Power grid secondary equipment fault tracing and operation and maintenance service method and system

Abstract

This invention discloses a method and system for fault tracing and maintenance services of secondary power grid equipment, relating to the field of power grid secondary equipment fault handling technology. The method includes: establishing an enterprise data gene library; fault knowledge extraction, data cleaning, and knowledge storage; establishing a business middle platform to acquire data transmitted through channels; building a fault tracing platform; constructing a product fault tracing and maintenance service knowledge graph database, and designing and developing a knowledge graph search engine; implementing fault tracing; updating the knowledge graph database; and constructing a fault tracing domain knowledge graph for new products using rule-based reasoning algorithms. This invention solves the problem of enterprise data silos by integrating data from R&D, procurement, and production processes and rapidly building a fault tracing platform, thereby quickly locating the cause of faults and improving fault tracing efficiency; and updates the domain knowledge graph database based on business operations, equipment faults, and resolution status.

Description

Technical Field

[0001] This invention relates to the field of power grid secondary equipment fault handling technology, and in particular to a method and system for power grid secondary equipment fault tracing, operation and maintenance services. Background Technology

[0002] Currently, enterprises engaged in the research and development and manufacturing of secondary power grid equipment manage their products from sales to after-sales service according to the sales, design, procurement, production and service processes of manufacturing enterprises. The resulting business data is scattered in the corresponding business systems: ERP, MES, PLM, engineering service and other systems. Data exchange between systems is completed through interfaces, resulting in low information utilization.

[0003] When secondary equipment malfunctions on-site, it typically requires after-sales service personnel to arrive, analyze the symptoms, locate the fault, communicate with company technical support to clarify the problem, and then the company mails spare parts to the site. The after-sales service personnel then replace and test the parts to resolve the issue. This process heavily relies on the technical skills and experience of the after-sales personnel. Existing power equipment information systems primarily focus on business processes, neglecting the utilization of business data and exhibiting severe data silos. Current power equipment information systems only register equipment fault conditions. Fault tracing for power equipment requires integration with ERP, MES, and PLM systems. Since these systems store data in different formats, data conversion is necessary, hindering direct statistical analysis and resulting in low efficiency.

[0004] Chinese patent document CN113360555A discloses a "fault diagnosis and analysis method and system based on big data of power grid secondary equipment." This method extracts historical data from faulty substations from a unified power grid data platform, extracts instantaneous current and voltage values ​​at the time of the fault, and performs fault analysis to generate a fault analysis report and solution. However, this technical solution only focuses on business processes, merely registering equipment fault conditions, and lacks sufficient utilization of data from other business systems, resulting in a serious data silo problem. Fault tracing of power equipment requires integration with ERP, MES, and PLM systems. Since these systems store data in different formats, there is an urgent need for a method and system for power grid secondary equipment fault tracing and maintenance services that can comprehensively read real-time fault data from multiple business systems and directly perform statistical analysis. Summary of the Invention

[0005] The technical problem to be solved by this invention is to overcome the shortcomings of the existing technology, based on the full life cycle management model of power grid secondary equipment products, utilize the data accumulated by various information systems to form a knowledge base, apply knowledge graph and other technologies to the fault tracing of power grid secondary equipment, improve the intelligence level of fault tracing, and provide a method and system for fault tracing and operation and maintenance services of power grid secondary equipment.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0007] A method for fault tracing and maintenance services of secondary power grid equipment, characterized by comprising the following steps:

[0008] S1 business and data analysis, and the establishment of an enterprise data gene library;

[0009] S2 fault knowledge extraction, data cleaning, and knowledge storage: It extracts existing defect records and corresponding handling solutions from the information system in a rule-based manner, performs data analysis and cleaning on the extracted defect records and corresponding handling solutions, stores them, and establishes a knowledge base.

[0010] S3 establishes a business middle platform based on virtualization and container technologies. It connects to the data channel configured in the data gene library through connectors, obtains the data transmitted through the channel, and saves it in a data model. The data model is bound to a specific business model.

[0011] S4 builds a fault tracing platform, which is built on the basis of the completed data gene library and business middle platform;

[0012] The S5 knowledge graph database is constructed based on the data generated by the fault knowledge extraction. A product fault tracing and operation and maintenance service knowledge graph database is constructed. A graph database is used to store the object relationships of the knowledge graph, and the knowledge graph is displayed through a visual interface.

[0013] S6 fault tracing implementation;

[0014] The S7 knowledge graph database is updated continuously based on business operations, equipment failures, and their resolution, enabling closed-loop management for fault tracing.

[0015] S8 knowledge graph reasoning, based on the original product knowledge graph, constructs a fault tracing domain knowledge graph for new products with modified designs through rule-based reasoning algorithms.

[0016] Furthermore, the establishment of the enterprise data gene bank includes the following steps: S1.1 Organizing enterprise metadata, compiling an enterprise data asset catalog, and creating a data sharing and open catalog, including structured data and related semi-structured data from supplier, product, and engineering service knowledge bases; S1.2 Based on data search, providing information such as table usage instructions, data categories, data lineage, and field lineage to establish the enterprise data gene bank; S1.3 Building bridges between data, organically linking various elements such as business data, product design, and departmental responsibilities to form data standards.

[0017] Furthermore, the fault knowledge extraction, data cleaning, and knowledge storage include the following steps: S2.1 Semantic processing of defect records and handling solutions; S2.2 Extracting stems from the semantically processed data using natural language processing methods; S2.3 Data cleaning of the extracted data, including missing data determination, missing data processing, noise data detection, and noise data processing; S2.4 Establishing a defect-solution knowledge base by classifying entities, attributes, and relations into entity-relationship-entity triples and entity-attribute value pairs.

[0018] Furthermore, the missing data determination includes using heatmaps to quickly locate missing data, using the `info` method to find missing attribute values, and using the `apply` method to calculate the missing rate. The missing data processing includes directly deleting missing rows and columns, imputing with the mean, imputing with the nearest neighbor, and imputing with a hot card. The noise data detection includes the mean-standard deviation method and the upper and lower four median and median difference methods, classifying values ​​outside the intervals [mean - 2x standard deviation, mean + 2x standard deviation] and [lower four median - 1.5x median difference, upper four median + 1.5x median difference] as noise data. The noise data processing includes noise data deletion and noise data rewriting; noise data deletion includes deleting entire rows and columns of data; noise data rewriting includes replacing with the mean, replacing with the median, and replacing with the nearest maximum and minimum boundary values.

[0019] Furthermore, the knowledge storage includes: storing structured data using a relational database, and storing documents, drawings, and video data in file format.

[0020] Furthermore, the construction of the fault tracing platform includes the following steps: S4.1 Based on the completed data gene bank and business middle platform, data is extracted to the cloud computing platform through data collection and access to existing multi-source data, establishing a unified standard data modeling system; S4.2 A big data platform is built with business layering as the architecture, completing the construction of systems related to enterprise data value, such as system architecture, data graph, data quality, organizational structure, and standardized processes, to achieve rapid integration of data, business, and intelligence; S4.3 A unified service exit is provided to offer distinctive data services, enabling no-code development and intelligent performance optimization of data services, providing support for upper-layer business applications. S4.4 Utilizes data intelligence capabilities; S4.4 Identifies the production and management issues that need to be addressed, as well as the main business functions that the system needs to implement, focusing on enterprise procurement, planning, product details, raw material inventory, workshop order batch information, scheduled production plans, and supply requirements; S4.5 Adopts a microservice development model, based on the aforementioned big data platform and highly reusable software development platform, to complete the system's functional development. Through UI construction, various UI components are dragged and dropped to integrate the model's data, services, and functions into the fault tracing platform; S4.6 Publishes the completed fault tracing platform to the application platform through the publishing function for user use.

[0021] Furthermore, the featured data services include intelligent data query services, complex data query services, and real-time data push services.

[0022] Furthermore, the fault tracing implementation includes the following steps: S6.1 Sending a fault tracing form to the first-level user, who uses a knowledge graph search engine to query the cause of the fault and provides feedback to the second-level user; S6.2 The second-level user queries the corresponding logical relationship classification of the faulty product in the fault tracing platform and queries products from the same batch; S6.3 Tracing the problematic finished products, semi-finished products, and other user sites; S6.4 Tracing the materials and software used, determining whether these materials will affect other products, and forming a traceability system.

[0023] Furthermore, the corresponding logical relationship classification of the faulty products includes in stock, in-process, and delivered.

[0024] This invention also provides a fault tracing and maintenance service system for power grid secondary equipment, comprising: a data gene library module, providing information such as table usage instructions, data categories, data lineage, and field lineage; a knowledge base module, storing defect records and corresponding processing solutions; a business middle platform module, connecting to the data gene library's data channel via a connector to acquire data transmitted through the channel and storing it using a data model; a fault tracing platform module, enabling rapid integration of data, business, and intelligence, and providing specialized data services; and a knowledge graph library module, using a graph database to store the object relationships of the knowledge graph and displaying the knowledge graph through a visual interface.

[0025] The beneficial effects of this invention are:

[0026] 1. Improve the usability of the fault tracing knowledge base, making the display intuitive and easy to understand, reducing employee training and usage costs, and lowering after-sales service costs for enterprises; by utilizing the fault tracing and maintenance service system, the cause of faults can be quickly located, solving the problem of low tracing efficiency, improving customer service levels, and enhancing user satisfaction with products.

[0027] 2. To address the issues of multi-source and heterogeneous data encountered during digital transformation, we will sort out the data and business processes related to enterprise R&D, procurement, and production, establish data space and business space, and apply them in the rapid development of enterprise product intelligent service software to solve the problem of enterprise data silos.

[0028] 3. Deliver enterprise product intelligent services as a new service model to power grid companies for on-site application and practical use; build a private cloud platform for enterprise product intelligent services to provide remote operation and maintenance management services for small and medium-sized enterprises and reduce their manpower operation and maintenance costs.

[0029] 4. Based on the enterprise data gene library data platform and the business platform for developing intelligent and highly reusable software for multiple terminals on one cloud, a fault tracing platform can be quickly built, solving the problem of low efficiency in building a fault tracing platform.

[0030] 5. Enhance the level of intelligent fault tracing and form a knowledge graph for new products through knowledge reasoning. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a flowchart illustrating the method for tracing faults and providing maintenance services for secondary power grid equipment in this embodiment of the invention.

[0033] Figure 2 This is a schematic diagram of the structural composition of the power grid secondary equipment fault tracing and operation and maintenance service system in an embodiment of the present invention. Detailed Implementation

[0034] Exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that the following description is explanatory and exemplary in nature only and is not intended to limit the invention or its application or use. Unless otherwise specifically stated, the relative positions, numerical expressions, and values ​​of components and steps set forth in the embodiments do not limit the scope of the invention. Furthermore, techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but are intended to form part of the specification where appropriate.

[0035] This invention, based on the enterprise's full lifecycle management model for power grid secondary equipment, utilizes data accumulated from various enterprise information systems to form a knowledge base. Knowledge graph and other technologies are applied to fault tracing of power grid secondary equipment, improving the intelligence level of fault tracing. Based on the established data and business spaces, an "intelligent, highly reusable software development and operation platform for multiple cloud terminals" is adopted to develop industrial management software. This moves the offline fault tracing model online, achieving informatization and intelligence in fault tracing, and managing product ledger information, fault tracing, and early warning.

[0036] Figure 1 An exemplary flowchart illustrates a general method for fault tracing and maintenance services of power grid secondary equipment according to an embodiment of the present invention. Figure 1 As shown, the method includes the following steps:

[0037] S1 business and data analysis, establishing an enterprise data gene bank.

[0038] S2 fault knowledge extraction, data cleaning, and knowledge storage involves extracting existing defect records and corresponding handling solutions from the information system using a rule-based approach, performing data analysis and cleaning on the extracted defect records and corresponding handling solutions, storing them, and building a knowledge base.

[0039] S3 establishes a business middle platform. Based on virtualization and container technologies, it designs an industrial management software business middle platform. The business middle platform connects to the data channel of the data gene library through connectors, obtains the data transmitted through the channel, and saves it in a data model. The data model is bound to a specific business model to provide data support for the business model, thus becoming a set of highly reusable components for quickly building a fault traceability platform.

[0040] S4 builds a fault tracing platform, which is built on the basis of the completed data gene library and business middle platform. The fault tracing platform includes several sets of highly reusable components that are bound to data models and business models, providing data support for business models.

[0041] The S5 knowledge graph database is constructed based on the data generated from the extracted fault knowledge. A product fault tracing and maintenance service knowledge graph database is built, using a graph database to store the object relationships within the knowledge graph. The knowledge graph is displayed through a visual interface. A knowledge graph search engine is designed and developed to support rapid knowledge base searches. This search engine has a built-in fault keyword matching index matrix, which can automatically match industry synonyms, fixed terms, and Chinese and English abbreviations based on fault keywords. By setting keyword matching value parameters, the search results are sorted according to the matching value and returned to the user, enabling fast and accurate knowledge base searches.

[0042] S6 fault tracing implementation.

[0043] The S7 knowledge graph database is updated continuously based on business operations, equipment failures, and their resolution, enabling closed-loop management for fault tracing.

[0044] S8 knowledge graph reasoning, based on the original product knowledge graph, constructs a fault tracing domain knowledge graph for modified products using rule-based reasoning algorithms. Specifically, it employs a neural network-based knowledge graph reasoning model (NTN) to extract two actual power grid secondary equipment fault entities (e1, e2) (e1, e2 ∈ R) from the knowledge base. d (This is a vector representation of the entity), summarizing the relationship between the instantaneous values ​​of voltage, current, power, power factor, phase, and frequency of each fault entity and equipment parameters and service modules. Reasoning and calculation formula:

[0045] g(e1, R, e2) = u T R f(e T 1W R [1:k] e2+V R [e1e2]+b R

[0046] Compared to traditional standard linear neural network fault tracing methods, this knowledge graph reasoning model uses a bilinear tensor layer to associate two entity vectors. In this model, the entity vector is the average of all word vectors, meaning the number of words in an entity is much smaller than the number of entities, allowing for efficient reuse of word vectors to construct entity representations. Different slices of the relation tensor correspond to semantic relationships between different entity vectors, enhancing these relationships. By calculating the probability score of a relationship between two entities, this model can be applied to products with varying designs, enabling relation prediction and knowledge base completion in the fault tracing domain of new products.

[0047] In a preferred embodiment, the business and data analysis described in step S1 includes analyzing the current state of the company's business, including the company's R&D and manufacturing processes, production processes, supplier and customer information, product information, equipment operating status, historical faults and maintenance, intelligence level, data collection methods and capabilities, and equipment monitoring application level in the smart grid secondary equipment industry. The establishment of the enterprise data gene bank described in step S1 includes: analyzing enterprise metadata, compiling an enterprise data asset catalog, and creating a data sharing and open catalog, including structured data such as supplier, product, and engineering service knowledge bases, as well as related semi-structured data; based on data search, providing table usage instructions, data categories, data lineage, field lineage, and other information to establish the enterprise data gene bank; and building bridges between data sources to organically link business data, product design, departmental responsibilities, and other elements to form industrial enterprise data standards. Addressing the issues of multi-source and heterogeneous data encountered during digital transformation, this step analyzes the data and business related to the company's R&D, procurement, and production processes, establishing data and business spaces to solve the problem of enterprise data silos.

[0048] In a preferred embodiment, the fault knowledge extraction, data cleaning, and knowledge storage in step S2 include extracting existing defect records and corresponding processing solutions from various existing information systems such as the company's ERP system, MES system, PLM system, and engineering service system. Semantic processing is performed on the defect records and processing solutions, and natural language processing methods are used to extract the stem from the semantically processed data. Data cleaning is performed on the extracted data, including missing data determination, missing data processing, noise data detection, and noise data processing. A defect-solution knowledge base is established by organizing entities, attributes, and relationships into entity-relationship-entity triples and entity-attribute value pairs.

[0049] Specifically, the fault knowledge extraction described in step S2 uses the Harbin Institute of Technology Language Cloud Platform (LTP-Cloud). Through the latest version of the Chinese word segmentation interface, it combines the Conditional Random Fields (CFRs) algorithm with dictionary-based matching and statistical machine learning methods to segment the defect records and processing solutions. Then, it uses deep learning-based natural language processing technology to extract the stem. This platform has a fast running speed, high stem extraction accuracy, and good overall application effect.

[0050] The data cleaning described in step S2 includes four steps: missing value identification, missing value processing, outlier detection, and outlier handling. These steps aim to remove noisy outliers, appropriately fill in missing values, and accurately remove duplicate values, thereby improving data accuracy. Missing data identification includes: using heatmaps to quickly locate missing data, making the location and degree of missing values ​​immediately apparent; using the `info` method to discover missing attribute values, which is simple and fast, allowing for rapid location of missing attributes; and using the `apply` method to calculate the missing rate, displaying the results as a percentage for convenience and speed. Missing data processing includes directly deleting missing rows and columns, filling in missing values ​​using the mean, nearest distance, or hotcard methods. Noisy data detection includes the mean-standard deviation method and the upper / lower four median and median difference method, classifying values ​​outside the intervals [mean - 2x standard deviation, mean + 2x standard deviation] and [lower four median - 1.5x median difference, upper four median + 1.5x median difference] as noisy data. The noise data processing includes noise data deletion and rewriting; noise data deletion includes deleting entire rows and columns of data; noise data rewriting includes replacing the average value, median value, and nearest maximum and minimum boundary values. This data cleaning process has advantages such as low error, high efficiency, and low cost, effectively ensuring data stability and security.

[0051] The fault knowledge storage described in step S2 includes extracting and cleaning data from multiple systems with different storage formats according to the above rules, so that statistical analysis can be performed directly. Relational databases are used to store structured data, while documents, drawings, videos, and other data are stored in file format.

[0052] In a preferred embodiment, step S3, based on virtualization and container technologies, includes designing and developing an industrial management software business platform. The business platform connects to the data channel configured in the data gene library through a connector, acquires the data transmitted through the channel, and saves it using a data model. The data model is bound to a specific business model to provide data support for the business model, thus becoming a set of highly reusable components for quickly building a fault tracing platform.

[0053] In a preferred embodiment, the construction of the fault tracing platform in step S4 includes: building the fault tracing platform based on the completed data gene library and business middle platform; extracting data to the cloud computing platform through data collection and access to existing multi-source data; and establishing a unified standard data modeling system. A big data middle platform is constructed with a business-layered architecture, completing the system construction related to enterprise data value, including system architecture, data graphs, data quality, organizational structure, and standardized processes, to achieve rapid integration of data, business, and intelligence. A unified service exit provides three distinctive data services: intelligent data query service, complex data query service, and real-time data push service, enabling code-free development and intelligent performance optimization of data services, and providing data intelligence capabilities for upper-layer business applications. The system identifies production and management problems that need to be solved, as well as the main business functions that the system needs to implement, focusing on enterprise procurement, planning, product detail lists, raw material inventory, workshop order batch information, scheduled production plans, and supply requirements. The system adopts the microservice development model in existing technologies, and completes the functional development of the system based on the big data middle platform and the highly reusable software development platform. Through UI construction, various UI components are dragged and dropped to import the model's data, services and functions into the fault tracing platform. The completed application is then published to the application platform through the publishing function for use by all users.

[0054] In a preferred embodiment, the fault tracing implementation in step S6 includes: sending a fault tracing form to quality department personnel; quality department personnel using the knowledge graph search engine of the fault tracing and maintenance service system to query the cause of the fault and providing feedback to maintenance personnel; maintenance personnel querying the faulty product's corresponding logical relationship classification in the fault tracing and maintenance service system: querying products in the same batch in states such as in stock, in-process, and delivered; tracing the problematic finished products, semi-finished products, and other user sites, all the way back to the materials and software used, determining whether these materials will affect other products, thus forming a traceability system. When a product malfunctions on-site, the fault tracing platform can quickly trace the problematic finished products, semi-finished products, and other user sites, all the way back to the materials used, determining whether these materials will affect other products, thus forming a standardized traceability system.

[0055] Will Figure 2 The power grid secondary equipment fault tracing and maintenance service system shown is a new service model delivered to power grid companies for field application and practical use. It establishes a private cloud platform for intelligent services of enterprise products, providing remote maintenance management services for SMEs and reducing their manpower maintenance costs. Utilizing the fault tracing and maintenance service system (including defects and solutions), the causes of faults can be quickly located, improving customer service levels and enhancing user satisfaction with the products.

[0056] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A power grid secondary equipment fault tracing, operation and maintenance service method, characterized in that, Includes the following steps: S1 business and data analysis, and the establishment of an enterprise data gene library; S2 fault knowledge extraction, data cleaning, and knowledge storage: It extracts existing defect records and corresponding handling solutions from the information system in a rule-based manner, performs data analysis and cleaning on the extracted defect records and corresponding handling solutions, stores them, and establishes a knowledge base. S3 establishes a business middle platform based on virtualization and container technologies. It connects to the data channel configured in the data gene library through connectors, obtains the data transmitted through the channel, and saves it in a data model. The data model is bound to a specific business model. S4 builds a fault tracing platform, which is built on the basis of the completed data gene library, knowledge base, and business middle platform. S5 Knowledge Graph Database Construction: Based on the data generated by the extraction of fault knowledge, a product fault tracing and maintenance service knowledge graph database is constructed, and a knowledge graph search engine is designed and developed. S6 fault tracing implementation; The S7 knowledge graph database is updated continuously based on business operations, equipment failures, and their resolution, enabling closed-loop management for fault tracing. S8 knowledge graph reasoning, based on the original product knowledge graph, for modified designs of products, constructs a fault tracing domain knowledge graph for new products through a neural network-based knowledge graph reasoning NTN model; it extracts two actual power grid secondary equipment fault entities from the knowledge base, summarizes the relationships between the instantaneous values ​​of voltage, current, power, power factor, phase and frequency of each fault entity and equipment parameters and business modules, and uses a bilinear tensor layer to associate the two entity vectors to construct relationship prediction and knowledge base completion for the fault tracing domain knowledge graph of new products; The establishment of the enterprise data gene bank includes the following steps: S1.1 Organize enterprise metadata, compile an enterprise data asset catalog, and compile a data sharing and open catalog, including: structured data and related semi-structured data of supplier, product, and engineering service knowledge bases; S1.2 is based on data search and provides table usage instructions, data categories, data lineage, and field lineage information to establish an enterprise data gene library; S1.3 Build bridges between data, organically linking business data, product design, departmental responsibilities, and other elements to form data standards; The steps to build the fault tracing platform are as follows: S4.1 Based on the completed data gene bank and business middle platform, data is extracted to the cloud computing platform through data collection and access to existing multi-source data to establish a unified standard data modeling system; S4.2 Builds a big data platform with business layering as the architecture, and completes the construction of a system related to system architecture, data graph, data quality, organizational structure, standardized processes and enterprise data value, so as to achieve rapid integration of data, business and intelligence; S4.3 provides a unified service exit, offering distinctive data services, enabling no-code development and intelligent performance optimization of data services, and providing data intelligence capabilities for upper-layer business applications; S4.4 outlines the production and management issues that need to be addressed, as well as the main business functions that the system needs to achieve, focusing on enterprise procurement, planning, product details, raw material inventory, workshop order batch information, scheduled production plans, and supply demand. S4.5 adopts a microservice development model. Based on the big data platform and highly reusable software development platform, it completes the system's functional development. Through UI construction, various UI components are dragged and dropped to import the model's data, services, and functions into the fault tracing platform. S4.6 The established fault tracing platform is published to the application platform through the publishing function for user use; The fault tracing implementation includes the following steps: S6.1 Sends a fault tracing form to the first-level user, who uses a knowledge graph search engine to query the cause of the fault and provides feedback to the second-level user. S6.2 Second-level users can query the corresponding logical relationship classification of faulty products and query products in the same batch in the fault tracing platform; the corresponding logical relationship classification of faulty products includes in stock, in-process, and delivered. S6.3 Trace the problematic finished products, semi-finished products, and other user sites; S6.4 Trace the materials and software used, determine whether these materials will affect other products, and form a traceability system.

2. The grid secondary equipment fault backtracking, operation and maintenance service method according to claim 1, characterized in that, The fault knowledge extraction, data cleaning, and knowledge storage include the following steps: S2.1 Perform semantic processing on defect records and handling plans; S2.2 Use natural language processing methods to extract the stem from the semantically processed data; S2.3 Perform data cleaning on the extracted data, including missing data identification, missing data processing, noise data detection, and noise data processing; S2.4 Establish a knowledge base corresponding to defects and solutions by organizing entities, attributes, and relations into entity-relationship-entity triples and entity-attribute value pairs.

3. The method for fault tracing and maintenance services of power grid secondary equipment according to claim 2, characterized in that: The missing data determination includes: using heatmaps to quickly find missing data, using the info method to find missing attribute values, and using the apply method to calculate the missing rate; The missing data processing includes: direct deletion of missing row data and column attributes, filling with mean, filling with nearest distance, and filling with hot values ​​to fill missing values; The noise data detection includes: using the mean-standard deviation method and the upper and lower four medians and median difference method, determining the values ​​that are not in the interval [mean-2xstandard deviation, mean-2xstandard deviation] and the interval [lower four medians-1.5xmedian difference, upper four medians+1.5xmedian difference] as noise data; The noise data processing includes: noise data deletion and noise data rewriting; the noise data deletion includes deleting entire rows of data and deleting entire columns of data; the noise data rewriting includes replacing the average value, replacing the median value, and replacing the nearest maximum and minimum boundary values.

4. The grid secondary equipment fault backtracking, operation and maintenance service method according to claim 2, characterized in that, The knowledge storage includes: storing structured data using a relational database, and storing documents, drawings, and video data in file format.

5. The grid secondary equipment fault backtracking, operation and maintenance service method of claim 1, wherein, The featured data services include: intelligent data query service, complex data query service, and real-time data push service.

Images