A method, device and equipment for automatically updating a power grid enterprise knowledge graph

Abstract

The application discloses an automatic updating method, device and equipment for a power grid enterprise knowledge graph, and belongs to the technical field of knowledge graph updating, and comprises the following steps: identifying an exception and triggering an inspection signal, issuing exception information to a mobile inspection terminal, generating initial inspection information to control an inspection equipment to collect an image set, extracting a key inspection image set based on image quality and target detection parameters, obtaining an inspection analysis result, obtaining a knowledge graph analysis label, and updating the knowledge graph in combination with the inspection analysis result. The application triggers an inspection signal by identifying an exception, screens key inspection images of unmanned equipment, reduces data transmission volume, intelligently selects an uploading mode according to a network state, reduces network pressure while guaranteeing data integrity, improves transmission efficiency, selects a comparative analysis mode according to the integrity score of historical data, accurately judges an updating instruction, and improves the accuracy and timeliness of power system operation and maintenance decision-making.

Description

Technical Field

[0001] This invention relates to the field of knowledge graph updating technology, and in particular to an automatic updating method, apparatus and equipment for a power grid enterprise knowledge graph. Background Technology

[0002] With the continuous development of the power industry and the increasing level of informatization, power companies have accumulated a large amount of data. Knowledge graphs, as a technological means to effectively represent and manage knowledge, provide power companies with a comprehensive and intuitive knowledge system by graphically representing various entities and concepts in the power field and their relationships. This helps improve the operational efficiency of the power system and ensure the safe and stable operation of the power grid. However, the knowledge of power companies is dynamic and time-sensitive. The topology of the power grid changes constantly with the commissioning of new equipment, the decommissioning of old equipment, and network upgrades. Therefore, the knowledge graph of power companies needs to be updated and maintained in a timely manner to ensure its accuracy and completeness.

[0003] For example, the invention patent announcement CN117171364B discloses a method and apparatus for updating an operation and maintenance knowledge graph. The method involves receiving an update request for the knowledge graph, which includes operation and maintenance text and an update type of the knowledge graph. Based on the update type and the schema information of the corresponding schema layer of the knowledge graph, a target prompt set is determined. Then, based on the target prompt set, a large model is called to parse the operation and maintenance text, determine the first triplet contained in the operation and maintenance text, and then use the first triplet to update the knowledge graph.

[0004] For example, the invention patent announcement CN110019823B describes a method and apparatus for updating a knowledge graph. The method mainly includes: setting a change cycle for the attributes of entities in the data layer of the knowledge graph; determining the next update time of the attribute value corresponding to the attribute based on the current time and the change cycle of the attribute; and updating the corresponding attribute value in the data layer according to the next update time of the attribute value.

[0005] Among the aforementioned prior art, patents with publication number CN117171364B and CN110019823B have at least the following technical problems:

[0006] Current methods for updating and maintaining power knowledge graphs focus solely on data processing and analysis, neglecting the complex network environments in remote areas and the massive amounts of data generated by drone inspections. In remote areas, insufficient base station coverage, severe signal attenuation, low network bandwidth, high latency, and frequent interruptions prevent the knowledge graph from reflecting changes in the status of the target objects in a timely manner, severely impacting operational and maintenance decisions and reducing the efficiency and reliability of power system operations and maintenance.

[0007] In summary, traditional power grid knowledge graph update technology has significant drawbacks: reliance on manual inspection triggers delays, leading to sluggish anomaly responses; full transmission of original images causes bandwidth congestion in weak network areas; static historical data comparison mechanisms can lead to misjudgments or missed updates when data is incomplete; rigid data processing procedures lacking network adaptability exacerbate the risk of transmission interruptions; and cloud-based centralized processing models are ill-suited to the real-time requirements of mobile edge scenarios. Summary of the Invention

[0008] This invention aims to solve the problems of inefficient data transmission and delayed updates in the power grid knowledge graph under weak network conditions in the prior art, and provides an automatic update method, device and equipment for the knowledge graph of power grid enterprises.

[0009] The technical solution of the present invention:

[0010] The first aspect of this invention provides a method for automatically updating a knowledge graph of a power grid enterprise, comprising the following steps:

[0011] Identify anomalies and trigger inspection signals, then send anomaly information to the mobile inspection terminal;

[0012] Initial inspection information is generated, which is used to control the inspection equipment to perform data acquisition of the image set; the inspection equipment is a drone.

[0013] A key inspection image set is obtained, which is determined based on image quality parameters and target detection parameters.

[0014] Obtain inspection analysis results, which are determined based on the inspection information analysis tags obtained from the key inspection image set;

[0015] Obtain knowledge graph analysis tags and update the knowledge graph based on the inspection analysis results.

[0016] As a further optional solution to the automatic update method of the power grid enterprise knowledge graph, the identification of anomalies specifically includes: collecting real-time data through sensors deployed on the target detection object, analyzing the deviation between the real-time data and historical data using a data analysis model, and identifying anomalies and triggering inspection signals when the deviation exceeds a preset threshold.

[0017] As a further optional solution to the automatic update method of the power grid enterprise knowledge graph, the initial inspection information is generated based on a local line map library, which includes static data, dynamic data, and geographic information; the static data includes the power grid topology, the model of the inspection object, and its installation parameters; the dynamic data includes historical inspection records, status update logs, and maintenance records; and the geographic information includes the geographic coordinates and terrain elevation data of the inspection object.

[0018] As a further optional solution to the automatic update method of the power grid enterprise knowledge graph, the target detection parameters include the target location percentage and the target location coverage.

[0019] As a further optional solution to the automatic update method of the power grid enterprise knowledge graph, a key inspection image set is obtained. This key inspection image set is determined based on image quality parameters and target detection parameters extracted from the image set, specifically including:

[0020] A set of qualified images is selected based on image quality parameters, and a criticality score of the qualified image set is calculated based on target detection parameters. The set of images with criticality scores greater than a preset threshold is selected as the key inspection image set.

[0021] The criticality score is used to measure the value of the image set for the state analysis of the target detection object and the updating of the knowledge graph. The criticality score is determined by comprehensively considering the degree of comparison between the target detection parameters and the reference parameters and the influence of weight coupling.

[0022] It should be noted that the specific analysis process for obtaining the key inspection image set is as follows:

[0023] The mobile inspection terminal controls the drone to perform flight inspections and acquires a set of drone inspection images.

[0024] Collect image quality parameters from drone inspection images, and then extract a set of qualified inspection images.

[0025] Collect target detection parameters from qualified inspection images, including the proportion of target areas and the coverage of target areas in the qualified inspection images;

[0026] The key score of the qualified inspection image is obtained by analyzing and processing the target detection parameters of the qualified inspection image;

[0027] The criticality score of the qualified inspection image is used to measure the value of the qualified inspection image for the state analysis of the target detection object and the updating of the knowledge graph. Specifically, it is obtained by considering the proportion of the target part and the coverage of the target part in the qualified inspection image to reflect the richness of the key information of the image, and by comprehensively considering the comparison between the target detection parameters and the corresponding parameters and introducing the coupling influence of the weights of each parameter.

[0028] Extract the preset key scoring thresholds from the database;

[0029] Qualified inspection images with criticality scores greater than the criticality score threshold are recorded as critical inspection images;

[0030] The set of qualified inspection images is traversed sequentially to filter and extract the set of key inspection images.

[0031] As a further optional scheme for the automatic update method of the power grid enterprise knowledge graph, the image quality parameters include sharpness, contrast, exposure and noise level; the image quality index is calculated through the image quality parameters, and the combination of images with quality index greater than a preset threshold is selected as a qualified image set; the image quality index is determined by comprehensively considering the image detail presentation, color level, brightness and interference, combined with the degree of deviation of each parameter from the reference value and the influence of weight coupling.

[0032] It should be noted that the specific analysis steps for extracting the set of qualified inspection images are as follows:

[0033] Collect image quality parameters for drone inspections, including image sharpness, contrast, exposure, and noise level;

[0034] The quality index of UAV inspection images is obtained based on the analysis and processing of UAV inspection image quality parameters.

[0035] The UAV inspection image quality index is used to evaluate the quality of images captured by UAVs and their applicability to subsequent target detection analysis. Specifically, it considers the image's performance in detail presentation, color level differentiation, brightness control, and interference. Based on the image's clarity, noise level, and the degree of comparison with corresponding reference values, as well as the deviation of the image's contrast and exposure from corresponding ideal values, the coupling influence of the weights of each parameter is introduced to finally obtain the UAV inspection image quality index.

[0036] Extract preset image quality index thresholds from the database;

[0037] Drone inspection images with image quality indicators greater than the image quality indicator threshold are recorded as qualified inspection images.

[0038] The drone inspection image set is traversed sequentially to filter and extract the qualified inspection image set.

[0039] As a further optional solution to the automatic update method of the power grid enterprise knowledge graph, the inspection information analysis tag includes an information transmission compression ratio and an inspection information upload mode.

[0040] Network status characteristic indicators are calculated based on network status parameters, and the information transmission compression ratio and inspection information upload mode are determined based on the network status characteristic indicators.

[0041] It should be further explained that the specific process for determining the inspection information analysis tag is as follows:

[0042] The inspection information includes a set of key inspection images and basic information on drone inspections.

[0043] The basic information for drone inspection includes drone inspection time information and drone inspection location information;

[0044] Obtain network status parameters, including network latency, signal strength, network jitter, and packet loss rate;

[0045] Network state characteristic indicators are obtained through analysis and processing of network state parameters;

[0046] The network state characteristic index is used to measure the impact of network state on the transmission effect of inspection information. Specifically, it is obtained by considering the characteristics of the network in terms of data transmission speed, signal stability, fluctuation and data loss, based on the degree of comparison between the network state parameters and the reference network state parameters, and by introducing the coupling effect of corresponding weights.

[0047] Extract the compression ratio corresponding to each state feature index interval stored in the database, and map the compression ratio corresponding to the interval where the extracted network state feature index is located, denoted as the information transmission compression ratio.

[0048] The inspection information upload mode is determined based on network status characteristic indicators;

[0049] The information transmission compression ratio and the inspection information upload mode are combined as the inspection information analysis tags.

[0050] As a further optional solution to the automatic update method of the power grid enterprise knowledge graph, the inspection information upload mode includes direct upload mode, preprocessing upload mode and cache upload mode;

[0051] Preset the first and second verification indicators for network status;

[0052] If the network status characteristic index is less than or equal to the first network status verification index, then upload directly;

[0053] If the network status characteristic index is greater than the first network status verification index and less than the second network status verification index, then preprocess and upload.

[0054] If the network state characteristic indicator is greater than or equal to the second network state verification indicator, then upload to the cache.

[0055] It should be noted that the specific process for determining the inspection information upload mode is as follows:

[0056] The inspection information upload modes include direct upload, preprocessed upload, and cached upload;

[0057] Extract the preset first network status verification indicator and second network status verification indicator from the database;

[0058] If the network status characteristic index is less than or equal to the first network status verification index, then the inspection information upload mode is recorded as direct upload.

[0059] If the network status characteristic index is greater than the first network status verification index and less than the second network status verification index, then the inspection information upload mode is recorded as preprocessing upload.

[0060] If the network status characteristic index is greater than or equal to the second network status verification index, then the inspection information upload mode is recorded as cache upload.

[0061] As a further optional solution to the automatic updating method of the power grid enterprise knowledge graph, the acquisition of inspection analysis results specifically includes:

[0062] If the inspection information upload mode is direct upload, the inspection information will be compressed according to the information transmission compression ratio and transmitted to the cloud processing center for analysis to obtain the inspection analysis results.

[0063] If the inspection information upload mode is preprocessing upload, the inspection information is processed by the mobile inspection terminal to obtain the inspection analysis results and then compressed according to the information transmission compression ratio before being uploaded to the cloud processing center.

[0064] If the inspection information upload mode is cached upload, the inspection information is processed by the mobile inspection terminal and the inspection analysis results are cached, waiting to be uploaded.

[0065] It should be noted that the specific analysis process for obtaining the inspection analysis results is as follows:

[0066] If the inspection information upload mode is direct upload, the inspection information will be compressed according to the information transmission compression ratio and transmitted to the cloud processing center. The inspection information will be analyzed in the cloud processing center to obtain the inspection analysis results.

[0067] If the inspection information upload mode is preprocessing upload, the inspection information will be uploaded to the mobile inspection terminal, the inspection information analysis will be completed on the mobile inspection terminal, the inspection analysis results will be obtained, and the inspection analysis results information will be transmitted to the cloud processing center with a compression ratio.

[0068] If the inspection information upload mode is cached upload, the inspection information will be uploaded to the mobile inspection terminal. The mobile inspection terminal will then perform inspection information analysis, obtain the inspection analysis results, cache them on the mobile inspection terminal, and wait for them to be uploaded.

[0069] As a further optional solution to the automatic updating method of the power grid enterprise knowledge graph, the acquisition of knowledge graph analysis tags specifically includes:

[0070] Extract historical comparison data corresponding to the inspection analysis results and calculate their integrity score;

[0071] The knowledge graph is analyzed based on the completeness score, and the knowledge graph is updated based on the knowledge graph analysis tags.

[0072] It should be further explained that the specific analysis process for obtaining knowledge graph analysis tags is as follows:

[0073] Based on the inspection analysis results, the corresponding historical data in the knowledge graph of the power grid enterprise is extracted and recorded as historical comparison data.

[0074] Obtain the integrity parameters of historical comparison data and analyze the integrity score of historical comparison data;

[0075] The integrity score of the historical comparison data is used to measure the completeness of the historical comparison data in the knowledge graph of the power grid enterprise. By analyzing the completeness of the historical data in terms of quantity, time range and coverage of key parameters, the comparison results of the integrity parameters and corresponding reference values ​​are combined, and the coupling effect of the corresponding weights is introduced to finally obtain the integrity score of the historical comparison data.

[0076] Knowledge graph analysis tags are used to determine the integrity score of historical comparison data.

[0077] The knowledge graph analysis tags include direct comparison analysis, similarity comparison analysis, and no comparison analysis required.

[0078] Extract the first integrity verification score and the second integrity verification score stored in the database;

[0079] If the integrity score of the historical comparison data is greater than the first integrity verification score, then the knowledge graph analysis label will be recorded as direct comparison analysis;

[0080] If the integrity score of the historical comparison data is less than or equal to the first integrity verification score and greater than or equal to the second integrity verification score, then the analysis label of the knowledge graph will be recorded as similarity comparison analysis.

[0081] If the integrity score of the historical comparison data is less than the second integrity verification score, the analysis label of the knowledge graph will be marked as no comparison analysis required.

[0082] Automatic updates of knowledge graphs are achieved by analyzing tags based on knowledge graphs.

[0083] As a further optional solution to the automatic updating method of the power grid enterprise knowledge graph, the updating of the knowledge graph based on knowledge graph analysis tags specifically includes:

[0084] If the knowledge graph analysis tag is direct comparison analysis, then the knowledge graph is updated based on the inspection analysis results and historical comparison data;

[0085] If the knowledge graph analysis tag is similarity comparison analysis, then the knowledge graph is updated based on the inspection analysis results, historical comparison data, and similarity comparison data.

[0086] If the knowledge graph analysis tag is "no comparative analysis required", then the knowledge graph is updated by adding new knowledge graph nodes.

[0087] It should be noted that the specific analysis process for updating the knowledge graph based on knowledge graph analysis tags is as follows:

[0088] If the knowledge graph analysis tag is direct comparison analysis, the knowledge graph will be automatically updated based on the inspection analysis results and historical comparison data;

[0089] If the knowledge graph analysis tag is similarity comparison analysis, similarity comparison data is extracted based on the power enterprise knowledge graph, and the knowledge graph is automatically updated based on the inspection analysis results, historical comparison data, and similar comparison data;

[0090] If the analysis tag of the knowledge graph is "no comparative analysis required", then add a knowledge graph node, thereby directly performing automatic updates to the knowledge graph.

[0091] A second aspect of the present invention provides an automatic updating device for a knowledge graph of a power grid enterprise, comprising:

[0092] The inspection triggering module is used to identify anomalies and trigger inspection signals, and send anomaly information to the mobile inspection terminal; it is also used to generate initial inspection information, which is used to control the inspection equipment to perform data acquisition of the image set.

[0093] The key image extraction module is used to obtain a set of key inspection images, which is determined based on image quality parameters and target detection parameters.

[0094] The inspection result analysis module is used to obtain inspection analysis results, which are determined based on the inspection information analysis tags obtained from the key inspection image set.

[0095] The knowledge graph update module is used to obtain knowledge graph analysis tags and update the knowledge graph in combination with the inspection analysis results.

[0096] A third aspect of the present invention provides an electronic device, comprising: a processor, and memory and a network interface connected to the processor; the network interface being connected to non-volatile memory in a server; the processor, during operation, retrieves a computer program from the non-volatile memory via the network interface and runs the computer program via the memory to apply the method described above.

[0097] One or more technical solutions provided in this invention have at least the following technical effects or advantages:

[0098] 1. This invention provides an automatic update method for the knowledge graph of a power grid enterprise. In remote areas with complex network environments and difficult data transmission, it triggers drone inspections from the production command center to accurately locate abnormal areas. Key inspection images are selected using image quality parameters and target detection parameters to reduce data transmission volume. The upload mode is intelligently selected based on network status, ensuring data integrity while reducing network pressure and improving transmission efficiency. In the knowledge graph update stage, an appropriate comparative analysis method is selected based on the integrity score of historical data to accurately determine update instructions. Whether it's status updates, node additions, or inspection information updates, they can be completed promptly and accurately, ensuring that the knowledge graph reflects the real-time status of the target detection objects. This improves the accuracy and timeliness of power system operation and maintenance decisions, enhancing the safety and stability of power grid operation.

[0099] 2. By extracting key inspection image sets, this invention can optimize the data foundation for updating and maintaining the knowledge graph of power grid enterprises. In the case of massive image data generated by UAV inspections, the invention filters images based on image quality parameters and target detection parameters, effectively eliminating a large number of images with poor quality or insufficient key information. This reduces the burden of subsequent data transmission and processing, improves data transmission efficiency, reduces network pressure, and ensures the accuracy of data used for knowledge graph update analysis.

[0100] 3. This invention obtains inspection analysis results by determining the tags of inspection information. It allows for flexible adjustment of data processing and transmission strategies based on actual network conditions, effectively improving the efficiency and accuracy of knowledge graph updates for power grid companies. When network conditions are good, the direct upload mode fully utilizes the superior network environment to quickly transmit inspection information to the cloud processing center for analysis, ensuring data timeliness and enabling timely updates to the knowledge graph based on the latest data. This helps maintenance personnel quickly grasp the real-time status of the monitored objects. When network conditions are average, the preprocessing upload mode leverages the local processing capabilities of the mobile inspection terminal to perform preliminary analysis of the inspection information, extracting key content before uploading. This reduces network transmission pressure while ensuring valuable information is delivered to the cloud in a timely manner, providing a reliable basis for knowledge graph updates. In cases of poor network conditions, the cached upload mode avoids data loss or corruption due to network problems. Information analysis is completed locally and the results are cached, then uploaded once the network improves, ensuring data integrity and providing an accurate data foundation for subsequent precise knowledge graph updates. Attached Figure Description

[0101] Figure 1 A flowchart illustrating an automatic updating method for a power grid enterprise knowledge graph, provided as an embodiment of this application;

[0102] Figure 2 A schematic diagram of the structure of an automatic updating device for a knowledge graph of a power grid enterprise provided in this application embodiment;

[0103] Figure 3 This is a schematic diagram of the composition of an electronic device provided in an embodiment of this application. Detailed Implementation

[0104] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0105] Reference Figure 1 As shown, the first aspect of this invention provides a method for automatically updating a knowledge graph of a power grid enterprise, comprising the following steps:

[0106] S1 identifies anomalies and triggers inspection signals, sending anomaly information to the mobile inspection terminal.

[0107] S2, Generate initial inspection information, which is used to control the inspection equipment to perform data acquisition of the image set.

[0108] In this embodiment, the production command center identifies anomalies that trigger inspection signals and sends anomaly information to the mobile inspection terminal. The production command center then calls the local route map library to generate initial three-dimensional inspection information, thereby controlling the drone to perform inspections.

[0109] It should be noted that, in a specific embodiment, the production command center identifies anomalies and triggers inspection signals by: collecting data in real time through sensors deployed on the target detection object; using random forest to analyze the deviation between historical data and real-time data; and identifying anomalies if the deviation exceeds a preset deviation threshold in the database, thereby triggering an inspection signal.

[0110] The local line map library contains static data, dynamic data, and geographic information. The static data includes, but is not limited to, power grid topology, inspection object model, and installation parameters. The installation parameters include, but are not limited to, tower height and conductor specifications. The dynamic data includes, but is not limited to, historical inspection records, inspection object status update logs, and maintenance records. The geographic information includes, but is not limited to, the geographic coordinates and terrain elevation data of the inspection object.

[0111] In this embodiment, the abnormal information includes abnormal area location information and monitoring abnormal data.

[0112] In one specific embodiment, "when the production command center detects that the insulator temperature of a certain power tower (detection object ID: T012) exceeds the threshold continuously within a preset time window, an abnormal signal is triggered, and abnormal information containing the power tower coordinates (latitude X, longitude Y) and temperature data is generated."

[0113] The initial three-dimensional inspection information includes the location of the power tower, the distance to the line, and environmental markers.

[0114] An initial 3D inspection map is generated based on the initial 3D inspection information.

[0115] The method for generating the initial 3D inspection map is as follows: Based on the location information and geographical information of the detected objects in the local line map library, a 3D modeling tool is used to generate an initial 3D inspection map containing tower locations, line distances, and environmental markers. The tower locations include the tower coordinates and numbers in the abnormal area. The line distances include the conductor lengths and sag parameters between adjacent towers. The environmental markers are the distribution of obstacles in the abnormal area. The obstacles include, but are not limited to, trees, buildings, and other objects that may affect the UAV's flight inspection.

[0116] In a specific embodiment, controlling the drone to perform inspections involves generating a drone flight path based on an initial 3D inspection map using a path algorithm. The drone flight path is required to cover abnormal areas and avoid obstacles. Drone inspection control is then performed based on the drone flight path.

[0117] S3, Obtain a set of key inspection images, which is determined based on image quality parameters and target detection parameters.

[0118] In this embodiment, a set of key inspection images is obtained, and the specific analysis process is as follows:

[0119] The mobile inspection terminal controls the drone to conduct flight inspections and acquires a set of drone inspection images.

[0120] Collect image quality parameters from drone inspection images, and then extract a set of qualified inspection images.

[0121] In this embodiment, the set of qualified inspection images is extracted, and the specific analysis steps are as follows:

[0122] Collect image quality parameters for drone inspections, including image sharpness, contrast, exposure, and noise level.

[0123] In a specific embodiment, the image quality parameters of the UAV inspection can be obtained through analysis using OpenCV (a cross-platform computer vision library).

[0124] The quality index of UAV inspection images is obtained by analyzing and processing the quality parameters of UAV inspection images.

[0125] The UAV inspection image quality index is used to evaluate the quality of images captured by UAVs and their applicability to subsequent target detection analysis. Specifically, it considers the image's performance in detail presentation, color level differentiation, brightness control, and interference. Based on the image's clarity, noise level, and the degree of comparison with corresponding reference values, as well as the deviation of the image's contrast and exposure from corresponding ideal values, the coupling influence of the weights of each parameter is introduced to finally obtain the UAV inspection image quality index.

[0126] It should be noted that the image quality index for UAV inspection, obtained through analysis and processing of image quality parameters, takes into account the interrelationships between these parameters. For example, inappropriate exposure, such as overexposure or underexposure, can lead to loss of image detail, affecting sharpness and blurring the edges of normally sharp objects. Simultaneously, abnormal exposure alters image contrast; overexposed areas exhibit reduced contrast, while underexposed areas may fail to effectively display contrast. High noise levels interfere with image details, reduce sharpness, and disrupt the original contrast and color balance, impacting both contrast and exposure performance.

[0127] Extract reference image quality parameters stored in the database, including reference sharpness, ideal contrast, ideal exposure, and reference noise level.

[0128] The system extracts preset sharpness weights, contrast weights, exposure weights, and noise level weights from the database. It's important to understand that these weights range from 0 to 1. The extraction method involves constructing mapping sets between the image's sharpness, contrast, exposure, and noise level and their corresponding weights. When using the system, the sharpness, contrast, exposure, and noise level of the real-time acquired image are input into these mapping sets to extract the corresponding sharpness weights, contrast weights, exposure weights, and noise level weights.

[0129] In a specific embodiment, the image quality indicators for drone inspection are specifically described as follows:

[0130] ,

[0131] in, For the quality indicators of drone inspection images. For image clarity, For image contrast, For the image's exposure, The noise level of the image. For reference image sharpness, For ideal image contrast, For the ideal exposure of the image, As a reference noise level for the image, For sharpness weights, For contrast weights, For exposure weighting, denoted by , where is the noise level weight and e is the natural constant.

[0132] Extract preset image quality index thresholds from the database.

[0133] Drone inspection images with image quality indicators greater than the image quality indicator threshold are recorded as qualified inspection images.

[0134] It is important to understand that the drone inspection image quality index represents the quality of the image captured by the drone. The higher the drone inspection image quality index, the better the image quality, the easier the subsequent analysis, and the higher the accuracy. The image quality index threshold is the critical indicator for acceptable image quality. If the image quality index of a drone inspection image is greater than the image quality index threshold, it means that the image can be further analyzed and processed in terms of image quality.

[0135] The drone inspection image set is traversed sequentially to filter and extract the qualified inspection image set.

[0136] Collect target detection parameters from qualified inspection images, including the proportion of target areas and the coverage of target areas in the qualified inspection images.

[0137] It should be noted that, in specific embodiments, the target detection parameters can be obtained through analysis using the SSD (Single Shot MultiBox Detector) algorithm.

[0138] The target part proportion refers to the proportion of pixels occupied by the target detection object in the image, and the target part coverage refers to the proportion of the area of ​​the target detection object in the image to the total area of ​​the target detection object.

[0139] Extract the reference target detection parameters stored in the database, including the reference target area percentage and the reference target area coverage.

[0140] The criticality score of the qualified inspection image is obtained by analyzing and processing the target detection parameters of the qualified inspection image.

[0141] The criticality score of the qualified inspection image is used to measure the value of the qualified inspection image for the state analysis of the target detection object and the updating of the knowledge graph. Specifically, it is obtained by considering the proportion and coverage of the target part in the qualified inspection image to reflect the richness of key information of the image, and by comprehensively considering the comparison between the target detection parameters and the corresponding parameters and introducing the coupling influence of the weights of each parameter.

[0142] In a specific embodiment, the key score of a qualified inspection image is represented as follows:

[0143] ,

[0144] in, The key score for qualified inspection images, The percentage of target areas in qualified inspection images. To ensure the coverage of the target area in the qualified inspection image. For reference, the proportion of the target area For reference, the coverage of the target area, The weighting is based on the proportion of the target area. The weight of the coverage of the target area.

[0145] It should be noted that the target part proportion weight and target part coverage weight are both between 0 and 1. When using them, the pre-set values ​​can be directly extracted from the database. For example, the target part proportion and target part coverage are mapped one by one with the corresponding target part proportion weight and target part coverage weight. When using them, the target part proportion and target part coverage obtained in real time are input into the mapping set to extract the corresponding target part proportion weight and target part coverage weight.

[0146] Extract the preset key scoring thresholds from the database.

[0147] Qualified inspection images with a criticality score greater than the criticality score threshold are recorded as critical inspection images.

[0148] It's important to understand that a higher criticality score indicates that the image contains richer key information about the detected object, providing more accurate data for subsequent analysis. The criticality score threshold is the critical standard for determining whether a qualified inspection image can become a critical inspection image. If the criticality score of a qualified inspection image is greater than the criticality score threshold, it means that the image can provide data for subsequent analysis.

[0149] The set of qualified inspection images is traversed sequentially to filter and extract the set of key inspection images.

[0150] S4, Obtain the inspection analysis results, which are determined based on the inspection information analysis tags obtained from the key inspection image set.

[0151] In this embodiment, the process for determining the inspection information analysis tag is as follows:

[0152] The inspection information includes a set of key inspection images and basic information on drone inspections.

[0153] The basic information for drone inspection includes drone inspection time information and drone inspection location information.

[0154] The inspection information analysis tag includes the information transmission compression ratio and the inspection information upload mode.

[0155] The inspection information upload modes include direct upload, preprocessed upload, and cached upload.

[0156] Obtain network status parameters, including network latency, signal strength, network jitter, and packet loss rate.

[0157] The network latency, network jitter, and packet loss rate can be obtained using network testing tools, while the signal strength can be obtained using the signal strength detection module built into the mobile inspection terminal.

[0158] Extract the reference network state parameters stored in the database, including reference network latency, reference signal strength, reference network jitter, and reference packet loss rate.

[0159] Network state characteristic indicators are obtained by analyzing and processing network state parameters.

[0160] The network state characteristic index is used to measure the impact of network state on the transmission effect of inspection information. Specifically, it is obtained by considering the characteristics of the network in terms of data transmission speed, signal stability, fluctuation, and data loss, based on the degree of comparison between the network state parameters and the reference network state parameters, and by introducing the coupling effect of corresponding weights.

[0161] In specific embodiments, network state characteristic indicators are obtained through analysis and processing of network state parameters. This is because the interrelationships between network state parameters are taken into account. For example, signal strength directly affects network latency and packet loss rate. When the signal strength is weak, data transmission is easily interfered with, leading to a significant increase in network latency and a marked increase in the probability of packet loss. Network jitter and network latency are mutually influential; the more severe the network jitter, the greater the fluctuation in network latency, and the longer the average latency may become. Network jitter also exacerbates packet loss because an unstable network environment makes it difficult for data transmission to proceed stably. Furthermore, an increase in the packet loss rate indirectly affects network latency, as some data loss necessitates retransmission, further increasing network latency.

[0162] In a specific embodiment, the network state characteristic indicators are represented as follows:

[0163] ,

[0164] in, As a network state characteristic indicator, For network latency duration, For signal strength, For network jitter, For packet loss rate, For reference network latency, For reference signal strength, To reference network jitter, For reference packet loss rate, The network latency duration is the weight. For signal strength weights, As a weight for network jitter, is the packet loss rate weight, and e is the natural constant.

[0165] It should be noted that the weights for network latency, signal strength, network jitter, and packet loss rate all range from 0 to 1. These pre-defined values ​​can be directly extracted from the database. Specifically, the extraction method involves constructing mapping sets for each of the network latency, signal strength, network jitter, and packet loss rate, one by one. During use, the real-time acquired network latency, signal strength, network jitter, and packet loss rate are input into these mapping sets to extract the corresponding weights.

[0166] Extract the compression ratio corresponding to each state feature index interval stored in the database, and map the compression ratio corresponding to the interval where the network state feature index is located, denoted as the information transmission compression ratio.

[0167] It's important to understand that network status characteristic indicators are quantitative metrics used to measure network status. A higher network status characteristic indicator indicates a worse network condition and more obstacles to data transmission. To minimize data transmission volume, reduce network pressure, and improve transmission efficiency while ensuring data integrity and analyzability, a higher network status characteristic indicator corresponds to a higher information compression ratio. This allows for dynamic adjustments to the transmission strategy of inspection information based on the actual network conditions.

[0168] Extract the preset first network status verification indicator and second network status verification indicator from the database.

[0169] If the network status characteristic index is less than or equal to the first network status verification index, then the inspection information upload mode is recorded as direct upload.

[0170] If the network status characteristic index is less than or equal to the first verification index of network status, it indicates that the current network status is good, and the risk of data loss, damage, or delay during transmission is low. In this case, the mode of directly uploading inspection information is the most suitable. The inspection information can be directly transmitted to the cloud processing center at a certain information transmission compression ratio. This can make full use of the good network conditions, allowing the cloud to quickly obtain data and complete analysis, maximizing the timeliness of inspection information, and enabling the power grid company's knowledge graph to be updated in a timely manner based on the latest data.

[0171] If the network status characteristic index is greater than the first network status verification index but less than the second network status verification index, then the inspection information upload mode is recorded as preprocessing upload.

[0172] If the network status characteristic index is greater than the first verification index but less than the second verification index, it indicates that the current network status is average. Although the network can still maintain data transmission, there are some unstable factors, and data transmission may experience slight delays and a small amount of packet loss. In this case, the inspection information is first uploaded to the mobile inspection terminal. The local processing capabilities of the mobile inspection terminal are used to perform preliminary analysis of the information, extract key content, and then the processed results are transmitted to the cloud processing center at a compressed ratio. This reduces network transmission pressure, avoids problems caused by large-scale data transmission, and ensures that valuable information is delivered to the cloud in a timely manner, providing a basis for knowledge graph updates.

[0173] If the network status characteristic index is greater than or equal to the second network status verification index, then the inspection information upload mode is recorded as cache upload.

[0174] If the network status characteristic index is greater than or equal to the second verification index of network status, it indicates that the current network status is poor. In this case, directly uploading inspection information is highly likely to result in data loss or corruption during transmission, severely impacting subsequent analysis and knowledge graph updates. Therefore, it is necessary to choose cached upload. After uploading the inspection information to the mobile inspection terminal, the information analysis should be completed locally and the results cached. The information should then be uploaded to the cloud processing center once the network status improves. This ensures that the inspection information is not lost due to network problems, guarantees data integrity, and provides a basis for accurately updating the power grid company's knowledge graph.

[0175] In this embodiment, the inspection analysis results are obtained as follows:

[0176] If the inspection information is uploaded directly, the inspection information will be compressed according to the information transmission compression ratio and transmitted to the cloud processing center. The inspection information will be analyzed in the cloud processing center to obtain the inspection analysis results.

[0177] It should be noted that, in specific embodiments, the cloud processing center has built-in target detection and recognition algorithms, such as the YOLO (a real-time target detection algorithm) series of algorithms, which can quickly detect and recognize target objects (such as power towers, insulators, etc.) in inspection images and determine the location and status of the detected objects.

[0178] If the inspection information upload mode is preprocessing upload, the inspection information will be uploaded to the mobile inspection terminal, the inspection information analysis will be completed on the mobile inspection terminal, the inspection analysis results will be obtained, and the inspection analysis results information will be transmitted to the cloud processing center with a compression ratio.

[0179] It should be noted that, in specific embodiments, the mobile inspection terminal has built-in PP-PicoDet (PaddlePaddlePicoDet, a lightweight target detection algorithm), which can analyze and process inspection images locally.

[0180] If the inspection information upload mode is cached upload, the inspection information will be uploaded to the mobile inspection terminal. The mobile inspection terminal will then perform inspection information analysis, obtain the inspection analysis results, cache them on the mobile inspection terminal, and wait for them to be uploaded.

[0181] In a specific embodiment, if the inspection information analysis is completed on the mobile inspection terminal, the key inspection image corresponding to the inspection analysis result also needs to be cropped to obtain a key inspection image that only contains the target detection object, and the image and the inspection analysis result are uploaded to the cloud simultaneously.

[0182] S5, obtain knowledge graph analysis tags, and update the knowledge graph based on the inspection analysis results.

[0183] In this embodiment, knowledge graph analysis tags are obtained, and the knowledge graph is updated in conjunction with the inspection analysis results. The specific analysis process is as follows:

[0184] Based on the inspection analysis results, corresponding historical data are extracted from the knowledge graph of the power grid enterprise and recorded as historical comparison data.

[0185] Obtain the integrity parameters of historical comparison data and analyze the integrity score of historical comparison data.

[0186] The integrity score of the historical comparison data is used to measure the completeness of the historical comparison data in the knowledge graph of the power grid enterprise. By analyzing the completeness of the historical data in terms of quantity, time range and coverage of key parameters, the comparison results of the integrity parameters and corresponding reference values ​​are combined, and the coupling effect of the corresponding weights is introduced to finally obtain the integrity score of the historical comparison data.

[0187] It should be noted that analyzing the integrity score of historical comparative data based on its integrity parameters takes into account the interrelationships between these parameters. For example, data coverage affects the presentation of time span and key parameter integrity. Low data coverage, with a large amount of missing key data, can lead to discontinuities in key information within a time span, significantly compromising the integrity of key parameters and failing to fully reflect the true state of the target object at different times. Conversely, a short time span may not cover the various operational states of the target object under complex working conditions, resulting in insufficient data coverage and potentially affecting the integrity of key parameters, making it impossible to obtain complete key parameters of the target object. Missing key parameter integrity affects the effective assessment of data coverage and makes it impossible to accurately determine the value of data within a time span, rendering data coverage and time span based on incomplete key parameters meaningless.

[0188] In a specific embodiment, the integrity score of historical comparison data is obtained using the following method:

[0189] Obtain the integrity parameters of historical comparison data, including data coverage, time span, and integrity of key parameters.

[0190] The data coverage refers to the ratio of the amount of historical data already existing in the current knowledge graph to the amount of data that should theoretically be stored. The time span refers to the ratio of the historical data storage period to the theoretically required storage period. The key parameter completeness refers to the ratio of the actual number of key parameters contained in the knowledge graph to the total number of key parameters required by the detection object.

[0191] It should be noted that the key parameters refer to parameters in the knowledge graph that are closely related to the inspection analysis results and play an important role in determining the status of the target detection object and determining the knowledge graph update instructions.

[0192] Extract the reference integrity parameters stored in the database, including reference data coverage, reference time span, and reference key parameter integrity.

[0193] Extract the preset data coverage weight, time span weight, and key parameter integrity weight from the database.

[0194] The completeness score for historical comparison data is represented as follows:

[0195] ,

[0196] in, Score the completeness of historical comparison data. For historical comparison data coverage, For the time span of historical comparison data, To ensure the integrity of key parameters in historical comparison data, For reference data coverage, For reference time span, For the sake of completeness of key parameters, For data coverage weight, Weighted by time span This is the integrity weight for key parameters.

[0197] In this embodiment, knowledge graph analysis tags are determined based on historical data of the power grid enterprise's knowledge graph. The automatic updating of the power grid enterprise's knowledge graph is then completed based on the inspection analysis results and the knowledge graph analysis tags. The specific analysis process is as follows:

[0198] Based on the inspection analysis results, corresponding historical data are extracted from the knowledge graph of the power grid enterprise and recorded as historical comparison data.

[0199] Obtain the integrity parameters of historical comparison data and analyze the integrity score of historical comparison data.

[0200] The integrity score of the historical comparison data is used to measure the completeness of the historical comparison data in the knowledge graph of the power grid enterprise. By analyzing the completeness of the historical data in terms of quantity, time range and key parameters, and taking into account the coupling effect of these parameters, the integrity score of the historical comparison data is finally obtained.

[0201] Knowledge graph analysis tags are used to determine the integrity score based on historical comparison data.

[0202] The knowledge graph analysis tags include direct comparison analysis, similarity comparison analysis, and no comparison analysis required.

[0203] Extract the first integrity verification score and the second integrity verification score stored in the database.

[0204] If the integrity score of the historical comparison data is greater than the first integrity verification score, then the knowledge graph analysis label will be recorded as direct comparison analysis.

[0205] If the integrity score of the historical comparison data is greater than the first integrity verification score, it indicates that the currently acquired historical comparison data has extremely high integrity. This historical data can provide a comprehensive and reliable reference for the inspection and analysis results. Therefore, at this time, the knowledge graph analysis label is marked as direct comparison analysis. Based on the inspection and analysis results and historical comparison data, the knowledge graph update instructions can be analyzed quickly and accurately, efficiently completing the knowledge graph update work of the power grid enterprise and ensuring that the knowledge graph can reflect the true state of the target detection object in a timely manner.

[0206] If the integrity score of the historical comparison data is less than or equal to the first integrity verification score and greater than or equal to the second integrity verification score, then the analysis label of the knowledge graph will be recorded as similarity comparison analysis.

[0207] If the integrity score of historical comparison data is less than or equal to the first integrity verification score but greater than or equal to the second integrity verification score, it indicates that while the historical comparison data has a certain degree of integrity, it has not yet reached an ideal state. In this case, relying solely on historical comparison data and inspection analysis results to directly analyze knowledge graph update instructions may not be accurate enough. Therefore, it is necessary to extract similar comparison data based on the power grid enterprise's knowledge graph. By integrating inspection analysis results, existing historical comparison data, and newly extracted similar comparison data, knowledge graph update instructions can be comprehensively analyzed from multiple perspectives. Under the existing historical data conditions, this approach maximizes the extraction of useful information, ensuring the accuracy and reliability of knowledge graph updates, and enabling the knowledge graph to better serve the power grid enterprise's operation and maintenance decisions.

[0208] If the integrity score of the historical comparison data is less than the second integrity verification score, the analysis label of the knowledge graph will be marked as no comparison analysis required.

[0209] If the integrity score of historical comparison data is lower than the second integrity verification score, it indicates that the integrity of historical data is poor, making it difficult to accurately determine the update status of the knowledge graph by comparing it with the inspection analysis results. The inspection analysis results may include information on newly appearing detection objects, status changes, etc.

[0210] Automatic updates of knowledge graphs are achieved by analyzing tags based on knowledge graphs.

[0211] In this embodiment, the knowledge graph is automatically updated based on knowledge graph analysis tags. The specific analysis process is as follows:

[0212] If the knowledge graph analysis tag is direct comparison analysis, the knowledge graph will be automatically updated based on the inspection analysis results and historical comparison data.

[0213] If the knowledge graph analysis tag is similarity comparison analysis, similarity comparison data is extracted based on the power grid enterprise's knowledge graph, and the knowledge graph is automatically updated based on the inspection analysis results, historical comparison data, and similarity comparison data.

[0214] In a specific embodiment, the cloud processing center can use the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering algorithm to group data points with similar characteristics together. During direct comparison analysis, DBSCAN can cluster the inspection analysis results and historical comparison data. During similarity comparison analysis, DBSCAN can cluster the inspection analysis results, historical comparison data, and similar comparison data. By observing the distribution and changes in clusters, the state change trend of the target detection object can be understood. If the clustering results show that the target detection object's state data gradually clusters towards an abnormal area, a corresponding abnormal result is generated. At this time, the cloud processing center will output a state update command to update the target detection object's state and simultaneously output an inspection record command to update the inspection record. If the clustering results show that the target detection object's state data is in a normal area, a normal result is generated. At this time, the cloud processing center only outputs an inspection record command to update the inspection record.

[0215] It is important to note that if the cloud processing center outputs a status update command, the current status of the target detection object in the knowledge graph will be updated based on the output results of the cloud processing center through direct comparison analysis or similar comparison analysis. In a specific embodiment, taking the external parameters of a wire as an example, if the output result of the cloud processing center after analysis and processing is wire damage, wire wear location, and wire wear degree, and there are no other abnormalities, when the knowledge graph receives the status update command, it will update the status of the wire to wear, and at the same time update the wire wear location and wire wear degree, and simultaneously add the possible cause of the abnormality identified by the production command center: wire wear.

[0216] If the cloud processing center outputs an inspection record update instruction, the inspection record of the target detection object in the knowledge graph will be updated, that is, a new inspection record will be added. The inspection record includes the inspection time, inspection reason and inspection result. At the same time, the production command center will identify possible causes of abnormalities: exclude abnormalities caused by abnormal external physical structure of the target detection object.

[0217] If the analysis tag of the knowledge graph is "no comparative analysis required", then add a knowledge graph node, thereby directly performing automatic updates to the knowledge graph.

[0218] If the analysis tag of the knowledge graph is "no comparative analysis required," directly adding new knowledge graph nodes can quickly incorporate this new information into the knowledge graph, ensuring that the knowledge graph can reflect the latest situation of the target detection object in a timely manner and avoid information lag due to insufficient data. For example, if an inspection discovers a new target detection object or a state change that has not occurred before, but historical data is missing and cannot provide a reference, directly adding a node can record these new discoveries, providing accurate basis for subsequent operation and maintenance, decision-making, etc., and ensuring the real-time performance and usability of the power grid company's knowledge graph.

[0219] If the analysis label of the knowledge graph is "no comparative analysis required", adding a knowledge graph node means adding a new associated node of the target detection object to the knowledge graph and inputting the inspection analysis results of the target detection object, the current inspection record and the corresponding key inspection image set.

[0220] It should be noted that if the cloud processing center does not receive the corresponding key inspection image set, it will send a transmission instruction to the mobile inspection terminal and wait for the network status to support it before uploading the key inspection image set to the cloud processing center, thereby inputting the image data into the newly added associated node of the target detection object in the knowledge graph.

[0221] See Figure 2 As shown, a second aspect of the present invention provides an automatic updating device for a power grid enterprise knowledge graph, comprising:

[0222] The inspection trigger module is used by the production command center to identify abnormalities and trigger inspection signals, and to send abnormal information to the mobile inspection terminal; it is also used to generate initial inspection information, which is used to control the inspection equipment to perform data acquisition of the image set.

[0223] The key image extraction module is used to obtain a set of key inspection images, which is determined based on image quality parameters and target detection parameters.

[0224] The inspection result analysis module is used to obtain inspection analysis results, which are determined based on the inspection information analysis tags obtained from the key inspection image set.

[0225] The knowledge graph update module is used to obtain knowledge graph analysis tags and update the knowledge graph in combination with the inspection analysis results.

[0226] See Figure 3As shown, a third aspect of the present invention provides an electronic device, comprising: a processor, and memory and a network interface connected to the processor; the network interface being connected to non-volatile memory in a server; the processor, during operation, retrieves a computer program from the non-volatile memory via the network interface and runs the computer program via the memory to apply the method described above.

[0227] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0228] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0229] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0230] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0231] 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.

[0232] 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 automatically updating a knowledge graph of a power grid enterprise, characterized in that, Includes the following steps: Identify anomalies and trigger inspection signals, then send anomaly information to the mobile inspection terminal; Generate initial inspection information, which is used to control the inspection equipment to perform data acquisition of the image set; A key inspection image set is obtained, which is determined based on image quality parameters and target detection parameters. The inspection analysis results are obtained, which are determined based on the inspection information analysis tags obtained from the key inspection image set and network status parameters. The inspection information analysis tags include the information transmission compression ratio and the inspection information upload mode. Network status characteristic indicators are calculated based on the network status parameters, and the information transmission compression ratio and the inspection information upload mode are determined based on the network status characteristic indicators. The network status parameters include network latency, signal strength, network jitter, and packet loss rate. Obtain knowledge graph analysis tags and update the knowledge graph based on the inspection analysis results; The acquisition of knowledge graph analysis tags specifically includes: Extract historical comparison data corresponding to the inspection analysis results and calculate their integrity score; The knowledge graph is analyzed based on the completeness score, and the knowledge graph is updated based on the knowledge graph analysis tags. The determination of knowledge graph analysis tags based on completeness scores specifically includes: If the integrity score of the historical comparison data is greater than the first integrity verification score, then the knowledge graph analysis label will be recorded as direct comparison analysis; If the integrity score of the historical comparison data is less than or equal to the first integrity verification score and greater than or equal to the second integrity verification score, then the analysis label of the knowledge graph will be recorded as similarity comparison analysis. If the integrity score of the historical comparison data is less than the second integrity verification score, the analysis label of the knowledge graph will be marked as no comparison analysis required. The method of updating the knowledge graph based on knowledge graph analysis tags specifically includes: If the knowledge graph analysis tag is direct comparison analysis, then the knowledge graph is updated based on the inspection analysis results and historical comparison data; If the knowledge graph analysis tag is similarity comparison analysis, then the knowledge graph is updated based on the inspection analysis results, historical comparison data, and similarity comparison data. If the knowledge graph analysis tag is "no comparative analysis required", then the knowledge graph is updated by adding new knowledge graph nodes.

2. The automatic updating method for the knowledge graph of power grid enterprises as described in claim 1, characterized in that: A key inspection image set is obtained, which is determined based on image quality parameters and target detection parameters, and includes: A set of qualified images is selected based on image quality parameters, and a criticality score of the qualified image set is calculated based on target detection parameters. The set of images with criticality scores greater than a preset threshold is selected as the key inspection image set. The criticality score is used to measure the value of the image set for the state analysis of the target detection object and the updating of the knowledge graph. The criticality score is determined by comprehensively considering the degree of comparison between the target detection parameters and the reference parameters and the influence of weight coupling.

3. The automatic updating method for the knowledge graph of power grid enterprises as described in claim 2, characterized in that: The image quality parameters include sharpness, contrast, exposure, and noise level; image quality indices are calculated using these parameters, and image combinations with quality indices greater than a preset threshold are selected as a qualified image set.

4. The automatic updating method for the knowledge graph of power grid enterprises as described in claim 1, characterized in that: The inspection information upload modes include direct upload mode, preprocessing upload mode, and cache upload mode.

5. The automatic updating method for the knowledge graph of power grid enterprises as described in claim 4, characterized in that: The acquisition of inspection analysis results specifically includes: If the inspection information upload mode is direct upload, the inspection information will be compressed according to the information transmission compression ratio and transmitted to the cloud processing center for analysis to obtain the inspection analysis results. If the inspection information upload mode is preprocessing upload, the inspection information is processed by the mobile inspection terminal to obtain the inspection analysis results and then compressed according to the information transmission compression ratio before being uploaded to the cloud processing center. If the inspection information upload mode is cached upload, the inspection information is processed by the mobile inspection terminal and the inspection analysis results are cached, waiting to be uploaded.

6. An apparatus for applying the automatic updating method for a power grid enterprise knowledge graph as described in any one of claims 1-5, characterized in that, include: The inspection triggering module is used to identify anomalies and trigger inspection signals, and send anomaly information to the mobile inspection terminal. It is also used to generate initial inspection information, which is used to control the inspection equipment to perform data acquisition of the image set; A key image extraction module is used to obtain a set of key inspection images, which is determined based on image quality parameters and target detection parameters. The inspection result analysis module is used to obtain inspection analysis results, which are determined based on the inspection information analysis tags obtained from the key inspection image set. The knowledge graph update module is used to obtain knowledge graph analysis tags and update the knowledge graph in combination with the inspection analysis results.

7. An electronic device, characterized in that, include: A processor, and memory and a network interface connected to the processor; The network interface is connected to the non-volatile memory in the server; When the processor runs, it retrieves a computer program from the non-volatile memory via the network interface and runs the computer program via the memory to perform the method as described in any one of claims 1-5.

Images