environments, or ignoring the accelerating effect of wind-induced vibration on loosening of connectors under strong wind conditions.

[0062] At this point, an adaptive weighted fusion mechanism based on preset weight configuration information can be adopted to quantitatively couple the abnormal intensity index in the equipment feature data with the meteorological stress index in the environmental monitoring data according to electrical importance and scene sensitivity, providing influencing factor input for the priority ranking of target inspection points.

[0063] During implementation, the structured feature vectors of each anomaly can first be extracted from the equipment feature data. These structured feature vectors can include anomaly type identifiers, spatial location coordinates, morphological parameters, thermal intensity, and equipment electrical importance weights.

[0064] Simultaneously, the humidity, temperature, and wind speed values ​​at the current moment are obtained from environmental monitoring data, and these values ​​are mapped to a predefined environmental stress level range.

[0065] Furthermore, based on the equipment state weight coefficient and environmental factor weight coefficient obtained in step S231, the environmental stress levels of different dimensions of the equipment feature vector can be linearly weighted respectively.

[0066] Furthermore, by eliminating dimensional differences through normalization, a set of multidimensional target inspection impact factors under a unified scale is finally obtained. Each element in this set of multidimensional target inspection impact factors corresponds to a component of equipment to be inspected, and the numerical values ​​of the elements comprehensively characterize the overall risk level of the component under the current environmental conditions.

[0067] It should be noted that each influencing factor in this invention includes two fields: equipment component identification and fusion risk score.

[0068] S3, generate multiple initial inspection candidate points based on the target inspection impact factor set; Currently, video surveillance locations are primarily determined through preset fixed camera angles or static coverage strategies based on equipment records. These methods typically rely on human experience to delineate key areas or allocate locations by configuring a global priority list, making it difficult to accurately focus on high-risk components.

[0069] Taking the scenario of a localized temperature rise in the high-voltage switchgear area, but which does not trigger the traditional alarm threshold as an example, when microcracks appear at the bushing root in this area and are accompanied by weak heat accumulation, the existing inspection system will still take pictures of the circuit breaker body according to the regular cycle because the location is fixed and it does not perceive the potential risk of this combination of abnormalities. Ultimately, this will lead to the continuous failure to detect key early defects in multiple rounds of inspections.

[0070] In scenarios involving strong winds and heavy rain, if the inspection points are not replanned based on the impact of current wind speed and humidity on the external insulation of the equipment, the cameras may continue to focus on low-risk metal support areas, while ignoring the insulator skirts or terminal parts that are most severely affected by wind and rain, ultimately leading to a delay in fault warning.

[0071] Furthermore, although some existing solutions attempt to introduce risk scoring mechanisms to guide site selection, these solutions simply score equipment status and environmental data independently and then perform a simple ranking, without establishing a physical coupling relationship between the two.

[0072] In the embodiments of the present invention, the fusion risk score corresponding to each equipment component in the target inspection impact factor set is first analyzed.

[0073] Furthermore, high-risk scoring components are mapped to their precise spatial coordinates in the 3D device model.

[0074] Furthermore, by combining the camera's field of view, occlusion relationships, and focal length constraints, a set of initial candidate inspection points that meet the requirements of visibility and resolution is generated.

[0075] Furthermore, a closed-loop mapping from risk quantification to visual scheduling is ultimately achieved.

[0076] Among them, the high-risk scoring components mentioned above may include key primary equipment connection parts such as main transformer bushings, disconnector contacts, and surge arrester flanges.

[0077] It should be noted that the initial inspection candidate points generated by this solution can be automatically sorted according to the fusion risk score, and can exclude invisible points caused by equipment structure obstruction or installation blind spots, thereby effectively avoiding the defects existing in the prior art.

[0078] The aforementioned initial inspection candidate points can be understood as a list of locations to be observed that are generated by the target inspection impact factor set and have spatial coordinates and priority labels. Each candidate point in the initial inspection candidate points is associated with an equipment component identifier, three-dimensional spatial coordinates, expected imaging resolution, and risk priority value.

[0079] S4. Based on multiple initial inspection candidate points, substation equipment status data and environmental monitoring data, multiple point optimization processes are performed to obtain the target inspection point and the coverage assessment results corresponding to the target inspection point. The target inspection point includes multiple historical candidate points before the target inspection point and the historical coverage assessment results corresponding to each historical candidate point. In some embodiments, step S4 involves multiple point optimization processes based on multiple initial candidate inspection points, substation equipment status data, and environmental monitoring data to obtain the target inspection point and the corresponding coverage assessment results, including: S41, determine the optimization round M corresponding to multiple point optimization processes, where M is an integer greater than 1; In step S41, the optimization round M refers to the number of iterative optimization operations set after the initial inspection candidate points are generated, in order to gradually eliminate problems such as field-of-view conflicts or motion path redundancy between points.

[0080] For example, the value of optimization round M can be read directly from the pre-configured optimization strategy configuration file in the substation edge computing node.

[0081] It can also communicate with the main control room operation and maintenance management system and dynamically set the M value according to the urgency level of the current inspection task. For example, M can be set to 2 in routine inspections and to 4 in emergency inspections triggered by typhoon warnings or equipment alarms.

[0082] In addition, in some scenarios, the M value can be determined by monitoring the rate of change of the comprehensive score of the set of points after each round of optimization in real time. For example, when the score improvement of two consecutive rounds of optimization is lower than the preset convergence threshold, the subsequent rounds are terminated in advance and the current M value is locked.

[0083] It should be noted that the preset convergence threshold mentioned in this section can be set according to the actual needs of relevant technical personnel and in combination with the point optimization benefit decay curve in historical inspection tasks, and this invention does not limit it.

[0084] It should also be noted that the value of the optimization round M is always an integer greater than 1. This setting ensures that at least one initial point conflict resolution operation or one imaging quality verification is completed.

[0085] S42, based on multiple initial inspection candidate points, substation equipment status data and environmental monitoring data, perform M-1 point optimization processes to determine M-1 historical candidate points and the historical coverage assessment results corresponding to each historical candidate point; S43. Based on M-1 historical candidate points, the historical coverage assessment results corresponding to each historical candidate point, and multiple initial inspection candidate points, generate target inspection points; S44 generates coverage assessment results corresponding to the target inspection points through the point assessment model.

[0086] In some embodiments, the steps for establishing the site assessment model are as follows: A1 receives the target inspection point, the corresponding coverage score, and the coverage analysis text, and initializes the input buffer of the point evaluation model. A2. Determine whether the coverage score is greater than or equal to the preset effective coverage threshold. If it is true, proceed to step A3; otherwise, proceed to step A7. The preset effective coverage threshold can be set according to the actual needs of relevant technical personnel through the statistical relationship between imaging quality and defect detection rate in historical inspection tasks. This invention does not limit it. A3. Perform keyword matching on the covered analysis text to determine whether it contains three positive semantic tags: "unobstructed", "key components are fully visible", and "clear edges". If all of them exist, proceed to step A4; otherwise, proceed to step A7. A4. Extract the equipment status weight coefficient and environmental factor weight coefficient from the target inspection impact factor set corresponding to the point, and calculate the weighted risk intensity index. If the index is greater than the preset high-risk threshold, proceed to step A5; otherwise, proceed to step A6. The preset high-risk threshold can be set according to the actual needs of relevant technical personnel through the equipment failure history database and risk level classification standards. This invention does not limit it. A5. Mark the location as a high-priority target location object and record its spatial coordinates, equipment component identification, coverage score, coverage analysis text and weighted risk intensity index, then proceed to step A8. A6. Mark the location as a regular priority target location object, and record its spatial coordinates, equipment component identification, coverage score, coverage analysis text and weighted risk intensity index, then proceed to step A8. A7. Mark this location as an invalid location and exclude it from the subsequent planning data generation process. Proceed to step A8. A8. Determine whether all target inspection points have been processed. If not, return to step A1 to continue processing the next point. If all points have been processed, the point evaluation model construction process is complete.

[0087] In some embodiments, step S42 involves performing M-1 point optimization processes based on multiple initial inspection candidate points, substation equipment status data, and environmental monitoring data to determine M-1 historical candidate points and the historical coverage assessment results corresponding to each historical candidate point, including: S421, Based on the target inspection impact factor set, generate the first historical coverage assessment result corresponding to the first historical candidate point. The first historical candidate point is any one of multiple initial inspection candidate points. S422, Based on the j-th historical candidate point, the j-th historical coverage assessment result, and multiple initial inspection candidate points, generate the (j+1)-th historical candidate point, where j is an integer greater than or equal to 1; S423. Based on the set of impact factors of the target inspection, generate the (j+1)th historical coverage assessment result corresponding to the (j+1)th historical candidate point, until j is M-2, and obtain the (M-1)th historical coverage assessment result.

[0088] It should be noted that if the final inspection path is generated directly based on the initial inspection candidate points without multiple rounds of coverage assessment and point iterative correction, some high-risk equipment components may not be effectively observed because the overlap of camera fields of view, mechanical gimbal movement conflicts, or insufficient local imaging resolution are not taken into account.

[0089] That is, although microcracks and weak heat accumulation anomalies have been identified at the root of the main transformer bushing in the first round of site planning, if the actual imaging area of ​​the site is missing due to the obstruction of the adjacent surge arrester, and the system does not adjust the viewing angle offset through subsequent rounds of optimization, then directly locking the site for inspection will result in the missed detection of high-risk defects.

[0090] The aforementioned historical coverage assessment results can be understood as a quantitative score of the area of ​​equipment components that can be effectively observed and the imaging quality of a certain historical candidate point under the current camera parameter configuration. This quantitative score is determined by four indicators: the pixel ratio of the target area in the image, edge sharpness, illumination uniformity, and occlusion rate.

[0091] It should be noted that the historical candidate points refer to the intermediate inspection positions generated and recorded during the j-th round of optimization, which are used to participate in the next round of point generation; the (j+1)-th historical candidate point is a new candidate position after dynamic adjustment based on the spatial deviation and coverage shortcomings of the previous round of points.

[0092] In some embodiments, the generation of target inspection points based on M-1 historical candidate points, the historical coverage assessment results corresponding to each historical candidate point, and multiple initial inspection candidate points mentioned in step S43 includes: S431, Obtain the preset multi-round point optimization template; It should be noted that the multi-round point optimization template involved in step S431 refers to a set of predefined structured iterative rules in the substation video inspection task, which is used to achieve gradual convergence from the initial candidate points to the final inspection path.

[0093] In this invention, the multi-round site optimization template is configured as an operational framework that guides the mapping relationship between the generation of historical candidate sites and the feedback of coverage evaluation results in each round.

[0094] In practical applications, multi-round point optimization templates can be directly loaded through the policy storage area built into the local edge computing node of the substation.

[0095] Alternatively, by establishing a secure communication channel with the main control room's operation and maintenance management system, optimized template versions can be dynamically downloaded according to equipment type and weather scenario combination.

[0096] In addition, in some scenarios, the corresponding configuration can be retrieved and activated from the local template library by receiving the template identifier attached to the inspection task instruction issued by the dispatch center.

[0097] It should be noted that the above-mentioned multi-round site optimization templates are stored in the form of structured configuration files. Each template is clearly bound to a specific device family and a typical environmental stress mode, and is managed according to a unified version number. The typical environmental stress modes mentioned here can include high temperature and high humidity, strong wind and dust, and low temperature condensation.

[0098] S432, determine the current candidate point from multiple initial inspection candidate points. The current candidate point is the candidate point other than the M-1 historical candidate points from the multiple initial inspection candidate points. S433: Based on the multi-round point optimization template, integrate M-1 historical candidate points, the historical coverage assessment results corresponding to each historical candidate point, and the current candidate points to obtain the target inspection point.

[0099] It should be noted that if high-risk scoring points are directly selected from the initial inspection candidate points as the final inspection points without integrating the coverage feedback information accumulated during multiple rounds of optimization, the real impact of dynamic perspective adjustment on imaging effectiveness may be ignored. That is, although a significant temperature rise of the isolating switch contact has been identified in the first round of point planning, if the contact area is partially obscured by the metal bracket in the actual image due to the limited pitch angle of the pan-tilt unit, and the system does not combine the quantitative feedback of "effective pixel coverage rate is only 78%" from the previous M-1 rounds of historical coverage evaluation results to make corrections, if the point is still retained as the target inspection point, the thermal anomaly area will not be clearly captured, thereby weakening the confidence of fault identification.

[0100] The aforementioned current candidate points can be understood as supplementary observation locations selected from the remaining points in the initial candidate point set that have not yet participated in historical iterations during the Mth round of integration, in order to make up for the blind spots in spatial coverage of historical candidate points.

[0101] It should also be noted that the target inspection points refer to the final set of executable inspection locations generated after multiple rounds of template optimization, fusion of historical coverage evaluation feedback and current candidate points. Each target inspection point meets three hard constraints: no severe obstruction in the field of view, pixel coverage of key components not less than 90%, and edge sharpness score higher than a preset threshold. These target inspection points are used to drive the camera pan-tilt unit to perform high-precision automatic inspection tasks. S5 integrates and processes the target inspection impact factor set, target inspection points, and coverage assessment results to obtain the inspection point planning data corresponding to the substation. The inspection point planning data is used to configure the camera deployment parameters of the video inspection system.

[0102] In this embodiment of the invention, the coverage assessment result includes a coverage score and a coverage analysis text; In some embodiments, step S5 involves integrating the target inspection impact factor set, target inspection points, and coverage assessment results to obtain the substation corresponding inspection point planning data, including: S51. Based on the target inspection points, coverage score, and coverage analysis text, determine the target point objects. The coverage analysis text is the filtered text. S52, in the target inspection point, obtain multiple historical point objects, including historical candidate points, historical coverage score values ​​and historical coverage analysis text; S53 integrates and processes the set of target inspection influencing factors, target location objects, and multiple historical location objects to obtain inspection location planning data.

[0103] It should be noted that, in order to address the potential issues of fragmented decision-making basis for inspection points and untraceable evaluation results, it is necessary to clarify which points truly possess high effectiveness and feasibility in multiple rounds of optimization; that is, the target points should be identified first.

[0104] Once the target location is confirmed, a joint structured encapsulation of historical optimized trajectories and the current optimal solution can be carried out. This helps to reasonably address the problem of repeated scheduling or omission of key areas caused by the lack of contextual association during the inspection path generation process, and reduces the strategy fluctuations and resource waste caused by relying solely on the results of a single round of evaluation.

[0105] In some embodiments, target location selection can be achieved using a dual criterion based on coverage quality and semantic consistency.

[0106] For example, filtering rules can be used to select texts with coverage scores higher than a preset threshold and containing positive keywords such as "unobstructed", "clear edges" and "key components fully visible".

[0107] In practice, threshold screening can be performed on the coverage score value corresponding to each point in the target inspection points. For example, in the main transformer bushing inspection task, points with a coverage score value greater than 0.92 are retained, and low-scoring points where only part of the bushing root is in the picture due to the deviation of the pan-tilt angle are removed.

[0108] Based on this, keyword filtering processing is further implemented on the coverage analysis text. For example, continuing the aforementioned scenario, natural language rule matching is performed on the coverage analysis text of the retained points. Only when "the root of the casing is visible" and "no metal support obstruction" appear in the text at the same time will it be included in the candidate set of target points.

[0109] Subsequently, after the initial screening of target point objects is completed, the corresponding historical point objects are extracted from the M-1 round of optimization.

[0110] Ultimately, by structurally integrating the set of target inspection impact factors, target location objects, and multiple historical location objects, the planning data for inspection locations can be generated.

[0111] The aforementioned inspection point planning data can be understood as a final set of inspection instructions driven by target inspection impact factors, verified through multiple rounds of coverage, and structurally encapsulated.

[0112] It should be noted that the patrol point planning data in this invention can include six types of fields: spatial coordinates of each target point, equipment component identification, coverage score, coverage analysis text, historical optimization trajectory index, and environmental stress context.

[0113] Example 2, refer to Figure 2 This embodiment also provides an automatic planning system for substation video inspection points based on multi-source data fusion, including: The data acquisition module is used to acquire the substation equipment status data and environmental monitoring data to be processed. The substation equipment status data includes equipment temperature anomaly information and appearance anomaly information obtained based on video image recognition. The environmental monitoring data includes humidity, temperature and wind speed. The data fusion module is used to fuse substation equipment status data and environmental monitoring data to obtain a set of target inspection impact factors; The candidate site generation module is used to generate multiple initial inspection candidate sites based on the target inspection impact factor set; The site optimization module is used to perform multiple site optimization processes based on multiple initial inspection candidate sites, substation equipment status data, and environmental monitoring data to obtain the target inspection site and the coverage assessment results corresponding to the target inspection site. The target inspection site includes multiple historical candidate sites before the target inspection site and the historical coverage assessment results corresponding to each historical candidate site. The planning data generation module is used to integrate and process the target inspection impact factor set, target inspection points and coverage assessment results to obtain the inspection point planning data corresponding to the substation. The inspection point planning data is used to configure the camera deployment parameters of the video inspection system.

[0114] The above-mentioned unit modules can be embedded in the processor of the electronic device in hardware form or independent of it, or they can be stored in the memory of the electronic device in software form, so that the processor can call and execute the corresponding operations of the above modules.

[0115] This embodiment also provides an electronic device, which can be a terminal, and its internal structure diagram can be as follows. Figure 2As shown, the electronic device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements an automatic planning method for substation video inspection points based on multi-source data fusion. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the device's casing, or an external keyboard, touchpad, or mouse.

[0116] This embodiment also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, it performs the following steps: Acquire the substation equipment status data and environmental monitoring data to be processed. The substation equipment status data includes equipment temperature anomaly information and appearance anomaly information obtained based on video image recognition. The environmental monitoring data includes humidity, temperature and wind speed. The substation equipment status data and environmental monitoring data are fused and processed to obtain a set of target inspection influencing factors. Based on the set of target inspection impact factors, multiple initial inspection candidate locations are generated; Based on multiple initial inspection candidate points, substation equipment status data, and environmental monitoring data, the points are optimized multiple times to obtain the target inspection point and the coverage assessment results corresponding to the target inspection point. The target inspection point includes multiple historical candidate points before the target inspection point and the historical coverage assessment results corresponding to each historical candidate point. The target inspection impact factor set, target inspection points, and coverage assessment results are integrated and processed to obtain the inspection point planning data corresponding to the substation. The inspection point planning data is used to configure the camera deployment parameters of the video inspection system.

[0117] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0118] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0119] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for automatic planning of substation video inspection points based on multi-source data fusion, characterized in that, include: Acquire substation equipment status data and environmental monitoring data to be processed. The substation equipment status data includes equipment temperature anomaly information and appearance anomaly information obtained based on video image recognition. The environmental monitoring data includes humidity, temperature and wind speed. The substation equipment status data and environmental monitoring data are fused and processed to obtain a set of target inspection influencing factors; Based on the set of target inspection impact factors, multiple initial inspection candidate locations are generated; Based on the multiple initial inspection candidate points, the substation equipment status data, and the environmental monitoring data, multiple point optimization processes are performed to obtain the target inspection point and the coverage assessment result corresponding to the target inspection point. The target inspection point includes multiple historical candidate points before the target inspection point and the historical coverage assessment result corresponding to each historical candidate point. The target inspection impact factor set, the target inspection points, and the coverage assessment results are integrated and processed to obtain the inspection point planning data corresponding to the substation. The inspection point planning data is used to configure the camera deployment parameters of the video inspection system.

2. The automatic planning method for substation video inspection points based on multi-source data fusion as described in claim 1, characterized in that, The process of performing multiple site optimization processes based on the multiple initial candidate inspection sites, the substation equipment status data, and the environmental monitoring data to obtain the target inspection site and the coverage assessment result corresponding to the target inspection site includes: Determine the optimization round M corresponding to the multiple point optimization processes, where M is an integer greater than 1; Based on the multiple initial inspection candidate points, the substation equipment status data, and the environmental monitoring data, M-1 point optimization processes are performed to determine M-1 historical candidate points and the historical coverage assessment results corresponding to each historical candidate point; The target inspection point is generated based on the M-1 historical candidate points, the historical coverage assessment results corresponding to each historical candidate point, and the multiple initial inspection candidate points. The coverage assessment results corresponding to the target inspection points are generated using the point assessment model.

3. The automatic planning method for substation video inspection points based on multi-source data fusion as described in claim 2, characterized in that, The step of generating the target inspection point based on the M-1 historical candidate points, the historical coverage assessment results corresponding to each historical candidate point, and the multiple initial inspection candidate points includes: Obtain preset multi-round point optimization templates; Determine the current candidate point from the plurality of initial inspection candidate points. The current candidate point is the candidate point other than the M-1 historical candidate points from the plurality of initial inspection candidate points. Based on the multi-round point optimization template, the M-1 historical candidate points, the historical coverage evaluation results corresponding to each historical candidate point, and the current candidate point are integrated and processed to obtain the target inspection point.

4. The automatic planning method for substation video inspection points based on multi-source data fusion as described in claim 3, characterized in that, The process of performing M-1 point optimization processes based on the multiple initial inspection candidate points, the substation equipment status data, and the environmental monitoring data to determine M-1 historical candidate points and the historical coverage assessment results corresponding to each historical candidate point includes: Based on the target inspection impact factor set, generate the first historical coverage assessment result corresponding to the first historical candidate point, wherein the first historical candidate point is any one of the plurality of initial inspection candidate points; Based on the j-th historical candidate point, the j-th historical coverage assessment result, and the multiple initial inspection candidate points, the (j+1)-th historical candidate point is generated, where j is an integer greater than or equal to 1. Based on the set of target inspection impact factors, generate the (j+1)th historical coverage assessment result corresponding to the (j+1)th historical candidate point, until j is M-2, and obtain the (M-1)th historical coverage assessment result.

5. The automatic planning method for substation video inspection points based on multi-source data fusion as described in claim 4, characterized in that, The coverage assessment results include a coverage score and a coverage analysis text; the target inspection impact factor set, the target inspection locations, and the coverage assessment results are integrated and processed to obtain the inspection location planning data corresponding to the substation, including: Based on the target inspection point, the coverage score, and the coverage analysis text, the target point object is determined, and the coverage analysis text is filtered text. In the target inspection point, multiple historical point objects are obtained, including historical candidate points, historical coverage score values, and historical coverage analysis text; The target inspection impact factor set, the target location object, and the multiple historical location objects are integrated and processed to obtain the inspection location planning data.

6. The automatic planning method for substation video inspection points based on multi-source data fusion as described in claim 5, characterized in that, The process of fusing the substation equipment status data and environmental monitoring data yields a set of target inspection influencing factors, including: The substation equipment status data is subjected to image enhancement processing to obtain enhanced equipment status image data. The image enhancement processing adopts a nonlinear adaptive enhancement algorithm based on a second Taylor series. Feature extraction is performed on the enhanced device status image data to obtain device feature data; The equipment feature data and the environmental monitoring data are normalized and fused to obtain the set of target inspection influencing factors.

7. The automatic planning method for substation video inspection points based on multi-source data fusion as described in claim 6, characterized in that, The step of normalizing and fusing the equipment feature data with the environmental monitoring data to obtain the target inspection impact factor set includes: Obtain preset weight configuration information, which includes device status weight coefficient and environmental factor weight coefficient; Based on the weight configuration information, the equipment feature data and the environmental monitoring data are weighted and fused to obtain the target inspection impact factor set.

8. An automatic planning system for substation video inspection points based on multi-source data fusion, using the method described in any one of claims 1 to 7, characterized in that, include: The data acquisition module is used to acquire substation equipment status data and environmental monitoring data to be processed. The substation equipment status data includes equipment temperature anomaly information and appearance anomaly information obtained based on video image recognition. The environmental monitoring data includes humidity, temperature and wind speed. The data fusion module is used to fuse the substation equipment status data and environmental monitoring data to obtain a set of target inspection influencing factors. The candidate point generation module is used to generate multiple initial inspection candidate points based on the target inspection impact factor set; The site optimization module is used to perform multiple site optimization processes based on the multiple initial inspection candidate sites, the substation equipment status data, and the environmental monitoring data to obtain the target inspection site and the coverage assessment result corresponding to the target inspection site. The target inspection site includes multiple historical candidate sites before the target inspection site and the historical coverage assessment result corresponding to each historical candidate site. The planning data generation module is used to integrate and process the target inspection impact factor set, the target inspection points, and the coverage assessment results to obtain the inspection point planning data corresponding to the substation. The inspection point planning data is used to configure the camera deployment parameters of the video inspection system.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the automatic planning method for substation video inspection points based on multi-source data fusion as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the automatic planning method for substation video inspection points based on multi-source data fusion as described in any one of claims 1 to 7.

Images