Intelligent management method and system for power grid field investigation based on artificial intelligence

Abstract

The invention discloses a power grid field investigation intelligent management method and system based on artificial intelligence. The method comprises the steps of obtaining multi-source data and writing the multi-source data into a graph database; obtaining candidate workplaces based on associated data and rules of the equipment and the places, and generating a workplace candidate set; generating a work content set corresponding to the target exploration list; calculating and generating a risk assessment result; generating a power failure range list based on the graph database; outputting a safety measure text in combination with the work content set and the power failure range list; acquiring field data and exploration content to obtain field acquisition data; performing consistency and integrity verification on the field acquisition data and other data to generate a verification result; and aggregating the result and the data to generate a report file. Through the technical scheme of the invention, the working efficiency of on-site investigation is greatly improved, the accuracy of a power failure range and the comprehensiveness of safety measures are ensured, the time for manually filling a report is shortened, and the standardization and accuracy of report contents are ensured.

Description

Technical Field

[0001] This invention relates to the field of power industry management technology, and in particular to an intelligent management method for power grid field survey based on artificial intelligence and an intelligent management system for power grid field survey based on artificial intelligence. Background Technology

[0002] With the continuous expansion of the power grid, especially in the substation, transmission, and distribution sectors, the complexity of on-site surveys is also gradually increasing. Traditional on-site survey management methods rely on manual operation and often use paper or simple electronic forms for recording, which faces the following challenges: (1) Low efficiency of data statistics and processing: Existing survey data processing often relies on manual filling and calculation. The data volume is large and has many dimensions, especially the multi-professional and multi-level power grid survey data (such as equipment ledgers, historical survey forms, work plans, etc.). In the absence of efficient calculation tools, the traditional statistical process is very time-consuming and prone to human error, affecting the work progress and data accuracy.

[0003] (2) Large errors in filling in the survey content: The survey content, such as equipment information, survey results, risk level, and pre-control measures, often depends on the experience of the survey personnel. Different personnel may have different descriptions of the same equipment, resulting in non-standard survey content and missing or incorrect data. The manual filling process is cumbersome and prone to omissions and errors, which leads to inaccuracies in the subsequent maintenance plan preparation and work order issuance.

[0004] (3) Risk assessment relies on experience: Risk level assessment in on-site surveys usually relies on the experience of the person in charge of the survey or the on-site personnel. Different persons in charge have different judgments on the weight of assessment dimensions such as voltage level and scope of work, which leads to deviations in risk assessment results, affects the safety management of operations, and may result in high-risk scenarios not being identified in time.

[0005] (4) Lack of closed-loop control in the exploration process: In the traditional exploration process, there are information silos in each link, data flow is not smooth, and there is a lack of a unified control mechanism. When exploration forms are transferred between different departments, problems such as data lag and process delays often occur, making it difficult to achieve full-process tracking and feedback updates.

[0006] To address the aforementioned issues, there is an urgent need for an intelligent management system for field surveys based on artificial intelligence (AI) technology. This system should be able to improve the efficiency of data statistics and processing through intelligent means, automatically generate survey content, reduce human error, and achieve accurate risk level determination and matching of pre-control measures through multi-dimensional risk assessment and dynamic rule base support, ultimately improving the efficiency and accuracy of power grid field survey operations. Summary of the Invention

[0007] To address the aforementioned issues, this invention provides an intelligent management method for power grid field surveys based on artificial intelligence. Through automated data collection, work content generation, risk assessment, and report generation, it significantly improves the efficiency of field surveys. Intelligent analysis based on historical data and association rules automatically generates work content consistent with the survey task, automatically generates a list of power outage areas and safety measures texts, ensuring the accuracy of power outage areas and the comprehensiveness of safety measures. The automatic generation of standardized survey reports reduces the time spent manually filling out reports, ensuring the standardization and accuracy of the report content.

[0008] To achieve the above objectives, the present invention provides an intelligent management method for power grid field survey based on artificial intelligence, comprising: Acquire equipment ledgers, historical survey reports, work plans, GIS spatial and equipment status data, and write them into a graph database to form a basic dataset; Based on the association data and rules between devices and locations in the aforementioned basic dataset, candidate work locations are obtained, and a candidate work location set is generated. By combining the basic dataset and the candidate work location set, a work content set corresponding to the target survey order is generated; Based on the basic dataset and the work content set, a scoring detail and risk level are calculated, and a risk assessment result is generated accordingly. Based on the device topology relationships in the graph database, a list of power outage areas is generated according to the work objects in the work content set; With the support of the rule base and terminology base, the safety measures text corresponding to the hazard points and prevention and control measures are output by combining the work content set and the power outage range list; Collect on-site images, videos, environmental and equipment operation data, and obtain the survey content submitted by the survey personnel. Combine the collected data with the survey content to obtain the on-site collected data. The consistency and integrity of the field-collected data are verified with the basic dataset, the work content set, and the safety measures text, and verification results are generated. The basic dataset, the candidate set of work locations, the set of work content, the risk assessment results, the list of power outage areas, the safety measures text, the on-site collected data, and the verification results are aggregated to generate a report file associated with the target survey form.

[0009] In the above technical solution, preferably, the specific process of acquiring equipment ledgers, historical survey reports, work plans, GIS spatial and equipment status data, and writing them into the graph database includes: Equipment ledger data is extracted from the power grid equipment management system. The equipment ledger data includes equipment number, usage time, voltage level, equipment type, equipment model, equipment maintenance record, operating status, bay information, and equipment installation location. Historical survey order data is extracted from the historical survey management system. The historical survey order data includes survey location, survey equipment name, maintenance type, risk level, pre-control measures, and process status information. Extract work plan data from the work planning system, the work plan data including planned time, person in charge, and work scope information; Obtain GIS geospatial data from the GIS platform, wherein the GIS geospatial data includes the geographical location information of the device and the topological relationship of the device; The equipment ledger data, the historical survey data, the work plan data, and the GIS geospatial data are standardized and written into a graph database, which is a Neo4j graph database. When storing data, a model of nodes and relationships is established, and the equipment ledger data, the historical survey form data, and the work plan data are stored as different node types respectively. Extract the relationships between different nodes, and construct the relationships between nodes based on the relationships.

[0010] In the above technical solution, preferably, the specific process of generating the candidate set of work locations includes: Based on the "exploration equipment-work location" association data in the historical survey data, the association rule mining algorithm is used to train the data, extract the association rules between the exploration equipment and the work location, and generate a rule base that can be used for work location prediction. When the survey equipment information is input, a candidate set of work locations is automatically generated based on the trained rule base, and sorted according to the matching confidence. The work location with the highest confidence is output as the suggested survey location. The steps of the association rule mining algorithm include: Preprocess historical survey data to extract effective correlation information between survey equipment and work sites; The frequent itemset mining method is used to extract the combination of exploration equipment and work location from historical exploration data, calculate the corresponding confidence level, and generate association rules. Based on the association rules, the device information is automatically matched, and a candidate set of work locations that meet the conditions is output. If the input exploration equipment information cannot be matched in the existing rules, a prompt message will be sent to the exploration personnel to manually input the work location. New combinations of exploration equipment and work locations will be automatically added to the rule base to update and optimize the generation of subsequent candidate sets.

[0011] In the above technical solution, preferably, the process of generating the work content set specifically includes: Based on the survey equipment, work location, and maintenance type, a work content prediction model is constructed. The work content prediction model is trained using the "survey equipment-work location-maintenance type-work content" data from historical survey orders to extract the prediction rules for the work content. The training of the work content prediction model adopts the gradient boosting tree algorithm. The feature variables in the training process include: the type of survey equipment, the specific information of the work location, the maintenance type, the operating status of the equipment, and the job type. After inputting the survey equipment, work location, and maintenance type information, the work content prediction model generates corresponding work content information based on the trained rules. The work content information includes maintenance items, work steps, and work requirements. The work content information is filtered and optimized, and the final work content is determined based on the matching degree and operability of the work steps and requirements. If the generated work content fails to pass the optimization screening, a prompt message will be sent to the survey personnel to manually fill in the work content. The new combination of "survey equipment-work location-maintenance type-work content" will be automatically included in the subsequent rule base to continuously update and optimize the work content prediction model.

[0012] In the above technical solution, preferably, the specific process of generating the risk assessment result includes: A multi-dimensional risk assessment model is constructed based on preset assessment dimensions, and the weight coefficient of each dimension is calculated using the analytic hierarchy process. According to the evaluation dimensions, scores are given based on the preset scoring criteria, and the risk outcome of each task is calculated in conjunction with the weighting coefficients. Based on the risk results and the preset risk level classification criteria, each task is classified into low, medium, or high risk levels.

[0013] In the above technical solution, preferably, the specific process of generating the power outage range list includes: Based on the equipment ledger data and the work plan data, the scope of the power outage equipment is calculated; Based on the equipment topology, combined with the job type and job content, the power outage range of the job area is determined; Based on the electrical connections and voltage levels between the equipment, determine a list of outage areas that includes all outage equipment and their connected equipment; The power outage areas in the power outage area list are displayed using visual icons.

[0014] In the above technical solution, preferably, the process of generating the safety measures text specifically includes: Based on the work content set and the power outage range list, and combined with the safety measure templates in the rule base, corresponding safety measure texts are automatically generated. The safety measure texts include pre-control measures for survey equipment, workers and the environment. Match the safety measures text with the corresponding work plan, work scope, and survey equipment; If the generated safety measures do not meet the requirements, provide the surveyors with a prompt to modify or supplement them.

[0015] In the above technical solution, preferably, the specific process of on-site data collection and verification result generation includes: Real-time acquisition of on-site images, videos, environmental parameters, and equipment operating status data via mobile terminals and sensors; The collected data is automatically labeled with metadata including time, location, and device number, and then stored. Through data synchronization, all field-collected data is automatically populated into the corresponding fields of the relevant survey records; The rule engine performs consistency and integrity checks on the on-site collected data, the basic dataset, the work content set, and the safety measures text, including data format verification, field integrity verification, and logical consistency verification, to ensure that there is no inconsistency or missing information among all data and that they meet the actual operational requirements. A verification report is generated based on the verification results. Inconsistencies are marked as errors and fed back to the surveyors for modification or supplementation.

[0016] In the above technical solution, preferably, the specific process of generating the report file includes: Based on the surveying profession, select the appropriate report template, summarize the basic dataset, the candidate set of work sites, the set of work content, the risk assessment results, the list of power outage areas, the safety measures text, the field-collected data, and the verification results, and automatically generate a standardized report file; The report includes a cover, survey overview, detailed work content, risk assessment, safety measures, equipment parameters, and verification results; The report file is format-checked, supports the generation of PDF or Word files, and supports electronic signatures and approval processes for surveyors.

[0017] This invention also proposes an intelligent management system for power grid field surveys based on artificial intelligence, applying the intelligent management method for power grid field surveys based on artificial intelligence disclosed in any of the above technical solutions, including: The data acquisition module is used to acquire equipment ledgers, historical survey reports, work plans, GIS spatial and equipment status data, and write them into the graph database to form a basic dataset; The location candidate module is used to obtain candidate work locations based on the association data and rules between devices and locations in the basic dataset, and generate a candidate set of work locations; The work content module is used to combine the basic dataset and the candidate work location set to generate a work content set corresponding to the target survey order; The risk assessment module is used to calculate and form a scoring detail and risk level based on the basic dataset and the work content set, and generate a risk assessment result accordingly; The power outage list module is used to generate a power outage range list based on the device topology relationship in the graph database and the work objects in the work content set. The safety measure text module, supported by the rule base and terminology base, is used to output safety measure texts corresponding to hazard points and pre-control measures, in combination with the work content set and the power outage range list; The on-site data acquisition module is used to collect on-site images, videos, environmental and equipment operation data, and obtain the survey content submitted by the survey personnel. The on-site data is obtained by combining the collected data with the survey content. The result verification module is used to verify the consistency and integrity of the field-collected data with the basic dataset, the work content set, and the safety measures text, and generate verification results. The report generation module is used to aggregate the basic dataset, the candidate set of work locations, the set of work content, the risk assessment results, the list of power outage areas, the safety measures text, the on-site collected data, and the verification results to generate a report file associated with the target survey form.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) Improved the efficiency of on-site investigation This invention significantly improves the efficiency of on-site surveys through automated data acquisition, work content generation, and risk assessment. Surveyors can quickly obtain recommended work locations and tasks, avoiding the tedious process of manual input and calculation, thus reducing preparation time. The automated management of the entire survey process significantly reduces the time for survey data statistics from several hours to minutes.

[0019] (2) Reduced human error and data inconsistency This invention ensures the integrity and consistency of field survey data through a rule engine and an automatic data verification mechanism, reducing errors from manual data entry. It automatically verifies data format, field integrity, and logical consistency, ensuring the accuracy and compliance of each survey form and significantly reducing the risks associated with subsequent work due to data inconsistencies or errors.

[0020] (3) Improved the accuracy of risk assessment and safety management By employing a multi-dimensional risk assessment model (such as voltage level, work scope, work content, and number of personnel involved) and weighted scoring using the analytic hierarchy process (AHP), this invention can accurately calculate and determine the level of operational risk. This process eliminates the bias inherent in traditional risk assessments that rely on human experience, thus improving the accuracy of risk assessment. Especially in high-risk work environments, this invention provides surveyors with a more scientific risk control plan, ensuring operational safety.

[0021] (4) Dynamically optimized and intelligent generation of work content This invention utilizes association rule mining algorithms and machine learning techniques to dynamically optimize the candidate set of work locations and the prediction model for work content. As the amount of survey tasks and equipment information increases, the prediction model can be continuously updated and optimized, improving the accuracy of work content generation. This not only adapts to existing operational needs but also effectively handles new equipment, new tasks, and special operational scenarios.

[0022] (5) Automatically generate standardized reports, reducing manual workload. This invention can automatically generate standardized survey reports based on real-time collected field data, work content, risk assessment results, and power outage area lists. Once generated, the survey report can be confirmed through electronic signatures and approval processes, reducing time and errors associated with manual report writing and approval. This feature significantly improves report generation efficiency and ensures the consistency and compliance of report content.

[0023] (6) Enhanced adaptability and scalability This invention exhibits excellent adaptability, supporting applications across various power grid devices, operation types, and working environments. With continuous use and updates, new combinations of surveying equipment and work sites can be further optimized through dynamically updated rule bases, continuously improving performance over long-term use. Furthermore, it can be expanded to more fields based on actual needs, demonstrating strong scalability. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating an intelligent management method for power grid field survey based on artificial intelligence, as disclosed in one embodiment of the present invention. Figure 2This is a schematic diagram of the overall architecture of an intelligent management method for power grid field survey based on artificial intelligence, as disclosed in one embodiment of the present invention. Figure 3 This is a schematic diagram of the workflow of an intelligent analysis module disclosed in one embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the process of generating a power outage range list according to an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] The present invention will now be described in further detail with reference to the accompanying drawings: like Figure 1 and Figure 2 As shown, the intelligent management method for power grid field survey based on artificial intelligence provided by the present invention includes: Acquire equipment ledgers, historical survey reports, work plans, GIS spatial and equipment status data, and write them into a graph database to form a basic dataset; Candidate work locations are obtained based on the association data and rules between devices and locations in the basic dataset, and a candidate work location set is generated. By combining the basic dataset and the candidate set of work locations, a set of work content corresponding to the target survey order is generated; Based on the basic dataset and the set of work content, a detailed scoring system and risk level are calculated, and a risk assessment result is generated accordingly. Based on the device topology relationships in the graph database, a list of power outage areas is generated according to the work objects in the work content set; With the support of the rule base and terminology base, the system combines the work content set and the power outage range list to output the safety measures text corresponding to the hazard points and pre-control measures; Collect on-site images, videos, environmental and equipment operation data, and obtain the survey content submitted by the survey personnel. Combine the collected data and the survey content to obtain the on-site collected data. The consistency and integrity of the on-site collected data are verified with the basic dataset, the work content set, and the safety measures text, and verification results are generated. The basic dataset, candidate work locations, work content set, risk assessment results, power outage range list, safety measures text, on-site data collection, and verification results are aggregated to generate a report file associated with the target survey form.

[0027] In this implementation, the efficiency of on-site investigation is greatly improved by automating data collection, work content generation, risk assessment, and report generation. Intelligent analysis based on historical data and association rules can automatically generate work content that matches the investigation task, automatically generate a list of power outage areas and safety measures text, ensuring the accuracy of the power outage area and the comprehensiveness of safety measures. Automatically generating standardized investigation reports can reduce the time spent on manual report filling and ensure the standardization and accuracy of the report content.

[0028] like Figure 3 As shown, specifically, this method mainly supports the automated and intelligent management of power grid field survey tasks through the collection, processing, and intelligent analysis of various data. The specific implementation process is as follows: (1) Data acquisition and graph database storage The data acquisition module automatically retrieves relevant information about power grid equipment from multiple data sources. First, it extracts equipment ledger data from the power grid equipment management system, including equipment number, voltage level, equipment type, model, maintenance records, operating status, and installation location. Then, it retrieves historical survey data from the historical survey management system, including survey location, surveyed equipment name, maintenance type, risk level, and preventative measures. Work plan data is extracted from the work plan system, containing information such as work time, responsible person, and work scope. GIS geospatial data is collected through a GIS platform to obtain the geographical location and topological relationships of the equipment. All this data, after standardization, is stored in the Neo4j graph database. The graph database uses a node and relationship model to store different types of data as different nodes and constructs connections between equipment based on relationships.

[0029] (2) Generate a candidate set of work locations Based on the "exploration equipment-work location" association data in historical survey reports, an association rule mining algorithm is used to train the historical data and extract the association rules between equipment and work locations. When surveyors input survey equipment information, the algorithm automatically matches the rule base and generates a candidate set of work locations. The work locations in the candidate set are sorted according to the confidence of the match, and the work location with the highest match confidence is output as the suggested location.

[0030] (3) Generate a set of work content and risk assessment By combining candidate work locations and equipment inventory data, a set of work content corresponding to the target survey order is automatically generated. Using a multi-dimensional risk assessment model, the risk level of each task is calculated based on factors such as task content, equipment type, and voltage level. The risk assessment results will display the risk level of the task and provide a reference for subsequent decision-making.

[0031] (4) Generate a list of power outage areas Based on the device topology in the graph database, a power outage range list is automatically generated according to the work objects (such as equipment and lines) in the work content set. The power outage range list includes all equipment that needs to be powered down and its associated equipment. The power outage range is automatically calculated based on the electrical connection relationships of the equipment and the work type to ensure the accuracy of the work scope.

[0032] (5) Generate safety measures text With the support of a rule base and terminology base, the system automatically generates corresponding safety measure texts by combining the work content set and the power outage range list. The safety measure texts include pre-control measures for different hazards (such as working at height, electric shock, equipment failure, etc.) to ensure the safety of the survey mission.

[0033] (6) On-site data collection and verification The on-site data acquisition module collects images, videos, environmental data, and equipment operation data. The collected data is automatically recorded in real-time via mobile terminals, including metadata such as time, location, and equipment number. The collected data is then verified for consistency and completeness against the base dataset, work content set, and safety measure texts to ensure that the on-site data aligns with the work plan and equipment status. The verification results generate a report and are provided to the survey personnel for correction or supplementation.

[0034] (7) Report document generation The system aggregates data from the base dataset, candidate work sites, work content set, risk assessment results, power outage area list, safety measures documents, on-site data collection, and verification results, and automatically generates a standardized survey report. The survey report can be generated in PDF or Word format and supports electronic signatures and approval processes to ensure compliance and accuracy.

[0035] In the above embodiments, preferably, the specific process of acquiring equipment ledgers, historical survey reports, work plans, GIS spatial and equipment status data, and writing them into the graph database includes: Extract equipment ledger data from the power grid equipment management system. The equipment ledger data includes equipment manufacturer (e.g., "XX Group"), equipment number (e.g., "T2201" "L2201"), usage time (e.g., "2018-05-01"), voltage level (e.g., "220kV" "110kV" "10kV"), equipment type (main transformer, line, switch), equipment model, equipment maintenance record, operating status, assigned bay (e.g., "#1 bay" "#2 bay") information, and equipment installation location. Historical survey order data is extracted from the historical survey management system. The historical survey order data includes survey location (such as "main transformer room of XX substation" or "near tower #10 of XX line"), survey equipment name (such as "main transformer #1" or "tower #15 of L2201"), maintenance type (planned maintenance, fault maintenance, preventive maintenance), risk level (high, medium, low), pre-control measures (such as "wearing safety belts" or "laying tarpaulins"), and process status (prepared, executed, completed, terminated). Extract work plan data from the work plan system. The work plan data includes the planned time (e.g., "2024-08-10"), the person in charge (e.g., "Zhang XX"), and the work scope (e.g., "#1 main transformer" and "L2201#10-#20 tower section"), and supports collection by plan status (pending execution, executed, canceled). Obtain GIS geospatial data from the GIS platform. The GIS geospatial data includes the geographical location information of the equipment and the topological relationship of the equipment. It covers the location of power grid equipment (latitude and longitude coordinates), roads (highway, ordinary road), railways (high-speed railway, ordinary railway), and power transmission channels (important channels, general channels) layer data, and supports collection by region (province, city, county). In addition, data can be obtained from the equipment status monitoring system, including equipment operating status (operation, maintenance, standby) and key parameters (such as main transformer oil temperature and line current), supporting real-time and historical data acquisition. The equipment ledger data, historical survey data, work plan data and GIS geospatial data are standardized and written into a graph database, namely Neo4j graph database. When storing data, a model of nodes and relationships is established to store equipment ledger data, historical survey data, and work plan data as different node types. Extract the relationships between different nodes and build connections between nodes based on these relationships.

[0036] During implementation, relevant data was extracted from the power grid equipment management system, historical survey management system, work plan system, and GIS platform. All extracted data underwent standardization processing and was then written into the Neo4j graph database. In the graph database, equipment ledger data, historical survey data, and work plan data were stored as different node types. Relationships between nodes (such as "dependency" and "containment") were established based on the physical and logical relationships between equipment, ensuring efficient data storage and facilitating correlated queries.

[0037] In addition, data validation can be performed on the above data. The rule engine builds three types of validation rule bases: Data format verification rules: The date format must be "YYYY-MM-DD", the equipment number must conform to the format "XX-XX-XXX" (such as "T-220-001"), and the voltage level must be one of "10kV", "35kV", "110kV", "220kV" or "500kV"; Field integrity verification rules: Survey forms must include fields for survey location, survey equipment, maintenance type, responsible person, and process status; equipment ledgers must include fields for equipment number, voltage level, and interval to which it belongs; Logical consistency verification rule: Survey orders in the "completed" status must be associated with work plans.

[0038] Validation result processing: For data that fails validation, generate prompts containing error types (format errors, missing fields, logical contradictions) and error fields (such as "planned time format error" or "survey form lacks responsible person field").

[0039] In the above embodiments, preferably, the specific process of generating a candidate set of work locations includes: Based on the "exploration equipment-work location" association data in historical survey order data, an association rule mining algorithm is used to train the data, extract the association rules between the exploration equipment and the work location, and generate a rule base that can be used for work location prediction. When the survey equipment information is input, a candidate set of work locations is automatically generated based on the trained rule base, and sorted according to the matching confidence. The work location with the highest confidence is output as the suggested survey location. The steps of association rule mining algorithms include: Preprocess historical survey data to extract effective correlation information between survey equipment and work sites; The frequent itemset mining method is used to extract the combination of exploration equipment and work location from historical exploration data, calculate the corresponding confidence level, and generate association rules. Based on association rules, the system automatically matches device information and outputs a candidate set of work locations that meet the criteria. If the input exploration equipment information cannot be matched in the existing rules, a prompt message will be sent to the exploration personnel to manually input the work location. New combinations of exploration equipment and work locations will be automatically added to the rule base to update and optimize the generation of subsequent candidate sets.

[0040] During implementation, an association rule mining algorithm was trained based on the "exploration equipment-work location" association data in historical survey reports. The algorithm uses frequent itemset mining to extract effective association information between survey equipment and work locations, and generates association rules based on confidence levels.

[0041] After surveyors input their survey equipment information, the system can automatically generate a candidate set of work locations based on a trained rule base. The work locations in the candidate set are sorted according to their matching confidence level, with the work location having the highest confidence level being prioritized as the recommended location.

[0042] Specifically, during the data training process pre-generated at the work site, data encoding includes: uniquely encoding the survey equipment (such as "#1 main transformer", "L2201#15 tower", "F101#8 pole switch") and the work site (such as "XX substation main transformer room", "near XX line #15 tower (next to the highway)", "XX community F101#8 pole"); and converting text data into numerical data (such as "#1 main transformer" being encoded as 1001, and "XX substation main transformer room" being encoded as 2001).

[0043] The frequent itemset mining process includes: setting a minimum support threshold of 0.2 (support = number of survey orders containing the equipment-location combination / total number of survey orders), mining frequent itemsets, for example, the combination "#1 main transformer - XX substation main transformer room" appears 300 times in 1000 survey orders, support = 300 / 1000 = 0.3 ≥ 0.2, and is listed as a frequent itemset.

[0044] The association rule generation process includes: based on frequent itemsets, setting a minimum confidence threshold of 0.7 (confidence = number of survey orders containing the equipment-location combination / number of survey orders containing the equipment), generating an association rule for "survey equipment → work location". For example, there are 350 survey orders containing "#1 main transformer", of which 300 contain "#1 main transformer - XX substation main transformer room". The confidence is 300 / 350≈0.85≥0.7, generating the rule "#1 main transformer → XX substation main transformer room".

[0045] The rule application process includes: when the survey equipment information (such as "#1 main transformer") is input, the system queries the associated rule library, matches all rules containing the equipment, and outputs the work location corresponding to the rule with the highest confidence (such as "XX substation main transformer room") as the pre-generated result, with an accuracy rate of over 80%; if there is no matching rule for the equipment (such as adding a new equipment), the survey personnel are prompted to manually input it, and the equipment-location combination is included in the subsequent training data to update the rule library.

[0046] In the above embodiments, preferably, the process of generating the work content set specifically includes: Based on the survey equipment, work location, and maintenance type, a work content prediction model is constructed. The work content prediction model is trained using the "survey equipment-work location-maintenance type-work content" data from historical survey orders to extract the prediction rules for work content. The training of the job content prediction model uses the gradient boosting tree algorithm. The feature variables during the training process include: the type of survey equipment, the specific information of the work site, the maintenance type, the operating status of the equipment, and the job type. After inputting the survey equipment, work location, and maintenance type information, the work content prediction model generates corresponding work content information based on the trained rules. The work content information includes maintenance items, work steps, and work requirements. The work content information is screened and optimized, and the final work content is determined based on the matching degree and operability of the work steps and requirements. If the generated work content fails to pass the optimization screening, a prompt message will be sent to the survey personnel to manually fill in the work content. The new combination of "survey equipment-work location-maintenance type-work content" will be automatically included in the subsequent rule base to continuously update and optimize the work content prediction model.

[0047] During implementation, an intelligent agent-based question answering system can be used to build a job content prediction model. The specific process includes: Feature variable selection: Select “survey equipment”, “work location” and “maintenance type” as feature variables, such as “survey equipment: L2201#15 tower”, “work location: near #15 tower of XX line”, “maintenance type: fault repair”; “line insulator replacement”, “main transformer oil sample testing”, and “switch characteristic test” as target variables. Classify and encode the target variables (e.g., “line insulator replacement” is coded as 101, “main transformer oil sample testing” is coded as 201, and “switch characteristic test” is coded as 301).

[0048] Training data preparation: Historical survey reports are used as training data, covering survey scenarios for various equipment types in substations (main transformers, switches, busbars), transmission lines (lines, towers, insulators), and distribution lines (feeders, switches, transformers); the training data is cleaned to remove samples with missing key fields (such as missing survey equipment or work content) to ensure data integrity.

[0049] Agent training: The training data is divided into 5 parts, with 4 parts as the training set and 1 part as the validation set. The process is repeated 5 times to ensure that each sample participates in the validation. The model accuracy is calculated using the validation set. Training is stopped when the accuracy is ≥85%. Finally, the accuracy of the model on the test set is stabilized at 85%-90%.

[0050] Model Application: When inputting feature variables (such as "Survey equipment: L2201#15 tower, work location: near XX line #15 tower, maintenance type: fault maintenance"), the system converts the feature variables into numerical codes (such as "L2201#15 tower" is coded as 4001, "near XX line #15 tower" is coded as 5001, and "fault maintenance" is coded as 2); the model outputs the target variable code (such as 101), and the system maps the code to the corresponding work content "line insulator replacement" and displays it to the survey personnel; the survey personnel can modify the pre-generated content according to the actual scenario, and the modified content will be included as new samples in the model training data for subsequent model iteration and optimization.

[0051] In the above embodiments, preferably, the specific process of generating risk assessment results includes: A multi-dimensional risk assessment model is constructed based on preset assessment dimensions, and the weight coefficient of each dimension is calculated using the analytic hierarchy process. According to the evaluation dimensions, scores are given based on the preset scoring criteria, and the risk outcome of each task is calculated by combining the weighting coefficients. Based on the risk results and the preset risk level classification standards, each operation is classified into low, medium or high risk levels.

[0052] During implementation, the risk level calculation process includes: Multi-dimensional assessment model construction: Four core assessment dimensions are determined: “voltage level, scope of work, content of work, and number of personnel involved”. The selection of each dimension is based on the safety management requirements of power grid site survey, covering equipment characteristics (voltage level), scale of work (scope of work, number of personnel involved) and complexity of work (content of work).

[0053] Determining Dimension Weights: The specific steps for determining the weights of each dimension are as follows: The scoring criteria for each dimension were developed with reference to the "Power Safety Work Regulations of State Grid Corporation of China" and the on-site inspection risk assessment specifications.

[0054] Voltage level dimensions: 220kV and above (8-10 points, of which 500kV is 10 points and 220kV is 8 points), 110kV (5-7 points), 35kV and below (1-4 points, of which 35kV is 4 points and 10kV is 2 points).

[0055] Scope of work: across multiple bays / lines (8-10 points, across 3 or more bays / lines is 10 points, across 2 is 8 points), single bay / single line (5-7 points), local equipment (1-4 points, single equipment is 4 points, component level is 1 point).

[0056] Job content dimensions: complex maintenance (8-10 points, main transformer / busbar maintenance is 10 points, cable laying is 8 points), routine maintenance (5-7 points, switch / line maintenance is 7 points, insulator replacement is 5 points), daily inspection (1-4 points, comprehensive inspection is 4 points, special inspection is 2 points).

[0057] Number of participants: 10 or more (8-10 points, 20 or more (10 points), 10-19 (8 points), 5-9 (5-7 points), 4 or less (1-4 points, 4 (4 points), 1 (1 point).

[0058] Risk level calculation and output: Overall risk level = voltage level score × 0.35 + work scope score × 0.25 + work content score × 0.25 + number of personnel involved score × 0.15; Based on the calculation results, risk levels are divided as follows: a total score ≥ 8 points is high risk, 5-7 points is medium risk, and ≤ 4 points is low risk.

[0059] The system automatically generates a risk level report, which includes scores, weights, calculation processes, and risk level conclusions for each dimension. For example, "Voltage level 220kV (8 points) × 0.35 = 2.8, work scope single line (7 points) × 0.25 = 1.75, work content routine maintenance (5 points) × 0.25 = 1.25, number of personnel involved 3 (5 points) × 0.15 = 0.75, total score 6.5 points, judged as medium risk." The person in charge of the survey can manually adjust the risk level according to the actual situation on site (such as severe weather or special equipment conditions), but must fill in the reason for the adjustment (such as "Due to heavy rain, the risk level of the operation has been upgraded from medium risk to high risk"). The adjustment record is automatically archived for future reference.

[0060] Based on the data obtained from the above implementation methods, a pre-control measures rule base unit and a safety measures terminology base unit are constructed.

[0061] The pre-control measures rule base unit constructs and optimizes the "work content-maintenance type-hazard point-pre-control measures" association rule base through association rule mining and incremental learning, ensuring that pre-control measures are accurately matched with the survey scenario.

[0062] First, an initial rule base is constructed by selecting samples from historical survey reports that contain complete fields for "work content, maintenance type, hazard points, and pre-control measures," covering various professional scenarios such as substation (e.g., "main transformer maintenance" and "switch commissioning"), transmission (e.g., "line maintenance" and "tower reinforcement"), and distribution (e.g., "feeder maintenance" and "transformer replacement"). The sample data is then standardized. For example, "high-altitude work" and "fall from height risk" are uniformly classified as the hazard point "fall from height," and "wearing a safety belt" and "wearing a safety belt" are uniformly classified as the pre-control measure "wearing a safety belt."

[0063] The rule base adopts an incremental update approach, collecting new survey orders quarterly. The new data is standardized to ensure field formats are consistent with the initial rule base. For new data itemsets, if a frequent itemset appears in the new data with a frequency greater than or equal to the minimum support (100 times / quarter) and does not exist in the initial rule base, it is identified as a new frequent itemset. After each quarterly update, 100 new survey orders are extracted and matched against the updated rule base for pre-control measures. If the matching accuracy is greater than or equal to 85% (initial accuracy 85%), the update takes effect. If the accuracy is less than 85%, the incremental data processing is reviewed, mining parameters are corrected (e.g., adjusting the minimum confidence to 0.65), and the update is repeated to verify the update results.

[0064] Secondly, rule-based matching is applied, and the matching process includes: After the intelligent analysis module generates the work content and maintenance type (e.g., "Work content: line insulator replacement, maintenance type: fault repair"), the system queries the pre-control measures rule base, matches the association rules according to the "work content + maintenance type" field, and filters out the top 3 rules with the highest confidence (e.g., "line insulator replacement + fault repair → fall from height → wearing a safety belt" (confidence 0.9), "line insulator replacement + fault repair → falling object → wearing a safety helmet" (confidence 0.85), "line insulator replacement + fault repair → electric shock → voltage detection and grounding" (confidence 0.8)). Results Display: The matched hazards and pre-control measures will be displayed to the surveyors in a list format. The list includes the hazard name, the content of the pre-control measure, and the rule confidence level, so that the surveyors can prioritize the pre-control measures with high confidence. The surveyors can manually add unmatched hazards and pre-control measures (such as "Add Hazard: Mechanical Injury, Pre-control Measures: Standard Crane Operation"). The added content will be included as incremental data in the rule base update for the next quarter.

[0065] In the above embodiments, preferably, the process of generating the safety text specifically includes: Based on the work content set and the list of power outage areas, and combined with the safety measure templates in the rule base, the corresponding safety measure text is automatically generated. The safety measure text includes pre-control measures for survey equipment, workers and the environment. Match the safety measures text with the corresponding work plan, work scope, and survey equipment; If the generated safety measures do not meet the requirements, provide the surveyors with a prompt to modify or supplement them.

[0066] Specifically, the safety measure terminology library unit constructs and supplements the "power outage range - circuit breaker to be pulled - disconnecting switch" safety measure terminology library, realizing the automated generation of safety measure content and reducing the workload of manual filling.

[0067] The initial terminology database construction process includes: Corpus Collection and Preprocessing: Safety measure texts were collected from historical professional survey reports (e.g., "Power outage on 110kV XX line, circuit breaker #1 and disconnector #2 should be disconnected" and "Power outage on main transformer #1 of 220kV XX substation, circuit breaker #1-1 and disconnector #1-2 should be disconnected"). The corpus was then preprocessed. Word segmentation: Use word segmentation tools to segment the Ancuo text. For example, segment the phrase "110kVXX line is de-energized, #1 circuit breaker and #2 disconnect switch should be pulled" into "110kVXX line", "de-energized", "should be pulled", "#1 circuit breaker", and "#2 disconnect switch". Entity recognition: A model was used for entity recognition, identifying three types of entities: "power outage area", "circuit breaker", and "disconnect switch". The model was trained using a pre-annotated corpus of safety monitoring texts (10,000 entries), and the entity recognition accuracy after training was ≥90%. Relationship Extraction: Construct an entity relationship extraction model and use a rule-based method to extract the association between "power outage range" and "circuit breaker" and "disconnect switch". For example, extract the entities "circuit breaker" and "disconnect switch" that are immediately followed by "should be pulled" after "power outage range" from the word segmentation results, and establish the association relationship "110kVXX line → #1 circuit breaker, #2 disconnect switch" and "220kVXX substation #1 main transformer → #1-1 circuit breaker, #1-2 disconnect switch". Terminology database storage: The extracted relationships are stored in the initial safety measure terminology database. The terminology database is stored using a graph database (Neo4j). "Power outage range" is the node, "circuit breaker" and "disconnect switch" are the associated nodes, and "should be pulled" is the relationship edge. The node attributes include entity name, voltage level, and affiliated unit, and the relationship edge attributes include association strength.

[0068] Ancuo content generation applications include: Generation process: After the topology calculation module determines the power outage area, the system queries the safety measure terminology library, matches similar "power outage area" entities, and extracts the associated "circuit breaker" and "disconnect switch" entities; Result Optimization: Based on the equipment list of the power outage area output by the topology calculation module (such as "110kV XX line #10-#20 section includes #1 circuit breaker and #3 disconnector"), the terminology database matching results are optimized. If the "disconnector" (#2 disconnector) associated with the terminology database is not in the equipment list, it is automatically removed, and "#1 circuit breaker and #3 disconnector" are retained. Finally, the safety measure content "110kV XX line #10-#20 section is under power outage, #1 circuit breaker and #3 disconnector should be disconnected" is generated and displayed to the survey personnel. The survey personnel can manually adjust the safety measure content.

[0069] like Figure 4As shown, in the above embodiment, preferably, the specific process of generating the power outage range list includes: Based on equipment ledger data and work plan data, the scope of equipment to be affected by power outages was calculated. By analyzing the equipment topology and combining it with the type and content of the work, the power outage range of the work area can be determined. Based on the electrical connections and voltage levels between the equipment, determine a list of outage areas that includes all outage equipment and their connected equipment; The power outage areas in the power outage area list are displayed using visual icons.

[0070] In this implementation, the topological association information in the equipment ledger data is first cleaned and standardized to lay the foundation for subsequent topological map construction. Equipment ledger data is extracted from the data warehouse, fields related to topological associations are filtered out, redundant data is removed, logical errors are corrected, and data formats are standardized. An equipment type coding table is constructed to uniformly encode equipment types; a voltage level coding table is also constructed to standardize voltage level representations. The cleaned and standardized data is stored in a preprocessed data table in the graph database (Neo4j) to achieve data standardization. The table structure includes fields such as Equipment ID (unique identifier), Equipment Number (standardized code), Equipment Type (code + name), Belonging Site, Voltage Level (code + value), Parent Equipment ID, Child Equipment ID, Operating Status, and Processing Time.

[0071] Based on the preprocessed device data, a visual device topology map is constructed to clearly show the relationships between devices. Nodes in the map represent devices, and edges represent the physical connections between devices.

[0072] Node attributes include device ID, device number, device type, voltage level, operating status, and bay information; Edge attributes include connection type (direct connection, indirect connection) and connection distance (unit: meters, line devices only).

[0073] The topology graph is constructed based on a graph database. It is created by creating nodes, edges, and hierarchical divisions. The topology graph visualization interface supports zooming and panning, node highlighting, and status labeling, which can intuitively view devices in different areas, local relationships, and distinguish device status.

[0074] Based on this, the system automatically calculates and highlights the power outage area according to the input survey equipment information and preset professional rules.

[0075] During implementation, the "State Grid Corporation of China Power Safety Work Regulations" and the power outage operation specifications of various specialties can be referenced to build a rule library for the power outage scope of substations, transmission lines, and distribution lines. Each rule includes the applicable scenarios, judgment conditions, and power outage scope definition standards.

[0076] The specific process for calculating the power outage area includes: Input information acquisition: Receive survey equipment information (equipment ID, equipment type, and related discipline) and survey task tags (such as "#1 main transformer three-sided switch open") transmitted by the intelligent analysis module; Rule matching: Based on the "major + survey task tag", the preset rule library is queried to match the corresponding power outage range rule; for example, if the survey equipment is the substation #1 main transformer and the tag is "three-sided switch open", the "main transformer maintenance rule" will be matched. Topology Traversal: Starting from the surveyed equipment nodes, traverse the topology map according to the matching rule definition criteria: Main Transformer Maintenance Scenario: Traverse from the three-sided disconnector nodes of the #1 main transformer (#1-2 disconnector, #1-3 disconnector, #1-4 disconnector) towards the main transformer side to obtain all directly or indirectly related equipment (#1 main transformer, #1 bay transformer, #1 main transformer side surge arrester); Line Maintenance Scenario: Traverse from the #10 tower node of the transmission line to the #20 tower node to obtain all lines, towers, and insulator nodes within the #10-#20 tower section; Range Verification: Verify the equipment list obtained through traversal with the bay / line information of the equipment in the preprocessed data table, and remove equipment that does not belong to the current survey task scope (such as the #2 bay transformer mixed in during the #1 main transformer maintenance); Result Output: Generate the final power outage range equipment list, including equipment ID, equipment number, equipment type, and bay information.

[0077] The list of equipment within the power outage area is fed back to the topology visualization interface. Equipment nodes and connecting edges within the power outage area can be highlighted in red, creating a clear contrast with non-power-outage equipment (green for operation, yellow for maintenance).

[0078] In the above embodiments, preferably, the specific process of collecting data on-site and generating verification results includes: Real-time acquisition of on-site images, videos, environmental parameters, and equipment operating status data via mobile terminals and sensors; The collected data is automatically labeled with metadata including time, location, and device number, and then stored. Through data synchronization, all field-collected data is automatically populated into the corresponding fields of the relevant survey records; The rule engine performs consistency and integrity checks on on-site collected data, basic datasets, work content sets and safety measure texts, including data format verification, field integrity verification and logical consistency verification, to ensure that there is no inconsistency or missing information among all data and that they meet the actual operational requirements. A verification report is generated based on the verification results. Inconsistencies are marked as errors and fed back to the surveyors for modification or supplementation.

[0079] During the implementation process, surveyors used mobile terminals to conveniently collect multi-source on-site information. Among them, image acquisition was carried out by using the mobile terminal's camera to capture key images such as the appearance of the survey equipment (e.g., tower body, insulators), the surrounding environment (geographical features, surrounding buildings), and the work site (work area, safety measures). Each image was automatically watermarked with information such as the shooting time, location (latitude and longitude), and associated equipment number, which facilitates subsequent query and analysis.

[0080] Video capture is used for complex equipment maintenance (such as maintenance of the main transformer cover in a substation) or hazardous operation scenarios (such as working near live conductors). The video is recorded on-site, stored locally on the mobile terminal, and simultaneously uploaded to the cloud server.

[0081] Environmental data acquisition is achieved through an external environmental monitoring module (which supports the detection of temperature, humidity, wind speed, air pressure, and harmful gas concentration) to collect real-time environmental data at the survey site. A set of data is automatically collected every 5 minutes, stored on the mobile terminal, and synchronized to the corresponding field on the survey form.

[0082] For power grid equipment that supports intelligent sensing (such as smart meters and smart switches), mobile terminals can connect to the equipment via Bluetooth or NFC technology to read the equipment's real-time operating data (voltage, current, power, and equipment health status), and automatically fill the equipment data fields in the survey form to achieve equipment data collection and ensure data accuracy and timeliness.

[0083] The mobile terminal loads the electronic form corresponding to the survey form template. Survey personnel fill in the survey content online according to the actual situation on site, including equipment information, hazard points and pre-control measures, safety measures, work content and progress and other information. After completion, the survey personnel automatically upload the data to the system server, store it in the corresponding record of the survey form, and update it synchronously to the progress tracking page of the process control module.

[0084] After the survey content is submitted, the system activates an automatic verification mechanism. Based on a preset rule base, it performs logical consistency, data integrity, and format standardization checks on key fields. Once the automatic verification is passed, the system synchronizes the survey content to the mobile device of the person in charge of the survey, who then conducts a second review, including content review, feedback, and signature confirmation. After the review is passed, both the person in charge of the survey and the survey personnel need to confirm the electronic signature on their mobile devices. The signature information is automatically appended to the end of the survey form, forming a "dual confirmation" mechanism to ensure that the responsibility for the survey content is traceable.

[0085] Furthermore, when the on-site working environment changes or the initial survey has omissions, a re-survey process needs to be initiated. This allows for the retrieval of historical records and updates to the re-survey content, with modifications and annotations made to any discrepancies. If special circumstances arise (such as work plan cancellation or on-site conditions not meeting safety requirements) that prevent the survey from continuing, the survey process must be terminated by initiating a termination request, which will be approved before termination is implemented.

[0086] Once the entire exploration process (including re-exploration) is completed and approved, the exploration completion stage begins. This stage involves archiving exploration records, generating reports, and applying the results.

[0087] In the above embodiments, preferably, the specific process of generating the report file includes: Based on the surveying profession, select the appropriate report template, summarize the basic dataset, candidate set of work sites, set of work content, risk assessment results, list of power outage areas, safety measures text, on-site data collection and verification results, and automatically generate a standardized report file; The report includes a cover page, an overview of the survey, detailed work content, risk assessment, safety measures, equipment parameters, and verification results; The system performs format checks on report documents, supports the generation of PDF or Word files, and allows surveyors to perform electronic signatures and approval processes.

[0088] The system automatically matches the corresponding report template based on the survey specialty (substation / transmission / distribution). The template contains fixed chapters, including a cover page, survey overview, detailed survey content, survey conclusions and recommendations, and attachments. Based on the report template, the system automatically fills in data, text, and inserts charts, and finally exports the report for signing and stamping to complete the survey report.

[0089] In addition, a refined access control system is built based on "organizational hierarchy + role" to ensure the security of access to and operation of survey data and prevent the leakage of sensitive information and unauthorized operations.

[0090] Specifically, based on the organizational structure of power grid companies, users can be divided into five levels, each corresponding to a different data viewing scope. When a user logs into the system, the system automatically matches the corresponding level of permissions according to their organization, without requiring manual configuration. If a user's organization changes (e.g., from County Bureau A to County Bureau B), the system automatically updates their organizational level permissions through data synchronization, avoiding permission lag issues caused by personnel transfers. If a user holds multiple roles (e.g., a user is both a work area administrator and a survey manager), the system follows the principle of "accessing the intersection of permissions and accessing the smallest possible data range."

[0091] Based on the aforementioned AI-based intelligent management method for power grid field surveys, the specific process for surveyors during on-site surveys includes: After receiving the work plan number and exploration equipment information input by the surveyor, the system automatically retrieves a set of candidate work locations related to the exploration equipment from the data repository, sorts the work locations according to association rules and confidence levels, and recommends the highest-ranked work location to the surveyor. After the surveyors select a work location, a corresponding set of work content is automatically generated based on the survey equipment, work location, and maintenance type. The generated work content is then optimized to ensure that it matches the on-site operation requirements. By connecting to the field acquisition equipment, the system receives images, videos, environmental data, and equipment operation status data uploaded by surveyors in real time, and compares them with the basic dataset, the work content set, and the safety measures text for completeness and consistency. Verify the data matching between the on-site collected data and the basic dataset, the work content set, and the safety measures text. If any inconsistency or omission is detected, generate an error message and provide feedback to the survey personnel. Based on the modified data entered by the surveyors and the data collected on site, the survey records are automatically updated, a survey report is generated, and the report is automatically submitted to the reviewers for approval.

[0092] This invention also proposes an intelligent management system for power grid field surveys based on artificial intelligence, applying the intelligent management method for power grid field surveys based on artificial intelligence disclosed in any of the above embodiments, including: The data acquisition module is used to acquire equipment ledgers, historical survey reports, work plans, GIS spatial and equipment status data, and write them into the graph database to form a basic dataset; The location candidate module is used to obtain candidate work locations based on the association data and rules between devices and locations in the basic dataset, and generate a candidate set of work locations; The work content module is used to combine the basic dataset and the candidate set of work locations to generate the work content set corresponding to the target survey order; The risk assessment module is used to calculate and generate a score breakdown and risk level based on the basic dataset and the set of work content, and to generate risk assessment results accordingly. The power outage list module is used to generate a list of power outage areas based on the equipment topology relationships in the graph database and the work objects in the work content set. The safety measures text module, supported by a rule base and a terminology base, is used to output safety measures texts corresponding to hazard points and pre-control measures, in conjunction with a work content set and a power outage range list. The on-site data acquisition module is used to collect on-site images, videos, environmental and equipment operation data, and obtain the survey content submitted by the survey personnel. The on-site data is obtained by combining the collected data and the survey content. The result verification module is used to verify the consistency and integrity of the field-collected data with the basic dataset, work content set and safety measure text, and generate verification results. The report generation module is used to aggregate the basic dataset, candidate set of work locations, set of work content, risk assessment results, list of power outage areas, safety measures text, on-site data collection and verification results to generate a report file associated with the target survey form.

[0093] The functions to be implemented by each module of the intelligent power grid field survey management system based on artificial intelligence disclosed in the above embodiments correspond to the steps of the intelligent power grid field survey management method based on artificial intelligence disclosed in the above embodiments. In the implementation process, the operation is carried out with reference to the above embodiments, and will not be repeated here.

[0094] The implementation process and effects of the intelligent management method and system for power grid field survey based on artificial intelligence disclosed in the above embodiments are illustrated through the following examples.

[0095] (1) Implementation scenario definition Surveying and mapping specialty: Substation specialty Object of investigation: 110kV XX substation #2 main transformer (equipment number T-110-002, commissioning time June 2019, voltage level 110kV, belonging to bay #2 main transformer bay) Survey Objective: To provide on-site data support for the planned overhaul of #2 main transformer (replacement of main body seals, oil quality testing). This requires completing the following: pre-generation of the work location, confirmation of work content, assessment of risk level, analysis of the power outage area (including marking of remaining energized parts), and matching of pre-control measures. Related data sources: Equipment ledger system (parameters of #2 main transformer), historical survey record system (126 maintenance and survey records of 110kV main transformers in the past 3 years), GIS platform (substation topology and geospatial data), and work plan system (plan number JH202410005, planned date October 15, 2024). (2) System deployment preparation: Data integration (based on the data acquisition module) Multi-source data access: Equipment ledger data: Synchronize all parameters of the #2 main transformer (manufacturer: XX Electric, rated capacity 50MVA, cooling method: forced oil circulation, belonging to the #2 main transformer bay) from the power grid equipment management system, and complete the format verification (equipment number conforms to the format "T-voltage level-serial number") and field integrity verification (no missing "voltage level" and "belonging bay" fields) through the data verification unit. Historical survey data: 126 110kV main transformer maintenance surveys were extracted from the historical survey management system. The correlation data of "survey equipment-work location" and "work content-prevention measures" were filtered for use in agent training and association rule construction. GIS data: Obtain the substation topology layer (including the physical connection relationship between #2 main transformer and #2-1 circuit breaker, #2-2 disconnector, and #2 busbar) and geospatial data (substation perimeter and the latitude and longitude of the #2 main transformer room) from the GIS platform. Data storage: Verified data is stored in a data warehouse. The equipment topology is stored using a Neo4j graph database (nodes: equipment ID / type / voltage level, edges: connection type / distance), and historical survey data is stored in MySQL partitioned tables (sharded by year).

[0096] (3) Initialization of intelligent agent and topology analysis model (based on intelligent analysis module and topology calculation module) Agent training: Training data: Samples of "Frequently Asked Questions" (e.g., "What fields need to be filled in for 110kV main transformer maintenance survey?" "How to determine the risk level of main transformer maintenance?") were selected from 126 historical survey reports. Preventive control measures rule base ("Main transformer maintenance → fall from height → wearing a safety belt") and substation maintenance clauses in the "State Grid Electric Power Safety Work Regulations" were also selected. Model selection: Gradient boosting tree was used to construct the question answering model. The "question text" was encoded as a feature vector, and the "answer text / rule ID" was used as the target variable. Five-fold cross-validation was used (4 training sets and 1 validation set) until the accuracy on the validation set was ≥85%. Question and answer function configuration: Supports three types of questions and answers: "process consultation", "data query" and "anomaly handling". For example, if you input "What pre-control measures should be matched for the maintenance of the #2 main transformer?", the intelligent agent will automatically associate the rule "work content = main transformer maintenance + maintenance type = planned maintenance" and output the top 3 pre-control measures with the highest confidence (confidence 0.92: fall from height → wear safety belt; 0.88: oil leak → lay oil-proof cloth; 0.85: electric shock → test for voltage and ground).

[0097] (4) Configuration of topology diagram single-line diagram generation model: Single-line diagram rule base construction: Refer to the "Substation Main Transformer Maintenance Rules" in the preset professional rule base and supplement the single-line diagram drawing specifications (equipment symbols: main transformer uses "□", circuit breaker uses "○", disconnector uses "—⊓—"; layout logic: arranged in order from high to low voltage level (110kV busbar → main transformer → 35kV busbar) and from left to right). Topology traversal algorithm configuration: The depth-first search (DFS) algorithm is adopted. Starting from the #2 main transformer node (device ID1002), the parent / child device nodes (#2-1 circuit breaker ID2002, #2-2 disconnector ID3002, #2 bus ID4002) are traversed to automatically generate a device connection relationship linked list for single-line diagram drawing.

[0098] (5) Exploration initiation phase: Intelligent question-and-answer assisted initialization User operation: Liu XX, the person in charge of the overhaul and survey of the #2 main transformer (with work area level authority and patent 2.2.6 role), logs into the system and enters the work plan number JH202410005. The system automatically fills in the plan information (planned time 2024-10-15, work scope #2 main transformer). Intelligent agent interaction: Liu XX asked through the system's "intelligent assistant" pop-up window: "What equipment data needs to be associated with the planned maintenance survey of #2 main transformer?" The intelligent agent retrieved the equipment ledger rule library and output the answer: "The ledger data (including operating status and voltage level) of #2 main transformer body, #2-1 circuit breaker, #2-2 disconnector, and #2 busbar need to be associated. You can click the [Automatic Association] button to synchronize the data." Survey order generation: Liu XX clicked on [Automatic Association], and the system retrieved the above equipment data from the data warehouse, matched it with the substation professional survey order template, generated survey order number KC-XX-110T-202410-002, and updated the status to "Pending Execution".

[0099] (6) Intelligent analysis stage: Pre-generation of work content and risk assessment Work location pre-generation: The system calls the work location pre-generation unit, inputs "survey equipment = #2 main transformer", and automatically matches the historical "survey equipment - work location" association rules (support 0.32, confidence 0.87: "#2 main transformer → XX substation #2 main transformer room") to pre-generate the work location; Liu XX had doubts about the pre-generated results and asked the agent: "Does the #2 main transformer room include the auxiliary equipment area?" The agent retrieved GIS data and replied: "The #2 main transformer room includes the main transformer body area (8m long × 6m wide) and the auxiliary oil tank area (3m long × 2m wide). The survey needs to cover the entire area." After Liu XX confirmed, he locked the work location.

[0100] Risk level calculation: The risk level calculation unit automatically extracts assessment dimension data: voltage level 110kV (5-7 points, take 6 points), work scope single bay (5-7 points, take 6 points), work content routine maintenance (5-7 points, take 6 points), number of personnel involved 6 (5-7 points, take 6 points). Calculated using the weights of the analytic hierarchy process (voltage level 0.35, work scope 0.25, work content 0.25, number of personnel involved 0.15): 6×0.35+6×0.25+6×0.25+6×0.15=6 points, which is judged as medium risk; Liu XX asked the AI ​​agent: "What additional control measures are needed for medium-risk areas?" The AI ​​agent retrieved the risk control rule base and replied: "For medium-risk areas, two measures need to be added: 'holding a safety briefing meeting before work' and 'recording equipment status every 2 hours.' These have been automatically added to the 'pre-control measures' field of the survey form."

[0101] (7) Topology calculation stage: intelligent topology diagram generation and power outage range / energized part analysis Automatic generation of topology diagrams and single-line diagrams: Data preprocessing: The system calls the data preprocessing unit of the topology calculation module (patent 2.2.4.1) to extract standardized data (equipment ID, type, connection relationship) of the associated equipment of the #2 main transformer from the graph database and remove invalid data (such as irrelevant nodes of the #3 main transformer). Single-line diagram drawing: The system automatically generates a single-line diagram of the 110kV#2 main transformer bay according to preset layout rules. ▪ Horizontal layout: 110kV busbar (ID4002) → #2-2 disconnector (ID3002) → #2-1 circuit breaker (ID2002) → #2 main transformer (ID1002) → 35kV busbar (ID5002); ▪ Status label: Operating equipment (green): 110kV busbar, 35kV busbar; Equipment under maintenance (yellow): #2 main transformer; Interactive function: When Liu XX clicks on the #2-1 circuit breaker node, the system pops up a window to display the equipment parameters (model: VS1-12, commissioning time: June 2019) and prompts through the intelligent agent: "This circuit breaker is the core control equipment of the #2 main transformer. It must be opened first during a power outage."

[0102] (8) Intelligent calculation of power outage range: Rule matching: The system receives the task tag "#2 main transformer planned maintenance" and automatically matches it with the "substation main transformer maintenance rules" in the preset professional rule library (tag "switch and maintenance": all equipment from the handcart switch to the grounding switch). Topology traversal: Starting from the #2 main transformer node (ID1002), the DFS algorithm is used to traverse the associated devices according to the rules: #2-1 circuit breaker (ID2002), #2-2 disconnector (ID3002), #2 main transformer body (ID1002), and #2 main transformer side current transformer (ID6002). Output of power outage range: Generate a list of power outage equipment (including equipment ID, name, and type) and mark the power outage range (covering the area of ​​#2-2 disconnector switch → #2-1 circuit breaker → #2 main transformer → instrument transformer) with a red solid line box on the single-line diagram.

[0103] (9) Retain intelligent labeling of live parts: Logical judgment: The system compares the "Full Equipment List of Substation" with the "Power Outage Scope List" and selects equipment not included in the power outage scope as the energized parts to be retained, specifically: 110kV bus (ID4002), 35kV bus (ID5002), and all equipment in the #1 main transformer bay (ID1001, etc.). Visual annotation: Mark the energized parts with yellow dashed boxes on the single-line map, and prompt through the intelligent agent pop-up window: "The energized parts are the 110kV bus (voltage 110kV) and the 35kV bus (voltage 35kV). Safety fences must be set up to isolate them during the survey. Crossing the yellow dashed area is prohibited." Verification and Confirmation: The system automatically verifies the logical consistency between the power outage area and the energized parts (e.g., "the power outage area does not include the busbar, which complies with the specification that the main transformer maintenance does not affect the power supply to the busbar"). After the verification is passed, the "#2 Main Transformer Maintenance Power Outage Area and Emergent Part Confirmation Form" is generated for Liu XX to download and archive.

[0104] (10) Survey Completion Stage: Report Generation and Archiving Automatic verification: The system performs a logical consistency check on the survey content (e.g., "The isolating switch has been marked as corroded, and rust removal and pre-control measures have been added, which is consistent with the above statements"). The verification passes. Report generation: The system extracts survey data from the data warehouse and automatically fills in the substation professional survey report template. The "Outage Area Analysis" section inserts a single-line diagram (with red / yellow box markings), and the "Intelligent Q&A Record" section includes an intelligent agent interaction log. Archiving and Signing: After Liu XX approves the document, he and Wang XX complete the electronic signature. The system will archive the survey form, report, and single-line map to the "Survey Record Library" (retention period ≥ 5 years) and synchronize it to the work plan system, updating the plan status to "Survey Completed".

[0105] (11) Implementation effect verification and efficiency improvement Data statistics efficiency: The time for compiling data related to the maintenance and inspection of 110kV main transformer (126 historical orders and 5 types of equipment ledgers) has been shortened from the traditional 2 hours to 8 minutes (meeting the requirement of "efficiency improvement of more than 60%)). The accuracy rate of pre-generated work location and work content is 87%. The intelligent agent reduces the amount of manual data entry by 60%, and the average number of survey tasks per person per day increases from 1.2 to 1.8 (in line with the "1.5 times increase"). Power outage area analysis: The time required for traditional manual drawing and calculation (1.5 hours) has been shortened to automatic generation by the system (10 minutes), improving efficiency by 89%.

[0106] (12) Accuracy guarantee Risk level assessment: The medium risk assessment in this study is consistent with the results of human experience assessment, with an accuracy rate of 100% (meeting the requirement of "≥90%)". Power outage area analysis: On-site verification confirmed that no power outage equipment was omitted or redundant, with an accuracy rate of 100% (meeting "≥95%)". Preventive measures matching: The preventive measures output by the intelligent agent cover all hazards (fall from height, oil leak, electric shock, corrosion), with a matching accuracy of 92% (meeting "≥85%"). The technical solution of this invention will be further described in detail below with reference to the accompanying drawings.

[0107] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An intelligent management method for power grid field survey based on artificial intelligence, characterized in that, include: Acquire equipment ledgers, historical survey reports, work plans, GIS spatial and equipment status data, and write them into a graph database to form a basic dataset; Based on the association data and rules between devices and locations in the aforementioned basic dataset, candidate work locations are obtained, and a candidate work location set is generated. By combining the basic dataset and the candidate work location set, a work content set corresponding to the target survey order is generated; Based on the basic dataset and the work content set, a scoring detail and risk level are calculated, and a risk assessment result is generated accordingly. Based on the device topology relationships in the graph database, a list of power outage areas is generated according to the work objects in the work content set; With the support of the rule base and terminology base, the safety measures text corresponding to the hazard points and the pre-control measures are output by combining the work content set and the power outage range list; Collect on-site images, videos, environmental and equipment operation data, and obtain the survey content submitted by the survey personnel. Combine the collected data with the survey content to obtain the on-site collected data. The consistency and integrity of the field-collected data are verified with the basic dataset, the work content set, and the safety measures text, and verification results are generated. The basic dataset, the candidate set of work locations, the set of work content, the risk assessment results, the list of power outage areas, the safety measures text, the on-site collected data, and the verification results are aggregated to generate a report file associated with the target survey form.

2. The intelligent management method for power grid field survey based on artificial intelligence according to claim 1, characterized in that, The specific process of acquiring equipment ledgers, historical survey reports, work plans, GIS spatial and equipment status data, and writing them into the graph database includes: Equipment ledger data is extracted from the power grid equipment management system. The equipment ledger data includes equipment number, usage time, voltage level, equipment type, equipment model, equipment maintenance record, operating status, bay information, and equipment installation location. Historical survey order data is extracted from the historical survey management system. The historical survey order data includes survey location, survey equipment name, maintenance type, risk level, pre-control measures, and process status information. Extract work plan data from the work planning system, the work plan data including planned time, person in charge, and work scope information; Obtain GIS geospatial data from the GIS platform, wherein the GIS geospatial data includes the geographical location information of the device and the topological relationship of the device; The equipment ledger data, the historical survey data, the work plan data, and the GIS geospatial data are standardized and written into a graph database, which is a Neo4j graph database. When storing data, by establishing a model of nodes and relationships, the equipment ledger data, the historical survey form data, and the work plan data are stored as different node types respectively; Extract the relationships between different nodes, and construct the relationships between nodes based on the relationships.

3. The intelligent management method for power grid field survey based on artificial intelligence according to claim 2, characterized in that, The specific process for generating the candidate set of work locations includes: Based on the "exploration equipment-work location" association data in the historical survey data, the association rule mining algorithm is used to train the data, extract the association rules between the exploration equipment and the work location, and generate a rule base that can be used for work location prediction. When the survey equipment information is input, a candidate set of work locations is automatically generated based on the trained rule base, and sorted according to the matching confidence. The work location with the highest confidence is output as the suggested survey location. The steps of the association rule mining algorithm include: Preprocess historical survey data to extract effective correlation information between survey equipment and work sites; The frequent itemset mining method is used to extract the combination of exploration equipment and work location from historical exploration data, calculate the corresponding confidence level, and generate association rules. Based on the association rules, the device information is automatically matched, and a candidate set of work locations that meet the conditions is output. If the input exploration equipment information cannot be matched in the existing rules, a prompt message will be sent to the exploration personnel to manually input the work location. New combinations of exploration equipment and work locations will be automatically added to the rule base to update and optimize the generation of subsequent candidate sets.

4. The intelligent management method for power grid field survey based on artificial intelligence according to claim 3, characterized in that, The process of generating the work content set specifically includes: Based on the survey equipment, work location, and maintenance type, a work content prediction model is constructed. The work content prediction model is trained using the "survey equipment-work location-maintenance type-work content" data from historical survey orders to extract the prediction rules for the work content. The training of the work content prediction model adopts the gradient boosting tree algorithm. The feature variables in the training process include: the type of survey equipment, the specific information of the work location, the maintenance type, the operating status of the equipment, and the job type. After inputting the survey equipment, work location, and maintenance type information, the work content prediction model generates corresponding work content information based on the trained rules. The work content information includes maintenance items, work steps, and work requirements. The work content information is filtered and optimized, and the final work content is determined based on the matching degree and operability of the work steps and requirements. If the generated work content fails to pass the optimization screening, a prompt message will be sent to the survey personnel to manually fill in the work content. The new combination of "survey equipment-work location-maintenance type-work content" will be automatically included in the subsequent rule base to continuously update and optimize the work content prediction model.

5. The intelligent management method for power grid field survey based on artificial intelligence according to claim 4, characterized in that, The specific process for generating risk assessment results includes: A multi-dimensional risk assessment model is constructed based on preset assessment dimensions, and the weight coefficient of each dimension is calculated using the analytic hierarchy process. According to the evaluation dimensions, scores are given based on the preset scoring criteria, and the risk outcome of each task is calculated in conjunction with the weighting coefficients. Based on the risk results and the preset risk level classification criteria, each task is classified into low, medium, or high risk levels.

6. The intelligent management method for power grid field survey based on artificial intelligence according to claim 5, characterized in that, The specific process for generating the power outage area list includes: Based on the equipment ledger data and the work plan data, the scope of the power outage equipment is calculated; Based on the equipment topology, combined with the job type and job content, the power outage range of the job area is determined; Based on the electrical connections and voltage levels between the equipment, determine a list of outage areas that includes all outage equipment and their connected equipment; The power outage areas in the power outage area list are displayed using visual icons.

7. The intelligent management method for power grid field survey based on artificial intelligence according to claim 6, characterized in that, The process of generating the safety measures text specifically includes: Based on the work content set and the power outage range list, and combined with the safety measure templates in the rule base, corresponding safety measure texts are automatically generated. The safety measure texts include pre-control measures for survey equipment, workers and the environment. Match the safety measures text with the corresponding work plan, work scope, and survey equipment; If the generated safety measures do not meet the requirements, provide the surveyors with a prompt to modify or supplement them.

8. The intelligent management method for power grid field survey based on artificial intelligence according to claim 7, characterized in that, The specific process of collecting data on-site and generating verification results includes: Real-time acquisition of on-site images, videos, environmental parameters, and equipment operating status data via mobile terminals and sensors; The collected data is automatically labeled with metadata including time, location, and device number, and then stored. Through data synchronization, all field-collected data is automatically populated into the corresponding fields of the relevant survey records; The rule engine is used to verify the consistency and integrity of the on-site collected data, the basic dataset, the work content set and the safety measures text, including data format verification, field integrity verification and logical consistency verification, to ensure that there is no inconsistency or missing information among all data and that they meet the actual operation requirements. A verification report is generated based on the verification results. Inconsistencies are marked as errors and fed back to the surveyors for modification or supplementation.

9. The intelligent management method for power grid field survey based on artificial intelligence according to claim 8, characterized in that, The specific process for generating the report file includes: Based on the surveying profession, select the appropriate report template, summarize the basic dataset, the candidate set of work sites, the set of work content, the risk assessment results, the list of power outage areas, the safety measures text, the field-collected data, and the verification results, and automatically generate a standardized report file; The report includes a cover, survey overview, detailed work content, risk assessment, safety measures, equipment parameters, and verification results; The report file is format-checked, supports the generation of PDF or Word files, and supports electronic signatures and approval processes for surveyors.

10. An intelligent management system for power grid field survey based on artificial intelligence, characterized in that, The method for intelligent management of power grid field survey based on artificial intelligence, as described in any one of claims 1 to 9, includes: The data acquisition module is used to acquire equipment ledgers, historical survey reports, work plans, GIS spatial and equipment status data, and write them into the graph database to form a basic dataset; The location candidate module is used to obtain candidate work locations based on the association data and rules between devices and locations in the basic dataset, and generate a candidate set of work locations; The work content module is used to combine the basic dataset and the candidate work location set to generate a work content set corresponding to the target survey order; The risk assessment module is used to calculate and form a scoring detail and risk level based on the basic dataset and the work content set, and generate a risk assessment result accordingly; The power outage list module is used to generate a power outage range list based on the device topology relationship in the graph database and the work objects in the work content set. The safety measure text module, supported by the rule base and terminology base, is used to output safety measure texts corresponding to hazard points and pre-control measures, in combination with the work content set and the power outage range list; The on-site data acquisition module is used to collect on-site images, videos, environmental and equipment operation data, and obtain the survey content submitted by the survey personnel. The on-site data is obtained by combining the collected data with the survey content. The result verification module is used to verify the consistency and integrity of the field-collected data with the basic dataset, the work content set, and the safety measures text, and generate verification results. The report generation module is used to aggregate the basic dataset, the candidate set of work locations, the set of work content, the risk assessment results, the list of power outage areas, the safety measures text, the on-site collected data, and the verification results to generate a report file associated with the target survey form.

Images