A patrol inspection method and system combining power operation scene characteristics

Abstract

The present application provides an inspection method and system combined with power operation scenario characteristics, which relates to the field of power safety technology. The method includes: obtaining power operation scenario characteristics; determining multi-type operation scenario characteristic clusters; obtaining inspection characteristics; determining inspection plans; obtaining abnormal inspection information; sending the abnormal inspection information to a twin digital platform as a training task to perform abnormal inspection task training; feeding back abnormal inspection parameters that have reached convergence results in the training to the inspection platform, and updating the parameters of the inspection plan. The method solves the technical problem in the prior art that the inspection plan has poor scene adaptability due to insufficient analysis of power operation scenario characteristics due to increased diversity and complexity of inspection data. The inspection plan is configured based on the operation scenario characteristics, and then the inspection plan is updated through the twin digital platform. The technical effect of improving the scene adaptability of the inspection plan is achieved without affecting the physical equipment.

Description

Technical Field

[0001] The present application relates to the field of electric power safety technology, and specifically to an inspection method and system that combines the characteristics of electric power operation scenarios. Background Art

[0002] With the rapid development of the power industry, the scale and complexity of power equipment and systems are increasing, and the requirements for inspection work are becoming higher and higher. Traditional inspection methods often rely on manual inspection and recording, which is not only inefficient but also prone to errors. With the rapid development of Internet of Things technology, the real-time collection and transmission, centralized storage and analysis of inspection data are achieved through the introduction of sensors and communication technologies. Although these technologies have improved the efficiency and quality of inspection work to a certain extent, there are still some problems in practical applications. For example, the operating scenario characteristics of different equipment and systems vary greatly, resulting in increased diversity and complexity of inspection data, insufficient consideration of the characteristics of power operation scenarios, and difficulty in making adaptive adjustments based on scenarios.

[0003] In summary, the existing technology has technical problems such as poor scene adaptability of inspection schemes due to increased diversity and complexity of inspection data and insufficient analysis of power operation scene characteristics. Summary of the invention

[0004] The present application provides an inspection method and system that combines the characteristics of power operation scenarios, so as to solve the technical problem in the prior art that the diversity and complexity of inspection data increase, the analysis of power operation scenario characteristics is insufficient, and the scene adaptability of the inspection scheme is poor.

[0005] According to the first aspect of the present application, a patrol method combined with power operation scene characteristics is provided, including: collecting power operation scene information, performing feature extraction, and obtaining power operation scene characteristics; clustering based on the power operation scene characteristics to determine multi-type operation scene feature clusters; performing patrol feature analysis on the multi-type operation scene feature clusters to obtain patrol features; configuring patrol plans using the patrol features to determine patrol plans; based on the patrol plans, extracting a patrol anomaly library, aggregating the patrol anomaly library, and obtaining abnormal patrol information; sending the abnormal patrol information to a twin digital platform as a training task to perform abnormal patrol task training; feeding back abnormal patrol parameters that have reached convergence results in the training to the patrol platform, and updating the parameters of the patrol plan.

[0006] According to a second aspect of the present application, a patrol system combining power operation scene features is provided, including: a scene feature extraction unit, the scene feature extraction unit is used to collect power operation scene information, perform feature extraction, and obtain power operation scene features; a feature clustering unit, the feature clustering unit is used to perform clustering based on the power operation scene features, and determine multiple types of operation scene feature clusters; a patrol feature parsing unit, the patrol feature parsing unit is used to perform patrol feature parsing on the multiple types of operation scene feature clusters to obtain patrol features; a patrol scheme configuration unit, the patrol scheme configuration unit is used to configure the patrol scheme using the patrol features and determine the patrol scheme; an abnormal aggregation unit, the abnormal aggregation unit is used to extract a patrol abnormality library based on the patrol scheme, aggregate the patrol abnormality library, and obtain abnormal patrol information; an abnormal patrol task training unit, the abnormal patrol task training unit is used to send the abnormal patrol information to the twin digital platform as a training task, and perform abnormal patrol task training; a parameter updating unit, the parameter updating unit is used to feed back the abnormal patrol parameters that have reached convergence results in the training to the patrol platform, and update the parameters of the patrol scheme.

[0007] Based on the above analysis, it can be seen that one or more technical solutions provided by this application can achieve the following beneficial effects:

[0008] Collect power operation scenario information, extract features, obtain power operation scenario features, perform clustering based on power operation scenario features, determine multi-type operation scenario feature clusters, perform inspection feature analysis on multi-type operation scenario feature clusters, obtain inspection features, use inspection features to configure inspection plans, determine inspection plans, extract inspection anomaly libraries based on inspection plans, aggregate inspection anomaly libraries, obtain abnormal inspection information, send training tasks for abnormal inspection information to the twin digital platform, perform abnormal inspection task training, feed back abnormal inspection parameters that have reached convergence results to the inspection platform, update inspection plan parameters, thereby configuring inspection plans based on operation scenario features, and then updating inspection plans through the twin digital platform, achieving the technical effect of improving the scene adaptability of inspection plans without affecting physical equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in this application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. The drawings constituting a part of this application are used to provide a further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation on this application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without creative work.

[0010] Figure 1 A flowchart of an inspection method combined with power operation scenario characteristics provided in an embodiment of the present application;

[0011] Figure 2 A schematic diagram of the structure of an inspection system combined with the characteristics of power operation scenarios provided in an embodiment of the present application.

[0012] Explanation of the reference numerals: scene feature extraction unit 11, feature clustering unit 12, inspection feature analysis unit 13, inspection plan configuration unit 14, anomaly aggregation unit 15, abnormal inspection task training unit 16, parameter updating unit 17. DETAILED DESCRIPTION

[0013] In order to make the purpose, technical solution and advantages of the present application more obvious, the exemplary embodiments of the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0014] The terms used in the specification are used to describe the embodiments, rather than to limit the present application. As used in the specification, the singular terms "a", "an" and "the" are intended to also include plural forms, unless the context clearly indicates otherwise. When used in the specification, the terms "comprise" and / or "include" specify the presence of steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other steps, operations, elements, components and / or groups thereof.

[0015] Unless otherwise defined, all terms (including technical and scientific terms) used in this specification shall have the same meaning as commonly understood by those skilled in the art to which this application belongs. Terms, such as those defined in commonly used dictionaries, should not be interpreted in an idealized or overly formal sense unless explicitly defined herein. Throughout the specification, the same reference numerals represent the same elements.

[0016] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0017] Embodiment 1

[0018] Figure 1 A diagram of an inspection method combined with the characteristics of a power operation scenario provided in an embodiment of the present application, the method comprising:

[0019] Collect power operation scenario information, perform feature extraction, and obtain power operation scenario features;

[0020] The power operation scenario information refers to the operation scenarios of all power equipment in a specific area. For example, the operation records of power equipment in historical time can be collected as the power operation scenario information. The specific area is the area that needs to be inspected, and the power operation scenario information is subjected to feature extraction, that is, the electrical parameters, operation time, equipment layout range, etc. of the equipment during operation are extracted as the power operation scenario features.

[0021] Clustering is performed based on the power operation scenario characteristics to determine multiple types of operation scenario characteristic clusters;

[0022] In a preferred embodiment, it also includes:

[0023] The multi-cluster center parameters are set, wherein the multi-cluster center parameters include operating equipment, equipment deployment range, operation participation target attributes, and operating time; based on the multi-cluster center parameters, a multi-level clustering execution model is obtained by training a historical data sample set; the multi-level clustering execution model is used to perform hierarchical clustering on the power operation scenario features to obtain a multi-level clustering result; according to the multi-level clustering result, the power operation scenario features of multiple power operation scenarios are clustered and labeled, and cluster division is performed according to the labeled values ​​of the cluster labels to obtain the multi-type operation scenario feature clusters.

[0024] In a preferred embodiment, it also includes:

[0025] According to the multi-level clustering results, the clustering distances corresponding to each level of the multi-level clustering results are obtained; coefficient conversion is performed according to the clustering distance to obtain the clustering coefficient, and a labeling value is generated based on the clustering coefficient to cluster and label the power operation scene characteristics; a core proportion coefficient of multiple cluster center parameters is preset, and the labeling value is matched in positive order based on the core proportion coefficient, and the matching result is projected with the preset cluster to construct the multi-type operation scene feature cluster.

[0026] Specifically, based on the characteristics of power operation scenarios, scenario clustering analysis is performed from multiple dimensions to obtain multi-type operation scenario feature clusters. The multi-type operation scenario feature clusters include different types of operation scenario feature sets. The specific method is described in detail as follows.

[0027] Specifically, set the multi-cluster center parameters, which include operating equipment, equipment deployment range, operation participation target attributes, and operation time, where the operating equipment can be understood as the type of power equipment, such as transformers, inverters, etc.; the equipment deployment range refers to the area where the power equipment is located; the operation participation target attribute refers to the participants when the equipment is running, such as the need for staff control to operate, or through machine control operation, and the operation of people and machines; the operation time refers to the length of time that the power equipment has been put into use. Based on the multi-cluster center parameters, a multi-level clustering execution model is obtained by training with a historical data sample set. The multi-level clustering execution model is an existing neural network model, which uses any one of the multi-cluster center parameters in turn to achieve multi-level classification.

[0028] The historical data sample set is a database consisting of power operation scene features and clustering result samples collected by professional and technical personnel in the field over a period of time in the past. The multi-level clustering execution model includes multiple execution branches, each of which performs clustering analysis based on a center parameter of multiple clustering center parameters. Exemplarily, multiple execution branches can be constructed based on the existing k-means, and the multi-level clustering execution model is trained and tested with the historical data sample set and clustering result samples to obtain a multi-level clustering execution model with an accuracy rate that meets the requirements. Furthermore, the multi-level clustering execution model is used to perform hierarchical clustering of the power operation scene features, that is, the power operation scene features are input into the multi-level clustering execution model, and the multi-level clustering results are output. The multi-level clustering results include clustering results obtained based on the multiple clustering center parameters respectively.

[0029] The power operation scenario characteristics of multiple power operation scenarios are clustered and labeled according to the multi-level clustering results, and cluster division is performed according to the labeled values ​​of the cluster annotations to obtain the multi-type operation scenario feature clusters, thereby realizing cluster division of operation scenario characteristics, facilitating the subsequent formulation of inspection plans based on the scenario characteristics and improving the scenario adaptability of the inspection plans.

[0030] Among them, the power operation scene characteristics of multiple power operation scenes are clustered and labeled according to the multi-level clustering results, and cluster division is performed according to the labeled values ​​of the cluster labels. The method for obtaining the multi-type operation scene feature clusters is: according to the multi-level clustering results, the cluster distances at each level corresponding to the multi-level clustering results are obtained, wherein the cluster distances at each level refer to the distances between the multi-level clustering results and the parameters of the multi-cluster center, such as the Euclidean distance, which can be calculated and obtained based on the existing technology when clustering. Then, coefficient conversion is performed according to the cluster distance to obtain the clustering coefficient. Specifically, the clustering coefficient can be obtained by calculating the ratio of the cluster distances at each level to the sum of the cluster distances at each level. Based on the cluster coefficient, a labeled value that can identify the cluster coefficient is generated, and the power operation scene characteristics are clustered and labeled. The core proportion coefficient of the preset multi-cluster center parameters is set by the professional and technical personnel in this field. It can be simply understood as the weight coefficient of the multi-cluster center parameters. That is, different cluster centers have different reference values ​​for equipment inspection and different importance. The higher the importance, the greater the core proportion coefficient. It is specifically set by the professional and technical personnel in this field based on actual experience.

[0031] Finally, based on the core proportion coefficient, the marked values ​​are matched in positive order, that is, the marked values ​​corresponding to the multi-level clustering results and the core proportion coefficients are matched one by one to obtain matching results, and the matching results are projected with the preset clusters. The preset clusters refer to the clusters that can be divided based on the marked values ​​and the core proportion coefficients pre-set by professional and technical personnel in this field, that is, the clusters corresponding to the matching results are extracted from the preset clusters as the multi-type operation scene feature clusters. In this way, the clustering of the operation scene features is realized, and the scene adaptability of the equipment inspection plan is improved.

[0032] Performing inspection feature analysis on the multi-type operation scenario feature clusters to obtain inspection features;

[0033] In a preferred embodiment, it also includes:

[0034] Based on the control inspection parameters of the inspection platform, a standard inspection mapping library is constructed, which includes control inspection parameters, inspection features, inspection feature description attributes and their mapping relationships; feature decomposition is performed on the multi-type operation scenario feature clusters to obtain feature description attributes of each type; the feature description attributes are compared and mapped using the standard inspection mapping library to obtain the inspection features.

[0035] The inspection feature is parsed for the multi-type running scenario feature clusters to obtain the inspection features, and the method is as follows:

[0036] Based on the control inspection parameters of the inspection platform, an inspection standard mapping library is constructed, which includes control inspection parameters, inspection features, inspection feature description attributes and their mapping relationships. The inspection platform is a system platform for managing inspection tasks. The control inspection parameters of the inspection platform refer to inspection parameters that can be adjusted and controlled, such as inspection cycles and inspection routes; inspection features refer to key features that need to be observed and measured, such as temperature and current; and inspection feature description attributes refer to the feature types of corresponding equipment, such as operating time. Control inspection parameters, inspection features, and inspection feature description attributes have a one-to-one mapping relationship.

[0037] The multi-type operation scenario feature clusters are decomposed respectively, and the multi-type operation scenario feature clusters contain multiple features, and their respective feature description attributes are extracted, such as the corresponding operation time. Then, the feature description attributes are compared and mapped using the inspection standard mapping library, that is, the control inspection parameters and inspection features corresponding to the feature description attributes are matched in the inspection standard mapping library as the inspection features.

[0038] Using the inspection features to configure an inspection plan and determine the inspection plan;

[0039] In a preferred embodiment, it also includes:

[0040] According to the inspection characteristics, various types of inspection parameters are numerically quantified to determine the scene feature inspection parameter values; an inspection scheme parameter configuration module is built, and the scene feature inspection parameter values ​​are used as input to perform inspection parameter matching through the inspection scheme parameter configuration module, and based on the matching relationship, the scheme is configured according to the scene feature inspection parameter values ​​to generate the inspection scheme.

[0041] In a preferred embodiment, it also includes:

[0042] The inspection plan parameter configuration module includes mandatory inspection parameters and additional inspection parameters. When the matching relationship satisfies the mandatory inspection parameters, the inspection plan configuration is performed. When the additional inspection parameters are empty, the inspection plan configuration is performed after supplementary assignment according to the preset parameter assignment rules.

[0043] The inspection features are used to configure the inspection plan and determine the inspection plan. The method is as follows:

[0044] According to the inspection characteristics, the numerical quantization of various types of inspection parameters is to quantify the control inspection parameters. If they are quantitative values, such as the inspection cycle, no processing is required. If they are not quantitative values, such as the inspection route, they need to be converted into numerical values. The technical personnel in this field shall set the conversion values ​​by themselves. The numerical quantization results of various types of inspection parameters are used as scene feature inspection parameter values. The inspection scheme parameter configuration module is built. The inspection scheme parameter configuration module is a model for generating inspection schemes. Specifically, the historical inspection parameter record data and the corresponding historical scene feature inspection parameter value record data can be retrieved to establish a matching mapping database and stored in the inspection scheme parameter configuration module.

[0045] Then, the scene characteristic inspection parameter value is used as input to match the inspection parameters through the inspection scheme parameter configuration module, and the scheme is configured according to the scene characteristic inspection parameter value based on the matching relationship. That is, based on the correspondence between historical inspection parameter record data and historical scene characteristic inspection parameter value record data, the inspection parameters corresponding to the scene characteristic inspection parameter value are obtained as the inspection scheme.

[0046] Among them, the inspection plan parameter configuration module includes mandatory inspection parameters and additional inspection parameters. The mandatory inspection parameters refer to the inspection parameters that must be extracted, that is, the inspection parameters that need to be obtained for each inspection feature, such as inspection cycle and inspection route; the additional inspection parameters are separately added inspection parameters set for different scene features. For example, a certain inspection feature requires electrical inspection, and electrical parameters need to be configured. When the matching relationship meets the mandatory inspection parameters, the inspection plan is configured, that is, the matched mandatory inspection parameters are directly added to the inspection plan. At the same time, the matched additional inspection parameters are also added to the inspection plan. When the matched additional inspection parameters are empty, the inspection plan is configured after supplementary assignment according to the preset parameter assignment rules, such as assigning 0 or 1 to the additional inspection parameters, which is specifically set by professional and technical personnel in this field.

[0047] This enables the matching of inspection plans based on scene characteristics, improves the scene adaptability of inspection plans, and provides a basis for subsequent inspection anomalies and parameter updates.

[0048] Based on the inspection scheme, extracting an inspection anomaly library, aggregating the inspection anomaly library, and obtaining abnormal inspection information;

[0049] Based on the inspection scheme, the inspection anomaly library is extracted, and the inspection anomaly library is aggregated to obtain abnormal inspection information, that is, after the inspection scheme is inspected, the abnormal information generated during the inspection process is recorded, such as the abnormal inspection location and occurrence time where the inspection cannot be carried out due to obstacles, and other parameter combinations are used as the inspection anomaly library. The abnormal information in the inspection anomaly library is classified and aggregated, and similar anomalies are classified into one category or aggregated into a set. This can be achieved based on existing clustering methods, such as the KNN algorithm, and the clustering results are used as abnormal inspection information.

[0050] Send the abnormal inspection information as a training task to the twin digital platform to perform abnormal inspection task training;

[0051] In a preferred embodiment, it also includes:

[0052] According to the abnormal inspection information, the inspection identification information is determined; based on the inspection identification information, data nodes are located in the twin digital platform, and data relationships are radiated by node positioning to determine the target positioning data network; according to the target positioning data network, training data sets are extracted to construct a twin task training set; based on the abnormal inspection information, the learning direction, learning objectives, and iteration cycle are determined, and according to the learning direction, learning objectives, and iteration cycle, the twin task training set is used to perform task learning training until the learning objectives or iteration cycle are reached.

[0053] In a preferred embodiment, it also includes:

[0054] Based on the abnormal inspection information and the iteration cycle, a determination cycle node is set; when the determination cycle node is reached, the learning target approximation relationship analysis is performed on the training results to determine the learning approximation trend; if the learning approximation trend does not meet the preset requirements, the learning direction is adjusted based on the learning approximation trend, wherein the adjustment of the learning direction includes adjusting according to one or more of the learning approximation trend direction, according to other optimization directions of the abnormal inspection information, and in a direction opposite to the current learning direction.

[0055] The abnormal inspection information is sent to the twin digital platform as a training task to perform abnormal inspection task training. The specific method is as follows:

[0056] Digital twins are essentially a copy, a software representation of various assets, information, and processes in the real world. They exist in the cloud. In this embodiment, a twin digital platform is built based on digital twin technology to train abnormal inspection information without causing any risk to power equipment in the real world. Specifically, the parameters such as the structure and location of the equipment in the specific area that needs to be inspected can be obtained and input into the existing digital twin modeling software to generate a twin digital platform.

[0057] According to the abnormal inspection information, the inspection identification information is determined, and the inspection identification information is the position or time node where the inspection anomaly is identified in the abnormal inspection information. Based on the inspection identification information, data node positioning is performed in the twin digital platform, that is, the data node that generates the abnormal inspection information is located, and the data relationship is radiated by node positioning to determine the target positioning data network, that is, the data nodes that generate the abnormal inspection information and the data transmission relationship between the nodes are connected to obtain the target positioning data network. Then, the training data set is extracted according to the target positioning data network, that is, the operation data generated by the target positioning data network is extracted to form a twin task training set.

[0058] Finally, based on the abnormal inspection information, the learning direction, learning goal, and iteration cycle are determined. According to the learning direction, learning goal, and iteration cycle, the twin task training set is used to perform task learning training until the learning goal or iteration cycle is reached. That is to say, the twin task training set needs to be used for task learning training to eliminate the impact of the abnormal inspection information, such as eliminating the abnormality by changing the inspection route, adjusting the inspection time, etc. The learning direction needs to be determined according to the abnormal inspection information. For example, if the inspection cannot be carried out due to an obstacle, the inspection route needs to be trained and updated, and the learning direction is the direction to avoid the obstacle. The learning goal and iteration cycle are both judgment conditions for the training to converge, such as the number of iterations. Using the twin task training set for task learning training is to continuously adjust the inspection parameters according to the learning direction until the learning goal or iteration cycle is reached. In this way, the analysis and update of the abnormal inspection parameters are realized, and the inspection effect is improved without affecting the operation of the actual equipment.

[0059] Specifically, based on the abnormal inspection information and the iteration cycle, the key time node is set according to the abnormal inspection information to extract the abnormal occurrence frequency, and the node is located in the iteration cycle based on the key time node to obtain the judgment cycle node. When the judgment cycle node is reached, the training result is analyzed for the learning target approximation relationship to determine the learning approximation trend, that is, the difference between the training result and the learning target under the judgment cycle node is used as the learning approximation trend. For the learning approximation trend that does not meet the preset requirements, that is, the learning approximation trend cannot reach the learning target, the learning direction is adjusted based on the learning approximation trend, wherein the learning direction is adjusted according to the learning approximation trend direction, according to other optimization directions of the abnormal inspection information, and in the opposite direction to the current learning direction. One or more adjustments are made, that is, by adjusting the learning direction, the learning approximation trend reaches the learning target. For example, the inspection cycle is currently adjusted to a larger value, and it can be adjusted to a smaller value later, or adjusted to a direction different from the inspection line, etc. In this way, the learning direction adjustment of the abnormal inspection parameters is realized to improve the learning efficiency.

[0060] The abnormal inspection parameters that have reached convergence results in training are fed back to the inspection platform, and the parameters of the inspection plan are updated.

[0061] Finally, the abnormal inspection parameters that have reached convergence results in training, that is, the inspection parameters that have reached the learning target or iteration cycle, are fed back to the inspection platform, the parameters of the inspection plan are updated, and the updated inspection plan is obtained to perform equipment inspection.

[0062] Based on the above analysis, it can be seen that one or more technical solutions provided by this application can achieve the following beneficial effects:

[0063] Collect power operation scenario information, extract features, obtain power operation scenario features, perform clustering based on power operation scenario features, determine multi-type operation scenario feature clusters, perform inspection feature analysis on multi-type operation scenario feature clusters, obtain inspection features, use inspection features to configure inspection plans, determine inspection plans, extract inspection anomaly libraries based on inspection plans, aggregate inspection anomaly libraries, obtain abnormal inspection information, send training tasks for abnormal inspection information to the twin digital platform, perform abnormal inspection task training, feed back abnormal inspection parameters that have reached convergence results to the inspection platform, update inspection plan parameters, thereby configuring inspection plans based on operation scenario features, and then updating inspection plans through the twin digital platform, achieving the technical effect of improving the scene adaptability of inspection plans without affecting physical equipment.

[0064] Embodiment 2

[0065] Based on the same inventive concept as the inspection method combined with the characteristics of the power operation scene in the above-mentioned embodiment, Figure 2 As shown, the present application also provides an inspection system combining the characteristics of power operation scenarios, the system comprising:

[0066] A scene feature extraction unit 11, wherein the scene feature extraction unit 11 is used to collect power operation scene information, perform feature extraction, and obtain power operation scene features;

[0067] A feature clustering unit 12, the feature clustering unit 12 is used to perform clustering based on the power operation scene features to determine multiple types of operation scene feature clusters;

[0068] An inspection feature parsing unit 13, the inspection feature parsing unit 13 is used to perform inspection feature parsing on the multi-type operation scenario feature clusters to obtain inspection features;

[0069] An inspection scheme configuration unit 14, the inspection scheme configuration unit 14 is used to configure the inspection scheme using the inspection features and determine the inspection scheme;

[0070] An abnormal aggregation unit 15, the abnormal aggregation unit 15 is used to extract an inspection abnormality library based on the inspection scheme, aggregate the inspection abnormality library, and obtain abnormal inspection information;

[0071] An abnormal inspection task training unit 16, the abnormal inspection task training unit 16 is used to send the abnormal inspection information to the twin digital platform as a training task to perform abnormal inspection task training;

[0072] The parameter updating unit 17 is used to feed back the abnormal inspection parameters that have reached convergence results through training to the inspection platform, and to update the parameters of the inspection scheme.

[0073] Furthermore, the feature clustering unit 12 further includes:

[0074] Setting multi-cluster center parameters, wherein the multi-cluster center parameters include operating equipment, equipment deployment range, operating participation target attributes, and operating time;

[0075] Based on the multi-cluster center parameters, a multi-level clustering execution model is obtained by training with a historical data sample set;

[0076] Using the multi-level clustering execution model to perform hierarchical clustering on the power operation scenario features to obtain a multi-level clustering result;

[0077] The power operation scene features of multiple power operation scenes are clustered and labeled according to the multi-level clustering results, and cluster division is performed according to the labeled values ​​of the cluster labels to obtain the multi-type operation scene feature clusters.

[0078] Furthermore, the feature clustering unit 12 further includes:

[0079] According to the multi-level clustering result, obtaining the clustering distances of each level corresponding to the multi-level clustering result;

[0080] Performing coefficient conversion according to the clustering distance to obtain a clustering coefficient, generating a labeling value based on the clustering coefficient, and clustering labeling the power operation scene features;

[0081] The core proportion coefficients of the multi-cluster center parameters are preset, and the labeled values ​​are matched in positive order based on the core proportion coefficients. The matching results are projected with the preset clusters to construct the multi-type operation scenario feature clusters.

[0082] Furthermore, the inspection feature analysis unit 13 also includes:

[0083] Based on the control inspection parameters of the inspection platform, an inspection standard mapping library is constructed, wherein the inspection standard mapping library includes control inspection parameters, inspection features, inspection feature description attributes and their mapping relationships;

[0084] Decomposing the feature clusters of the multiple types of operation scenarios respectively to obtain feature description attributes of each type;

[0085] The inspection standard mapping library is used to compare and map the feature description attributes to obtain the inspection feature.

[0086] Furthermore, the inspection scheme configuration unit 14 also includes:

[0087] Numerical quantification of various types of inspection parameters is performed according to the inspection characteristics to determine scene characteristic inspection parameter values;

[0088] Build an inspection plan parameter configuration module, use the scene feature inspection parameter value as input to perform inspection parameter matching through the inspection plan parameter configuration module, and configure the plan according to the scene feature inspection parameter value based on the matching relationship to generate the inspection plan.

[0089] Furthermore, the inspection scheme configuration unit 14 also includes:

[0090] The inspection plan parameter configuration module includes mandatory inspection parameters and additional inspection parameters. When the matching relationship satisfies the mandatory inspection parameters, the inspection plan configuration is performed. When the additional inspection parameters are empty, the inspection plan configuration is performed after supplementary assignment according to the preset parameter assignment rules.

[0091] Furthermore, the abnormal inspection task training unit 16 also includes:

[0092] Determining inspection identification information according to the abnormal inspection information;

[0093] Based on the inspection identification information, data nodes are located in the twin digital platform, data relationships are radiated by node positioning, and a target positioning data network is determined;

[0094] Extracting a training data set according to the target positioning data network and constructing a twin task training set;

[0095] Based on the abnormal inspection information, the learning direction, learning objectives, and iteration cycle are determined. According to the learning direction, learning objectives, and iteration cycle, the twin task training set is used to perform task learning training until the learning objective or iteration cycle is reached.

[0096] Furthermore, the abnormal inspection task training unit 16 also includes:

[0097] Based on the abnormal inspection information and the iteration cycle, setting a determination cycle node;

[0098] When the determination period node is reached, the training results are analyzed for the learning target approximation relationship to determine the learning approximation trend;

[0099] If the learning approximate trend does not meet the preset requirements, the learning direction is adjusted based on the learning approximate trend, wherein the adjustment of the learning direction includes adjusting according to one or more of the learning approximate trend direction, according to other optimization directions of abnormal inspection information, and in a direction opposite to the current learning direction.

[0100] The specific example of an inspection method combined with power operation scenario characteristics in the aforementioned embodiment 1 is also applicable to an inspection system combined with power operation scenario characteristics in this embodiment. Through the aforementioned detailed description of an inspection method combined with power operation scenario characteristics, technical personnel in this field can clearly understand an inspection system combined with power operation scenario characteristics in this embodiment, so for the sake of brevity of the specification, it will not be described in detail here.

[0101] It should be understood that various forms of the processes shown above can be used to reorder, add or delete steps, as long as the expected results of the technical solutions disclosed in this application can be achieved, and this document does not limit this.

[0102] Note that the above are only preferred embodiments of the present application and the technical principles used. Those skilled in the art will understand that the present application is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present application. Therefore, although the present application is described in more detail through the above embodiments, the present application is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present application, and the scope of the present application is determined by the scope of the appended claims.

Claims

1. A patrol inspection method combining the characteristics of power operation scenarios, characterized in that: include: Collect power operation scenario information, perform feature extraction, and obtain power operation scenario features; Clustering is performed based on the power operation scenario characteristics to determine multiple types of operation scenario characteristic clusters; Performing inspection feature analysis on the multi-type operation scenario feature clusters to obtain inspection features; Using the inspection features to configure an inspection plan and determine the inspection plan; Based on the inspection scheme, extracting an inspection anomaly library, aggregating the inspection anomaly library, and obtaining abnormal inspection information; Send the abnormal inspection information as a training task to the twin digital platform to perform abnormal inspection task training; Feedback the abnormal inspection parameters that have reached convergence results in the training to the inspection platform, and update the parameters of the inspection plan; Perform inspection feature analysis on the multi-type operation scenario feature clusters to obtain inspection features, including: Based on the control inspection parameters of the inspection platform, an inspection standard mapping library is constructed, wherein the inspection standard mapping library includes control inspection parameters, inspection features, inspection feature description attributes and their mapping relationships; Decomposing the feature clusters of the multiple types of operation scenarios respectively to obtain feature description attributes of each type; Using the inspection standard mapping library to compare and map the feature description attributes to obtain the inspection feature; The abnormal inspection information is sent to the twin digital platform as a training task to perform abnormal inspection task training, including: Determining inspection identification information according to the abnormal inspection information; Based on the inspection identification information, data nodes are located in the twin digital platform, data relationships are radiated by node positioning, and a target positioning data network is determined; Extracting a training data set according to the target positioning data network and constructing a twin task training set; Based on the abnormal inspection information, determine the learning direction, learning goal, and iteration cycle, and use the twin task training set to perform task learning training according to the learning direction, learning goal, and iteration cycle until the learning goal or iteration cycle is reached; Based on the abnormal inspection information and the iteration cycle, setting a determination cycle node; When the determination period node is reached, the training results are analyzed for the learning target approximation relationship to determine the learning approximation trend; If the learning approximate trend does not meet the preset requirements, the learning direction is adjusted based on the learning approximate trend, wherein the adjustment of the learning direction includes adjusting according to one or more of the learning approximate trend direction, according to other optimization directions of abnormal inspection information, and in a direction opposite to the current learning direction.

2. The method according to claim 1, characterized in that Clustering is performed based on the power operation scenario characteristics to determine multiple types of operation scenario feature clusters, including: Setting multi-cluster center parameters, wherein the multi-cluster center parameters include operating equipment, equipment deployment range, operating participation target attributes, and operating time; Based on the multi-cluster center parameters, a multi-level clustering execution model is obtained by training with a historical data sample set; Using the multi-level clustering execution model to perform hierarchical clustering on the power operation scenario features to obtain a multi-level clustering result; The power operation scene features of multiple power operation scenes are clustered and labeled according to the multi-level clustering results, and cluster division is performed according to the labeled values ​​of the cluster labels to obtain the multi-type operation scene feature clusters.

3. The method according to claim 2, characterized in that The power operation scene features of multiple power operation scenes are clustered and labeled according to the multi-level clustering results, and clusters are divided according to the labeled values ​​of the cluster labels to obtain the multi-type operation scene feature clusters, including: According to the multi-level clustering result, obtaining the clustering distances of each level corresponding to the multi-level clustering result; Performing coefficient conversion according to the clustering distance to obtain a clustering coefficient, generating a labeling value based on the clustering coefficient, and clustering labeling the power operation scene features; The core proportion coefficients of the multi-cluster center parameters are preset, and the labeled values ​​are matched in positive order based on the core proportion coefficients. The matching results are projected with the preset clusters to construct the multi-type operation scenario feature clusters.

4. The method according to claim 1, characterized in that The inspection features are used to configure the inspection plan and determine the inspection plan, including: Numerical quantification of various types of inspection parameters is performed according to the inspection characteristics to determine scene characteristic inspection parameter values; Build an inspection plan parameter configuration module, use the scene feature inspection parameter value as input to perform inspection parameter matching through the inspection plan parameter configuration module, and configure the plan according to the scene feature inspection parameter value based on the matching relationship to generate the inspection plan.

5. The method according to claim 4, characterized in that The inspection plan parameter configuration module includes mandatory inspection parameters and additional inspection parameters. When the matching relationship satisfies the mandatory inspection parameters, the inspection plan configuration is performed. When the additional inspection parameters are empty, the inspection plan configuration is performed after supplementary assignment according to the preset parameter assignment rules.

6. A patrol inspection system combining the characteristics of power operation scenarios, characterized in that: The system comprises: A scene feature extraction unit, the scene feature extraction unit is used to collect power operation scene information, perform feature extraction, and obtain power operation scene features; A feature clustering unit, the feature clustering unit is used to perform clustering based on the power operation scene features to determine multiple types of operation scene feature clusters; An inspection feature parsing unit, the inspection feature parsing unit is used to perform inspection feature parsing on the multi-type operation scenario feature clusters to obtain inspection features; An inspection scheme configuration unit, the inspection scheme configuration unit is used to configure the inspection scheme using the inspection features and determine the inspection scheme; An abnormal aggregation unit, the abnormal aggregation unit is used to extract an inspection abnormality library based on the inspection scheme, aggregate the inspection abnormality library, and obtain abnormal inspection information; An abnormal inspection task training unit, which is used to send the abnormal inspection information as a training task to the twin digital platform to perform abnormal inspection task training; A parameter updating unit, the parameter updating unit is used to feed back the abnormal inspection parameters that have reached convergence results in training to the inspection platform, and update the parameters of the inspection scheme; The inspection feature analysis unit also includes: Based on the control inspection parameters of the inspection platform, an inspection standard mapping library is constructed, wherein the inspection standard mapping library includes control inspection parameters, inspection features, inspection feature description attributes and their mapping relationships; Decomposing the feature clusters of the multiple types of operation scenarios respectively to obtain feature description attributes of each type; Using the inspection standard mapping library to compare and map the feature description attributes to obtain the inspection feature; The abnormal inspection task training unit also includes: Determining inspection identification information according to the abnormal inspection information; Based on the inspection identification information, data nodes are located in the twin digital platform, data relationships are radiated by node positioning, and a target positioning data network is determined; Extracting a training data set according to the target positioning data network and constructing a twin task training set; Based on the abnormal inspection information, determine the learning direction, learning goal, and iteration cycle, and use the twin task training set to perform task learning training according to the learning direction, learning goal, and iteration cycle until the learning goal or iteration cycle is reached; Based on the abnormal inspection information and the iteration cycle, setting a determination cycle node; When the determination period node is reached, the training results are analyzed for the learning target approximation relationship to determine the learning approximation trend; If the learning approximate trend does not meet the preset requirements, the learning direction is adjusted based on the learning approximate trend, wherein the adjustment of the learning direction includes adjusting according to one or more of the learning approximate trend direction, according to other optimization directions of abnormal inspection information, and in a direction opposite to the current learning direction.

Images