A method and system for intelligent verification of power work order content

Abstract

This application provides an intelligent verification method and system for power work orders. The method acquires the target content information of the power work order, then determines the target graph structure data corresponding to the target content information based on this information, and finally inputs the target graph structure data into a pre-trained dependency verification model to obtain a first verification result. The dependency verification model is a verification model trained based on the graph structure data corresponding to historical power work orders and the verification results corresponding to that graph structure data, used to obtain the dependency relationships between various entities in the target content information. In this technical solution, the dependency verification model can determine the dependency relationships between various entities in the power work order, and judge whether there are logical errors, missing dependencies, or dependency conflicts in the power work order, thus efficiently assisting users in determining the accuracy of the power work order.

Description

Technical Field

[0001] This application relates to the field of power technology, and in particular to an intelligent verification method and system for power work permit content. Background Technology

[0002] A work permit is a written instruction that allows work to be carried out on electrical equipment and power lines. It is also a written basis for clarifying safety responsibilities, providing safety instructions to workers, and ensuring their safety. The management of work permits is an important support for the development of power work, and the verification of work permits is of great significance for personal and property safety.

[0003] In existing technologies, it is necessary to rely on experienced power workers or safety managers to carefully review every field and detail of the work order to ensure that it complies with relevant safety specifications, operating procedures and legal requirements.

[0004] However, this method in the existing technology is time-consuming, labor-intensive, inefficient, and has the technical problem of low accuracy. Summary of the Invention

[0005] This application provides an intelligent verification method and system for power work permits to solve the problems of inaccurate verification, low efficiency, and time and labor costs in existing technologies.

[0006] In a first aspect, embodiments of this application provide an intelligent verification method for the content of power work orders, including:

[0007] Obtain the target content information of the power work order;

[0008] Based on the target content information of the power work order, determine the target graph structure data corresponding to the target content information;

[0009] The target graph structure data is input into a pre-trained dependency verification model to obtain a first verification result. The first verification result is used to indicate whether there are logical errors, missing dependencies, or dependency conflicts in the power work order. The dependency verification model is a verification model trained based on the graph structure data corresponding to historical power work orders and the verification results corresponding to the graph structure data, used to obtain the dependency relationships between various entities in the target content information.

[0010] In one or more embodiments, the target graph structure data is the graph structure of the power work order;

[0011] Accordingly, determining the target graph structure data corresponding to the target content information based on the target content information of the power work order includes:

[0012] The target content information is structured to obtain the entities involved in the target content information and the relationships between the entities;

[0013] Entities are treated as nodes in the target graph structure data, and the relationships between entities are treated as edges in the target graph structure data.

[0014] In one or more embodiments, after obtaining the target content information of the power work order, the method further includes:

[0015] The target content information of the power work order is input into a pre-trained preliminary verification model to obtain a second verification result; the second verification result is used to identify key information in the power work order to determine whether there are format or content errors in the power work order.

[0016] Feature extraction is performed on the target content information of the power work order to obtain vector features of each part of the target content information;

[0017] Based on a preset similarity algorithm, the semantic similarity between each vector feature is determined. The semantic similarity is used to describe the consistency, duplicate content, and similar content among the fields in the target content information.

[0018] The power work order is comprehensively evaluated based on the second verification result and the semantic similarity.

[0019] In one or more embodiments, after extracting features from the target content information of the power work order to obtain vector features of each part of the target content information, the method further includes:

[0020] The vector features are input into a pre-trained semantic understanding model to obtain a third verification result. The third verification result is used to describe the rationality and consistency of the logical relationship and contextual dependency between various fields in the target content information. The semantic understanding model is constructed using an attention mechanism and trained based on the vector features corresponding to historical power work tickets and the verification results corresponding to historical power work tickets.

[0021] In one or more embodiments, the method further includes:

[0022] Determine the optimal solution set among multiple optimization objectives;

[0023] During multiple runs of the target model, the parameters of the target model are adjusted in real time to obtain different output solution sets. The target model can be any one of the dependency verification model, semantic understanding model, and preliminary verification model.

[0024] The optimal target model is obtained by evaluating different output solution sets based on the optimal solution set among the multiple optimization objectives.

[0025] In one or more embodiments, the method further includes:

[0026] Determine the time series analysis information corresponding to the first power work order within the first preset time period before the current time node;

[0027] The time series analysis information is input into the prediction model to obtain the prediction result, which describes the quantity and type of the second power work order within a second preset time period after the current time node; the prediction model is trained based on historical power work orders.

[0028] In one or more embodiments, the method further includes:

[0029] The target content information is input into the anomaly detection model to obtain the abnormal data corresponding to the power work order. The anomaly detection model is trained from the content of multiple work orders and is used to determine the anomalies present in the target content information.

[0030] Based on the abnormal data, the abnormal result of the target content information is determined.

[0031] In one or more embodiments, the method further includes:

[0032] Based on the abnormal results, the target model is optimized to obtain the optimized target model.

[0033] In one or more embodiments, the method further includes:

[0034] The target verification result is encrypted using homomorphic encryption or differential privacy. The target verification result includes: a first verification result, a second verification result, and a third verification result.

[0035] Secondly, embodiments of this application provide an intelligent verification device for the content of power work orders, comprising:

[0036] The acquisition module is used to acquire the target content information of the power work order;

[0037] The determination module is used to determine the target diagram structure data corresponding to the target content information based on the target content information of the power work order;

[0038] The processing module is used to input the target graph structure data into a pre-trained dependency verification model to obtain a first verification result; the first verification result is used to indicate whether there are logical errors, missing dependencies, or dependency conflicts in the power work order; the dependency verification model is a verification model trained based on the graph structure data corresponding to historical power work orders and the verification results corresponding to the graph structure data, used to obtain the dependency relationships between various entities in the target content information.

[0039] In one or more embodiments, the target graph structure data is the graph structure of the power work order;

[0040] Accordingly, the determining module is specifically used for:

[0041] The target content information is structured to obtain the entities involved in the target content information and the relationships between the entities;

[0042] Entities are treated as nodes in the target graph structure data, and the relationships between entities are treated as edges in the target graph structure data.

[0043] In one or more embodiments, after obtaining the target content information of the power work order, the processing module is further configured to:

[0044] The target content information of the power work order is input into a pre-trained preliminary verification model to obtain a second verification result; the second verification result is used to identify key information in the power work order to determine whether there are format or content errors in the power work order.

[0045] Feature extraction is performed on the target content information of the power work order to obtain vector features of each part of the target content information;

[0046] Based on a preset similarity algorithm, the semantic similarity between each vector feature is determined. The semantic similarity is used to describe the consistency, duplicate content, and similar content among the fields in the target content information.

[0047] The power work order is comprehensively evaluated based on the second verification result and the semantic similarity.

[0048] In one or more embodiments, after performing feature extraction on the target content information of the power work order to obtain vector features of each part of the target content information, the processing module is further configured to:

[0049] The vector features are input into a pre-trained semantic understanding model to obtain a third verification result. The third verification result is used to describe the rationality and consistency of the logical relationship and contextual dependency between various fields in the target content information. The semantic understanding model is constructed using an attention mechanism and trained based on the vector features corresponding to historical power work tickets and the verification results corresponding to historical power work tickets.

[0050] In one or more embodiments, the processing module is further configured to:

[0051] Determine the optimal solution set among multiple optimization objectives;

[0052] During multiple runs of the target model, the parameters of the target model are adjusted in real time to obtain different output solution sets. The target model can be any one of the dependency verification model, semantic understanding model, and preliminary verification model.

[0053] The optimal target model is obtained by evaluating different output solution sets based on the optimal solution set among the multiple optimization objectives.

[0054] In one or more embodiments, the processing module is further configured to:

[0055] Determine the time series analysis information corresponding to the first power work order within the first preset time period before the current time node;

[0056] The time series analysis information is input into the prediction model to obtain the prediction result, which describes the quantity and type of the second power work order within a second preset time period after the current time node; the prediction model is trained based on historical power work orders.

[0057] In one or more embodiments, the processing module is further configured to:

[0058] The target content information is input into the anomaly detection model to obtain the abnormal data corresponding to the power work order. The anomaly detection model is trained from the content of multiple work orders and is used to determine the anomalies present in the target content information.

[0059] Based on the abnormal data, the abnormal result of the target content information is determined.

[0060] In one or more embodiments, the processing module is further configured to:

[0061] Based on the abnormal results, the target model is optimized to obtain the optimized target model.

[0062] In one or more embodiments, the processing module is further configured to:

[0063] The target verification result is encrypted using homomorphic encryption or differential privacy. The target verification result includes: a first verification result, a second verification result, and a third verification result.

[0064] Thirdly, this application provides an electronic device, including: a processor, and a memory communicatively connected to the processor;

[0065] The memory stores computer-executed instructions;

[0066] The processor executes computer execution instructions stored in the memory to implement the method as described in the first aspect or any of the above methods.

[0067] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the method described in the first aspect or any of the above-described methods.

[0068] Fifthly, this application provides a computer program, the computer program product including a computer program stored in a computer-readable storage medium, at least one processor being able to read the computer program from the computer-readable storage medium, the at least one processor executing the computer program being able to implement the method described in the first aspect or any of the above methods.

[0069] Sixthly, this application provides an intelligent verification system for the content of power work permits, comprising: an electronic device and a first device;

[0070] The electronic device obtains the power work order from the first device and executes the intelligent verification method for the power work order content as described in any of the first aspects and various embodiments above.

[0071] This application provides an intelligent verification method and system for power work orders. The method acquires the target content information of the power work order, then determines the corresponding target graph structure data based on this information, and finally inputs the target graph structure data into a pre-trained dependency verification model to obtain a first verification result. This first verification result indicates whether the power work order contains logical errors, missing dependencies, or dependency conflicts. The dependency verification model is trained based on graph structure data corresponding to historical power work orders and the corresponding verification results, and is used to obtain the dependency relationships between entities in the target content information. In this technical solution, the dependency verification model can determine the dependencies between entities in the power work order, identify whether there are logical errors, missing dependencies, or dependency conflicts, and efficiently assist users in determining the accuracy of the power work order. Attached Figure Description

[0072] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0073] Figure 1 A flowchart illustrating the intelligent verification method for power work order content provided in this application embodiment. Figure 1 ;

[0074] Figure 2 A flowchart illustrating the intelligent verification method for power work order content provided in this application embodiment. Figure 2 ;

[0075] Figure 3 A flowchart illustrating the intelligent verification method for power work order content provided in this application embodiment. Figure 3 ;

[0076] Figure 4 A flowchart illustrating the intelligent verification method for power work order content provided in this application embodiment. Figure 4 ;

[0077] Figure 5 A flowchart illustrating the intelligent verification method for power work order content provided in this application embodiment. Figure 5 ;

[0078] Figure 6 A schematic diagram of the structure of the intelligent verification system for power work order content provided in this application embodiment;

[0079] Figure 7 A schematic diagram of the intelligent verification device for power work order content provided in this application embodiment;

[0080] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0081] The accompanying drawings have illustrated specific embodiments of this disclosure, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concepts of this disclosure to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0082] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0083] Before introducing the embodiments of this application, the application background of the embodiments of this application will be explained first:

[0084] In the electrical field, a work permit is a written instruction that allows work to be carried out on electrical equipment and power lines. It is also a written basis for clarifying safety responsibilities, providing safety instructions to workers, and ensuring their safety. An operation permit is a written order for electrical operation personnel to perform operations. It is an important basis for preventing misoperation and ensuring personal safety, power grid safety, and equipment safety.

[0085] The management of work permits and operation permits serves as an important support for the work of various specialties. In actual operation and maintenance work, various task orders, on-duty records, safety measures cards, etc. are also used to support the work of these two permits.

[0086] In existing technologies, the manual filling of work tickets and operation tickets in the two-ticket management application is too much. It is generally based on manual verification, that is, relying on experienced power workers or safety management personnel to carefully review every field and detail of the work ticket to ensure that it complies with relevant safety specifications, operating procedures and legal requirements.

[0087] However, this method is time-consuming, labor-intensive, and inefficient, increasing the workload of frontline personnel; it lacks sufficient monitoring of on-site construction and cannot identify violations in a timely manner; and it does not make full use of intelligent equipment, failing to achieve real-time supervision and comparison.

[0088] To address the technical problems existing in the prior art, the inventors of this application propose the following: After obtaining an electronic work order, the content of the electronic work order can first be identified, treating the work order content as a graph structure, where nodes represent entities in the text (such as equipment, operations, safety measures, etc.), and edges represent the relationships between them (such as dependencies, conflicts, etc.). Then, based on a trained model, it can be determined whether there are logical errors, missing dependencies, or dependency conflicts. The training of this model can be based on the graph structure construction of historical electronic orders, and the model can be trained based on the corresponding results of historical electronic orders, for actual verification.

[0089] The parts not described in detail are disclosed in the following embodiments.

[0090] The technical solution of this application will now be described in detail through specific embodiments. It should be noted that the following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0091] It is worth noting that the application fields of the methods, apparatus, electronic devices, systems and storage media disclosed herein are not limited.

[0092] The subject of this application is an electronic device, specifically a terminal device, etc.

[0093] Figure 1 A flowchart illustrating the intelligent verification method for power work order content provided in this application embodiment. Figure 1 ,like Figure 1 As shown, the method may include the following steps:

[0094] Step 11: Obtain the target content information of the power work order;

[0095] The target content information can refer to the recording and management of various information in the maintenance and repair of power equipment, including equipment information, operation information, safety measures information, and working status.

[0096] Optionally, the following can also be done: input the target content information into the anomaly detection model to obtain the anomaly data corresponding to the power work order. The anomaly detection model is trained based on a large number of work orders and content information and is used to judge the anomalies in the target content information. Based on the anomaly data, the anomaly result of the target content information is determined.

[0097] In this embodiment, before inputting the target content information into the anomaly detection model, the target content information data needs to be preprocessed, including cleaning, standardizing or normalizing the data, removing noise and outliers, in order to ensure the quality and consistency of the data.

[0098] Then, a subsample is randomly selected from the target content information, and the data is recursively segmented. That is, a feature is randomly selected from the subsample, and a segmentation point is randomly selected within the value range of the feature for segmentation. Tree nodes are constructed based on the segmentation results, and the above steps are recursively repeated for each child node to construct an isolated tree, thereby constructing an isolated forest.

[0099] Furthermore, the path length of each data point in the target content information in each isolated tree is calculated using isolated forest. An anomaly threshold is set based on the path length output by the model, and an anomaly score is calculated based on the path length. If the anomaly score is greater than the preset anomaly threshold, then the data point is considered an anomaly.

[0100] Therefore, when the new power work order data is input into the anomaly detection model, the model will determine whether the data is abnormal based on the path length and give a corresponding anomaly score, thereby determining the abnormal result of the target content information, that is, determining whether the new work order data is abnormal.

[0101] It is understandable that by using anomaly detection models to detect anomalies in target content information, abnormal data in power work orders can be discovered in a timely manner, thereby identifying potential safety hazards and avoiding safety accidents caused by incorrect or incomplete work order information. Furthermore, anomaly detection models can effectively identify abnormal data points, thereby reducing errors caused by subjective human judgment and improving the accuracy and reliability of detection results.

[0102] In some embodiments, the distance between each data point in the target content information and its K nearest neighbors can be calculated using the K-Nearest Neighbors (KNN) method, and data points with larger distances can be regarded as anomalies, thereby completing the anomaly detection process.

[0103] Step 12: Based on the target content information of the power work order, determine the target diagram structure data corresponding to the target content information.

[0104] Graph structure data is a data representation method where data points (nodes) and the relationships between them (edges) are represented as a graph. Transforming the target content information of an electricity work order into graph structure data means representing each information item in the work order as a node in the graph, and the relationships between them as edges. Specifically, each information item in the electricity work order (such as equipment, operating procedures, safety measures, etc.) is defined as a node in the graph; edges between nodes are defined based on the relationships between the information items (such as the order of operating procedures, the association between safety measures and operating procedures, etc.). In some embodiments, attribute appending can also be applied to the graph structure data, that is, attaching relevant attribute information, such as node type and edge weight, to each node and edge.

[0105] For example, a power work order might contain the following information: Equipment A needs maintenance; operation steps 1, 2, and 3 need to be performed in sequence; and safety measures X and Y need to be completed before operation step 1. This information can be transformed into a corresponding graph structure: Nodes: Equipment A, Operation Step 1, Operation Step 2, Operation Step 3, Safety Measure X, Safety Measure Y; Edges: Operation Step 1 to Operation Step 2, Operation Step 2 to Operation Step 3, Safety Measure X to Operation Step 1, Safety Measure Y to Operation Step 1. This constructs a graph structure that reflects the content of the power work order for further analysis and processing.

[0106] Understandably, by converting the content information of power work orders into graph structure data, it is possible to more effectively identify anomalies in the work orders, such as missing key nodes or unreasonable connection methods of certain edges. On the other hand, through graph structure data analysis, bottlenecks and optimization points in the workflow can be discovered, thereby improving the workflow.

[0107] Step 13: Input the target graph structure data into the pre-trained dependency verification model to obtain the first verification result.

[0108] The first verification result is used to indicate whether there are logical errors, missing dependencies, or dependency conflicts in the power work order; the dependency verification model is a verification model trained based on the graph structure data corresponding to historical power work orders and the verification results corresponding to the graph structure data, used to obtain the dependency relationships between various entities in the target content information.

[0109] In this embodiment, the dependency verification model can be a Graph Neural Network (GNN), such as a Graph Convolutional Network (GCN) or a Graph Attention Network (GAT). By iteratively updating the node representations, the model can capture complex dependencies between nodes. Specifically, the target graph structure data of the current power work order is input into the pre-trained dependency verification model. The model analyzes the nodes and edges in the target graph structure data, checks the dependencies between them, and outputs the first verification result.

[0110] The first verification result output by the dependency verification model is used to indicate whether there are logical errors, missing dependencies, or dependency conflicts in the power work order. These problems correspond to the target graph structure data. Specifically, logical errors are manifested as logical inconsistencies or errors in the graph structure data, such as an operation step being executed without completing the preceding steps; missing dependencies are manifested as the absence of certain necessary dependencies in the graph structure data, such as an operation step lacking necessary safety measures; and dependency conflicts are manifested as the existence of conflicting dependencies in the graph structure data, such as a circular dependency between two operation steps.

[0111] It should be noted that the pre-training process of the dependency verification model includes: First, collecting a large amount of historical power work tickets and their corresponding graph structure data. Then, validating this graph structure data, identifying logical errors, missing dependencies, or dependency conflicts. Finally, using this labeled data to train the model, by learning and parsing the relationships in the graph, the model can capture complex dependencies between nodes, enabling it to automatically identify and verify dependencies in new graph structure data.

[0112] Therefore, after the model training is completed, the new power work order content is input into the model, and the model will verify the power work order based on the learned relationships to check for problems such as logical errors, missing dependencies, or conflicts.

[0113] In some embodiments, in addition to graph neural networks, dependency verification models can also be node embedding algorithms, such as DeepWalk, Node2Vec, LINE, etc., which capture the relationships between nodes by embedding nodes into a low-dimensional vector space, thereby using these embedded vectors for further dependency analysis and error detection.

[0114] Understandably, by inputting the target graph structure data into a pre-trained dependency verification model, logical errors, missing dependencies, and dependency conflicts in power work orders can be automatically identified. This process, based on historical data and machine learning models, can effectively learn and parse complex relationships in the graph, thereby improving the efficiency and accuracy of verification and ensuring the correctness and security of power work orders.

[0115] The intelligent verification method for power work orders provided in this application involves acquiring the target content information of the power work order, determining the target graph structure data corresponding to the target content information based on the target content information, and finally inputting the target graph structure data into a pre-trained dependency verification model to obtain a first verification result. This first verification result indicates whether the power work order contains logical errors, missing dependencies, or dependency conflicts. The dependency verification model is a verification model trained based on the graph structure data corresponding to historical power work orders and the verification results corresponding to the graph structure data, used to obtain the dependency relationships between various entities in the target content information. In this technical solution, the dependency verification model can determine the dependency relationships between various entities in the power work order, and determine whether the power work order contains logical errors, missing dependencies, or dependency conflicts, thus efficiently assisting users in determining the accuracy of the power work order.

[0116] Based on the above embodiments, the target graph structure data is the graph structure of an electricity work order; then Figure 2 A flowchart illustrating the intelligent verification method for power work order content provided in this application embodiment. Figure 2 ,like Figure 2 As shown, step 12 may include the following steps:

[0117] Step 21: Perform structured processing on the target content information to obtain the entities involved in the target content information and the relationships between the entities.

[0118] Structured processing refers to converting unstructured or semi-structured data (such as text, tables, etc.) into structured data. In this embodiment, structured processing means extracting information from the power work order and representing it in the form of a graph.

[0119] The entities involved in the target content information include equipment, tasks, personnel, time, and location. The relationships between these entities are of various types, such as a task being performed by a certain person or a device being used in a certain task.

[0120] It is understandable that by structuring the target content information, the information in the power work order can be better organized and managed, making it easier to query and analyze. In some embodiments, the obtained structured data can also be used for automated processing using algorithms, such as task scheduling and risk assessment.

[0121] Step 22: Treat the entities as nodes of the target graph structure data, and treat the relationships between the entities as edges of the target graph structure data.

[0122] In this embodiment, each entity can be represented as a node in the graph. For example, a specific device can be a node, and a specific task can be another node. Similarly, each relationship can be represented as an edge in the graph, connecting the relevant nodes. For instance, an edge between a task node and a personnel node can represent an "execution" relationship, and an edge between a task node and a device node can represent a "use" relationship.

[0123] Understandably, by transforming entities and the relationships between them into nodes and edges of the target graph structure data, the organization of the data becomes clearer and easier to understand and manage.

[0124] The intelligent verification method for power work orders provided in this application involves structuring the target content information to obtain the entities involved and the relationships between them. Then, the entities are used as nodes in the target graph structure data, and the relationships between the entities are used as edges. This technical solution implements the graph structure construction of the power work order, providing a foundation for accurately determining the verification results.

[0125] Based on the above embodiments, the target graph structure data is the graph structure of an electricity work order; then Figure 3 A flowchart illustrating the intelligent verification method for power work order content provided in this application embodiment. Figure 3 ,like Figure 3 As shown, after step 11, the following steps may be included:

[0126] Step 31: Input the target content information of the power work order into the pre-trained preliminary verification model to obtain the second verification result.

[0127] The second verification result is used to identify key information in the power work order in order to determine whether there are errors in the format or content of the power work order.

[0128] In this embodiment, the pre-trained preliminary verification model can refer to the Transformer model. Therefore, it is necessary to first convert the target content information of the power work order into a format that the model can process, such as converting text information into word vectors or embedding representations. Then, the Transformer model is used to extract features of the input data through a self-attention mechanism, capture key information and relationships, and generate a second verification result, including the identified key information, such as equipment name, task time, operation steps, etc., thereby further identifying possible format or content errors in the power work order.

[0129] Understandably, Transformer models are typically pre-trained on large-scale data, and process the input data based on self-attention mechanisms and deep learning capabilities. At the same time, they can process the input data in parallel, thus improving the speed and efficiency of verification.

[0130] Step 32: Extract features from the target content information of the power work order to obtain vector features of each part of the target content information.

[0131] Among them, feature extraction of the target content information of the power work order can be carried out by using natural language processing (NLP) techniques, such as word embeddings, term frequency-inverse document frequency (TF-IDF), and bidirectional encoder representations from transformers (BERT), to convert the text into a vector representation.

[0132] In this embodiment, the sentence vector representation function of BERT is used to extract features from the target content information (such as work tasks, safety measures, etc.) of the power work order. The result of feature extraction is usually represented in the form of vector, i.e. vector features. At the same time, multiple feature extractions are performed on the target content information of the power work order to finally obtain a comprehensive vector feature, i.e. a high-dimensional vector, which represents all the key information of this power work order.

[0133] Furthermore, the method also includes: inputting vector features into a pre-trained semantic understanding model to obtain a third verification result. The third verification result is used to describe the rationality and consistency of the logical relationship and contextual dependency between various fields in the target content information. The semantic understanding model is constructed using an attention mechanism and trained based on the vector features corresponding to historical power work tickets and the verification results corresponding to historical power work tickets.

[0134] In this embodiment, the pre-trained semantic understanding model can refer to a customized model specifically for electricity work order content, based on Transformer architectures such as BERT or Generative Pre-trained Transformer (GPT). This model contains more layers and larger embedding dimensions to capture more complex semantic information and contextual relationships. Specifically, using a self-attention mechanism to capture important information and long-distance dependencies in the input data (vector features) can enhance the model's focus on key information. By calculating the attention weights between each word and other words and dynamically assigning weights to different parts, the model can focus on the key parts of the input sequence.

[0135] Meanwhile, the model is trained based on a large amount of historical power work order data and its corresponding vector features and verification results. By learning from this historical data, the model can capture common patterns and logical relationships in power work orders. Specifically, the model is trained using a professional corpus containing a large amount of power work order text. Through two stages, pre-training and fine-tuning, the model can accurately understand the professional terminology, work processes, and safety regulations of the power industry.

[0136] For example, suppose an electricity work order contains information about the task, equipment, time, and personnel. The feature extraction process converts this information into vector features, which are then input into a semantic understanding model. The model might check the following logical relationships and contextual dependencies: whether the "task" is reasonably associated with the "equipment"; whether the "time" is within a reasonable working timeframe; and whether the "personnel" are qualified to perform the "task." If the model finds that the time is not within a reasonable range or that the personnel are not qualified to perform the task, it will mark these errors in the third verification result.

[0137] Understandably, this semantic understanding model can automatically identify key information such as equipment names, operating instructions, and safety measures, ensuring the accuracy and completeness of this information in the work order. Simultaneously, through deep learning techniques such as self-attention mechanisms, the model can capture the logical relationships and contextual dependencies between various fields in the power work order, thereby judging their consistency and rationality. This enables the model to efficiently and accurately identify logical errors and inconsistencies in the power work order, detecting potential problems such as missing key information, non-standard filling methods, or potential safety hazards, improving the accuracy and efficiency of verification, and enhancing security and compliance. Furthermore, by automatically prompting these issues, staff can promptly modify and improve the work order, providing crucial support for the efficient operation and maintenance of the power system.

[0138] Step 33: Determine the semantic similarity between each vector feature according to the preset similarity algorithm. The semantic similarity is used to describe the consistency, duplicate content, and similar content between each field in the target content information.

[0139] In this embodiment, the preset similarity algorithm can be a measurement method such as cosine similarity or Euclidean distance. During the verification process, it is used to verify the consistency between fields, detect duplicate or similar content, etc. Specifically, by calculating the semantic similarity between different fields in the work order, it can be determined whether they are semantically consistent. For example, whether the description in the safety measures field matches the description of the work task, and whether there are semantic conflicts or inconsistencies. Furthermore, in large power systems, multiple work orders may involve similar operations or equipment. By calculating the semantic similarity between these work orders, duplicate or similar content can be detected, thereby avoiding redundancy and confusion. In addition, power work orders typically follow certain templates and formats. By calculating the semantic similarity between the work order and the standard template, it can be used to help determine whether the power work order meets the prescribed template and format requirements.

[0140] It should be noted that cosine similarity measures the cosine of the angle between two vectors and determines the semantic similarity between the features of each vector based on the cosine of the angle. Euclidean distance measures the straight-line distance between two vectors; the smaller the value, the more similar the two vectors are.

[0141] Understandably, by determining the semantic similarity between various vector features, we can further verify the consistency between various fields in the work order, detect duplicate or similar content, and assist in template matching based on the preliminary verification model. By achieving in-depth verification, we can ensure the accuracy and completeness of the work order content.

[0142] Step 34: Based on the second verification result and semantic similarity, conduct a comprehensive evaluation of the power work order.

[0143] Understandably, by using a preliminary verification model to perform initial verification of the target content information of the power work order, a second verification result can be obtained. This can identify obvious errors in format or content, ensuring the elimination of basic errors and improving the accuracy of subsequent processing. By calculating the semantic similarity between various vector features using a pre-set similarity algorithm, potential issues such as content duplication, similarity, or inconsistency can be detected. Therefore, based on the second verification result and semantic similarity, a comprehensive evaluation of the power work order can be conducted, combining format errors, content errors, and semantic similarity to provide a holistic assessment result.

[0144] Furthermore, for the above implementation process, homomorphic encryption or differential privacy can also be used to encrypt the target verification result, which includes: a first verification result, a second verification result, and a third verification result.

[0145] In this embodiment, to protect data privacy, homomorphic encryption or differential privacy can be employed. Specifically, homomorphic encryption refers to using fully homomorphic or semi-homomorphic encryption to protect the security of sensitive data, allowing computations to be performed on encrypted data without decryption, thereby protecting data privacy. For example, the first, second, and third verification results can be directly used for verification in the encrypted state without exposing the original data. Differential privacy refers to protecting individual data privacy by adding noise. By setting a reasonable privacy budget, the relationship between privacy protection and data utility is balanced, making it difficult to deduce the original data even if the verification results are obtained.

[0146] Understandably, by employing homomorphic encryption or differential privacy technology, it is possible to prevent unauthorized access or tampering of power work tickets during transmission and storage, ensuring the confidentiality and integrity of data, while reducing the risk of human error and improving the overall security of the system.

[0147] The intelligent verification method for power work orders provided in this application involves inputting the target content information of the power work order into a pre-trained preliminary verification model to obtain a second verification result. This second verification result is used to identify key information in the power work order, determining format or content errors. Feature extraction is then performed on the target content information to obtain vector features of each part. Following this, a preset similarity algorithm is used to determine the semantic similarity between these vector features. The semantic similarity describes the consistency, repetition, and similarity of fields within the target content information. Finally, a comprehensive evaluation of the power work order is performed based on the second verification result and the semantic similarity. This technical solution, by combining the results of preliminary verification and semantic similarity, provides a comprehensive evaluation of the work order and can offer user modification suggestions or risk warnings.

[0148] Based on the above embodiments, the target graph structure data is the graph structure of an electricity work order; then Figure 4 A flowchart illustrating the intelligent verification method for power work order content provided in this application embodiment. Figure 4 ,like Figure 4 As shown, the method may also include the following steps:

[0149] Step 41: Determine the optimal solution set among multiple optimization objectives.

[0150] In this embodiment, for the dependency verification model, the optimization objectives can include minimizing dependency conflicts, maximizing task parallelism, and minimizing operation time. The optimal solution set is to find a balance point among these optimization objectives, so that dependencies are handled optimally, reducing conflicts and delays.

[0151] For semantic understanding models, optimization objectives can include improving semantic parsing accuracy, reducing misunderstandings and ambiguities, and increasing processing speed. The optimal solution set is to find a balance between these optimization objectives, so that semantic understanding is both accurate and efficient.

[0152] For the initial verification model, optimization objectives may include improving verification accuracy, reducing false positives and false negatives, and increasing verification speed. The optimal solution set is to find a balance between these optimization objectives, making the initial verification more accurate.

[0153] Furthermore, the optimal solution set among these objectives can be found by employing the Non-Dominated Sorting Genetic Algorithm II (NSGA-II), which balances multiple objectives through fast non-dominated sorting, crowding distance calculation, and elite retention strategies.

[0154] Understandably, by determining the optimal solution set among multiple optimization objectives, the model can adapt to complex and changing environments, thereby improving the system's adaptability and flexibility.

[0155] Step 42: During multiple runs of the target model, adjust the parameters of the target model in real time to obtain different output solution sets;

[0156] The target model can be any one of the following: dependency verification model, semantic understanding model, and preliminary verification model.

[0157] In this embodiment, during the algorithm's operation, it is necessary to collect output results and related performance indicators (such as accuracy, efficiency, error rate, etc.), and use this feedback to adjust model parameters and algorithm parameters in real time, such as population size, crossover probability, mutation probability, etc., in order to optimize algorithm performance.

[0158] Understandably, by adjusting parameters in real time, the model can better adapt to different input data and operating environments, improve the accuracy and reliability of the output results, and adapt to different scenarios and changing conditions, thereby improving the robustness and adaptability of the model and reducing dependence on specific parameter settings.

[0159] Step 43: Evaluate different output solution sets based on the optimal solution sets among multiple optimization objectives to obtain the optimal objective model.

[0160] In this embodiment, for multi-objective optimization, there are solution sets that cannot improve one objective without worsening others. These solution sets represent the optimal trade-offs among the various objectives. Therefore, different output solution sets are evaluated based on multiple optimization objectives. By comparing the performance of these solution sets on each objective, it is determined which solution sets belong to the optimal solution set and are selected as the optimal objective model. That is, the solution that best meets the current requirements is selected as the final solution for power work order verification.

[0161] Understandably, by determining the optimal solution set among multiple optimization objectives, a balance point can be found among different models, enabling the system to achieve optimal overall performance, rather than just optimizing a single model. Furthermore, by optimizing the model, data-driven methods can be used for decision-making, ensuring that the system is more intelligent and efficient in handling complex tasks.

[0162] In the above implementation, the multi-objective optimization approach comprehensively considers the output results of these different objective models. By balancing various optimization objectives (such as improving verification accuracy, reducing false alarm rate, and accelerating processing speed), the optimal verification strategy or parameter configuration is found. In this way, the entire verification system can more comprehensively and accurately verify the content of power work orders, thereby improving the safety production level of the power industry.

[0163] The intelligent verification method for power work orders provided in this application determines the optimal solution set among multiple optimization objectives for a target model. The target model can be any one of a dependency verification model, a semantic understanding model, and a preliminary verification model. Then, during multiple runs of the target model, the parameters of the target model are adjusted in real time to obtain different output solution sets. Finally, the different output solution sets are evaluated based on the optimal solution set among the multiple optimization objectives to obtain the optimal target model. This technical solution utilizes objective optimization to verify the dependency verification model, the semantic understanding model, and the preliminary verification model, thereby obtaining more accurate verification results.

[0164] Based on the above embodiments, the target graph structure data is the graph structure of an electricity work order; then Figure 5 A flowchart illustrating the intelligent verification method for power work order content provided in this application embodiment. Figure 5 ,like Figure 5 As shown, the method may also include the following steps:

[0165] Step 51: Determine the time series analysis information corresponding to the first power work order within the first preset time period before the current time node.

[0166] The current time point refers to the current time of the system, that is, the moment when analysis and decision-making are carried out.

[0167] The first preset duration refers to a fixed time period tracing back from the current point in time. For example, this period could be the past 24 hours, a week, a month, etc., and the specific duration is set according to business needs.

[0168] The first power work permit refers to a specific power work permit within a preset time period.

[0169] Time series analysis information refers to the information obtained by analyzing the time series data of power work orders within a preset time period. This information can include the submission time, approval time, execution time, completion time of the work order, as well as the intervals and trends between these time points.

[0170] In this embodiment, a time series analysis of historical work order data can be performed using an Autoregressive Integral Moving Average (ARIMA) model to predict the future quantity and type of work orders. Specifically, historical data on power work orders needs to be collected, including quantity, type, and processing time. Then, time series analysis is performed based on this data to construct a predictive model, thereby estimating the model's parameters, such as the difference order, autoregressive coefficients, and moving average coefficients, using historical data. Finally, the constructed model is used to predict the future quantity and type of work orders, providing a reference for the allocation and scheduling of power work orders.

[0171] Understandably, time series analysis can identify bottlenecks and delays in the power work order processing, optimize workflows, and improve overall work efficiency. At the same time, time series analysis can help predict future workload and resource requirements, allowing for advance preparation and avoiding resource waste and work backlog.

[0172] Step 52: Input the time series analysis information into the prediction model to obtain the prediction results.

[0173] The prediction results describe the number and type of second power work orders within a second preset time period after the current time node; the prediction model is trained based on historical power work orders.

[0174] Understandably, by using historical data and time series analysis information, predictive models can identify patterns and trends in the data, thereby improving the accuracy of predictions. Based on the prediction results, potential risks and problems can be identified, preventive measures can be taken in advance, the occurrence of failures and accidents can be reduced, and the safety and reliability of the system can be improved.

[0175] Optionally, the target model can be optimized based on the abnormal results to obtain an optimized target model.

[0176] It should be noted that abnormal results refer to unusual situations or deviations from expected results that occur during model execution. These anomalies may be caused by factors such as data noise, improper model parameter settings, or changes in the external environment. By analyzing abnormal results, the root causes of the anomalies can be identified, such as by checking data quality, model parameters, and model structure. Based on the diagnostic results, the target model can be optimized and adjusted, including retraining the model, adjusting model parameters, improving the model structure, and introducing new features.

[0177] Understandably, after optimization, the model is validated to ensure that its performance is improved, outliers are reduced, and its performance and stability are enhanced, enabling it to better handle abnormal situations and provide more accurate and reliable results.

[0178] The intelligent verification method for power work orders provided in this application determines the time series analysis information corresponding to the first power work order within a first preset time period before the current time node. This time series analysis information is then input into a prediction model to obtain a prediction result. The prediction result describes the quantity and type of the second power work order within a second preset time period after the current time node. The prediction model is trained based on historical power work orders. This technical solution enables the prediction of future electronic work orders, providing a reference for the allocation and scheduling of power work orders.

[0179] Optionally, based on the above embodiments, a reinforcement learning architecture can also be introduced, that is, defining a state space (such as the content of the current electronic work order, historical verification results, etc.) and an action space (such as adjusting the verification strategy, modifying parameters, etc.).

[0180] One approach is to train a policy network using the Q-learning algorithm, enabling it to select the optimal action based on the current state. Through continuous trial and error and feedback, the policy network gradually learns the optimal verification policy. During the verification process, the agent can obtain information on the effectiveness and problems of the current policy through a real-time feedback mechanism. Based on this feedback, the agent can dynamically adjust its behavioral policy to better adapt to the requirements of the verification task.

[0181] For different types of verification tasks, multiple strategies can be designed and integrated or switched according to actual needs during the verification process. This method of integrating and switching multiple strategies can further improve the flexibility and accuracy of verification.

[0182] In complex verification scenarios, a single agent may be insufficient to handle all tasks. In such cases, reinforcement learning algorithms involving multi-agent collaboration can be introduced. Through cooperation and competition among multiple agents, a more comprehensive and accurate verification of electronic work order content can be achieved.

[0183] For example, an agent system allows agents to learn and make decisions independently in a distributed environment, while simultaneously achieving information sharing and policy coordination through communication and collaboration mechanisms. In a multi-agent system, agents can collaborate to complete tasks or compete to improve their individual learning efficiency and performance. This combination of collaboration and competition can stimulate the learning potential of agents and drive continuous optimization of the verification task. In reinforcement learning algorithms involving multi-agent collaboration, the policies of each agent need to be collaboratively optimized to achieve the optimal performance of the overall system. This can be achieved by designing reasonable reward functions and collaboration mechanisms, thereby ensuring the accuracy and efficiency of the verification task. Deep learning and other technologies can be used to achieve knowledge sharing and transfer among agents. By sharing knowledge bases or model parameters, the learning ability and generalization ability of agents can be improved, thereby achieving better handling of complex verification tasks. Therefore, the aforementioned dependency verification model, preliminary verification model, semantic understanding model, and multi-objective optimization can all operate as independent agents, thus forming an agent system to achieve more comprehensive and efficient verification.

[0184] Based on a dependency verification model, a preliminary verification model, a semantic understanding model, and a combination of multi-objective optimization and reinforcement learning, an efficient and accurate intelligent verification method for power work orders can be constructed. This method can automatically identify errors and abnormal content in work orders, improve verification efficiency and accuracy, and provide strong protection for the safe production of enterprises.

[0185] In the above embodiments, Figure 6 This is a schematic diagram of the intelligent verification system for power work order content provided in the embodiments of this application, as shown below. Figure 6 As shown, the system includes: an electronic device 61 and a first device 62;

[0186] Electronic device 61 obtains the power work order from the first device 62 and executes the intelligent verification method for the power work order content of any of the above method embodiments.

[0187] The intelligent verification system for power work orders provided in this application is based on the principles and technical effects of the above embodiments, and will not be repeated here.

[0188] The following are device embodiments of the present application, which can be used to implement the method embodiments of the present application. For details not disclosed in the device embodiments of the present application, please refer to the method embodiments of the present application.

[0189] Figure 7 This is a schematic diagram of the intelligent verification device for power work order content provided in an embodiment of this application. Figure 7 As shown, the device includes:

[0190] Module 71 is used to obtain the target content information of the power work order;

[0191] The determination module 72 is used to determine the target diagram structure data corresponding to the target content information based on the target content information of the power work order;

[0192] The processing module 73 is used to input the target graph structure data into a pre-trained dependency verification model to obtain a first verification result. The first verification result is used to indicate whether there are logical errors, missing dependencies, or dependency conflicts in the power work order. The dependency verification model is a verification model trained based on the graph structure data corresponding to historical power work orders and the verification results corresponding to the graph structure data, used to obtain the dependency relationships between various entities in the target content information.

[0193] In one or more embodiments, the target graph structure data is a graph structure of an electrical work order;

[0194] Accordingly, module 72 is specifically used for:

[0195] The target content information is structured to obtain the entities involved in the target content information and the relationships between the entities;

[0196] Entities are treated as nodes in the target graph structure data, and the relationships between entities are treated as edges in the target graph structure data.

[0197] In one or more embodiments, after obtaining the target content information of the power work order, the processing module 73 is further configured to:

[0198] The target content information of the power work order is input into a pre-trained preliminary verification model to obtain a second verification result. The second verification result is used to identify key information in the power work order to determine whether there are format or content errors in the power work order.

[0199] Feature extraction is performed on the target content information of the power work order to obtain the vector features of each part of the target content information;

[0200] Based on the preset similarity algorithm, the semantic similarity between each vector feature is determined. The semantic similarity is used to describe the consistency, duplicate content, and similar content among the fields in the target content information.

[0201] Based on the second verification result and semantic similarity, a comprehensive evaluation of the power work order is conducted.

[0202] In one or more embodiments, after extracting features from the target content information of the power work order to obtain vector features of each part of the target content information, the processing module 73 is further configured to:

[0203] The vector features are input into a pre-trained semantic understanding model to obtain a third verification result. The third verification result is used to describe the rationality and consistency of the logical relationships and contextual dependencies between various fields in the target content information. The semantic understanding model is constructed using an attention mechanism and trained based on the vector features corresponding to historical power work tickets and the verification results corresponding to historical power work tickets.

[0204] In one or more embodiments, the processing module 73 is further configured to:

[0205] Determine the optimal solution set among multiple optimization objectives;

[0206] During multiple runs of the target model, the parameters of the target model are adjusted in real time to obtain different output solution sets. The target model can be any one of the dependency verification model, semantic understanding model, and preliminary verification model.

[0207] The optimal target model is obtained by evaluating different output solution sets based on the optimal solution sets among multiple optimization objectives.

[0208] In one or more embodiments, the processing module 73 is further configured to:

[0209] Determine the time series analysis information corresponding to the first power work order within the first preset time period before the current time node;

[0210] The time series analysis information is input into the prediction model to obtain the prediction results. The prediction results describe the number and type of the second power work order within the second preset time period after the current time node. The prediction model is trained based on historical power work orders.

[0211] In one or more embodiments, the processing module 73 is further configured to:

[0212] The target content information is input into the anomaly detection model to obtain the abnormal data corresponding to the power work order. The anomaly detection model is trained from the content of multiple work orders and is used to judge the anomalies in the target content information.

[0213] Based on the abnormal data, determine the abnormal results of the target content information.

[0214] In one or more embodiments, the processing module 73 is further configured to:

[0215] Based on the abnormal results, the target model is optimized to obtain the optimized target model.

[0216] In one or more embodiments, the processing module 73 is further configured to:

[0217] The target verification result is encrypted using homomorphic encryption or differential privacy. The target verification result includes: a first verification result, a second verification result, and a third verification result.

[0218] The apparatus provided in this application embodiment can be used to execute the methods in any of the above embodiments, and its implementation principle and technical effect are similar, so they will not be described again here.

[0219] It should be noted that the division of the various modules in the above device is merely a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, these modules can be implemented entirely in software via processing element calls; they can be fully implemented in hardware; or some modules can be implemented in software via processing element calls, while others are implemented in hardware. Additionally, these modules can be fully or partially integrated together, or implemented independently. The processing element here can be an integrated circuit with signal processing capabilities. During implementation, each step of the above method or each of the above modules can be completed through the integrated logic circuits in the hardware of the processor element or through software instructions.

[0220] Figure 8 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application, such as... Figure 8 As shown, the electronic device may include: a processor 81, a memory 82, and computer program instructions stored in the memory 82 and executable on the processor 81. When the processor 81 executes the computer program instructions, it implements the method provided in any of the foregoing embodiments.

[0221] Optionally, the various components of the electronic device can be connected via a system bus.

[0222] The memory 82 can be a separate memory unit or a memory unit integrated into the processor 81. The number of processors 81 can be one or more.

[0223] It should be understood that the processor 81 can be a Central Processing Unit (CPU), or other general-purpose processors 81, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor 81 can be a microprocessor 81, or any conventional processor 81. The steps of the method disclosed in this application can be directly manifested as being executed by the hardware processor 81, or being executed by a combination of hardware and software modules within the processor 81.

[0224] The system bus can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The system bus can be divided into address bus, data bus, control bus, etc. For ease of representation, only one thick line is used in the diagram, but this does not indicate that there is only one bus or one type of bus. Memory 82 may include random access memory (RAM) 82, and may also include non-volatile memory (NVM) 82, such as at least one disk storage device 82.

[0225] All or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a readable memory 82. When the program is executed, it performs the steps of the above method embodiments; and the aforementioned memory 82 (storage medium) includes: read-only memory 82 (ROM), RAM, flash memory 82, hard disk, solid-state hard disk, magnetic tape, floppy disk, optical disk, and any combination thereof.

[0226] The electronic device provided in this application embodiment can be used to execute the method provided in any of the above method embodiments. Its implementation principle and technical effect are similar, and will not be repeated here.

[0227] This application provides a computer-readable storage medium storing computer instructions that, when executed on a computer, cause the computer to perform the above-described method.

[0228] The aforementioned computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory, electrically erasable programmable read-only memory, erasable programmable read-only memory, programmable read-only memory, read-only memory, magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0229] Optionally, a readable storage medium can be coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Alternatively, the readable storage medium can be an integral part of the processor. Both the processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components within the device.

[0230] This application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium, and the at least one processor can implement the above-described method when executing the computer program.

[0231] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for intelligent verification of the contents of an electricity work order, characterized in that, include: Obtain the target content information of the power work order; Based on the target content information of the power work order, determine the target graph structure data corresponding to the target content information; The target graph structure data is input into a pre-trained dependency verification model to obtain the first verification result. The first verification result is used to indicate whether the power work order has logical errors, missing dependencies, or dependency conflicts; The dependency verification model is a verification model trained based on the graph structure data corresponding to historical power work tickets and the verification results corresponding to the graph structure data, used to obtain the dependency relationships between various entities in the target content information; After obtaining the target content information of the power work order, the method further includes: The target content information of the power work order is input into a pre-trained preliminary verification model to obtain a second verification result; the second verification result is used to identify key information in the power work order to determine whether there are format or content errors in the power work order. Feature extraction is performed on the target content information of the power work order to obtain vector features of each part of the target content information; Based on a preset similarity algorithm, the semantic similarity between each vector feature is determined. The semantic similarity is used to describe the consistency, duplicate content, and similar content among the fields in the target content information. The power work order is comprehensively evaluated based on the second verification result and the semantic similarity. After extracting features from the target content information of the power work order to obtain vector features of each part of the target content information, the method further includes: The vector features are input into a pre-trained semantic understanding model to obtain a third verification result. The third verification result is used to describe the rationality and consistency of the logical relationship and contextual dependency between various fields in the target content information. The semantic understanding model is constructed using an attention mechanism and trained based on the vector features corresponding to historical power work tickets and the verification results corresponding to historical power work tickets. The target graph structure data is the graph structure of the power work order; Accordingly, determining the target graph structure data corresponding to the target content information based on the target content information of the power work order includes: The target content information is structured to obtain the entities involved in the target content information and the relationships between the entities; Entities are treated as nodes in the target graph structure data, and the relationships between entities are treated as edges in the target graph structure data.

2. The method according to claim 1, characterized in that, The method further includes: Determine the optimal solution set among multiple optimization objectives; During multiple runs of the target model, the parameters of the target model are adjusted in real time to obtain different output solution sets. The target model can be any one of the dependency verification model, semantic understanding model, and preliminary verification model. The optimal target model is obtained by evaluating different output solution sets based on the optimal solution set among the multiple optimization objectives.

3. The method according to claim 1, characterized in that, The method further includes: Determine the time series analysis information corresponding to the first power work order within the first preset time period before the current time node; The time series analysis information is input into the prediction model to obtain the prediction result, which describes the quantity and type of the second power work order within a second preset time period after the current time node; the prediction model is trained based on historical power work orders.

4. The method according to claim 2, characterized in that, The method further includes: The target content information is input into the anomaly detection model to obtain the abnormal data corresponding to the power work order. The anomaly detection model is trained from the content of multiple work orders and is used to determine the anomalies present in the target content information. Based on the abnormal data, the abnormal result of the target content information is determined.

5. The method according to claim 4, characterized in that, The method further includes: Based on the abnormal results, the target model is optimized to obtain the optimized target model.

6. The method according to claim 5, characterized in that, The method further includes: The target verification result is encrypted using homomorphic encryption or differential privacy. The target verification result includes: a first verification result, a second verification result, and a third verification result.

7. An intelligent verification system for the contents of power work permits, characterized in that, include: Electronic devices and primary devices; The electronic device obtains the power work order from the first device and executes the intelligent verification method for the content of the power work order as described in any one of claims 1-6.

Images