A method and system for structuring extraction of power maintenance data based on a large model

Abstract

The application discloses a kind of based on the structured extraction method and system of power maintenance data of large model, comprising: collecting original power maintenance data and processing;Load the standardized naming mapping table of power system equipment and parse;Configuration and call large language model, the structured information of target field is obtained by feature extraction to preprocessed text to be analyzed;Using the semantic similarity calculation based on vector embedding, obtain the corresponding preliminary matching result;According to the standardized algorithm, further calculate the preliminary matching result, establish the standardized mapping index table of current plant and each level equipment in station;Standardized mapping index table is integrated, and it is packaged into system service or offline file form output.The application realizes the efficient, accurate structured extraction and standardized processing of power maintenance data, can process complex unstructured text input, through semantic understanding and multistage matching mechanism, the extracted information is accurately corresponding with standardized equipment and plant name.This process greatly improves the automation degree and accuracy of data processing.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system data processing, and particularly relates to a method and system for structuring extraction of power maintenance data based on a large model. BACKGROUND

[0002] In the daily operation of the power system, the accuracy, consistency and traceability of the maintenance data are crucial for ensuring the safe and stable operation of the power grid. However, a common problem at present is that the filling of maintenance content mainly relies on manual operation, which leads to many challenges in data quality and consistency:

[0003] 1. Data inconsistency: Due to the differences in personal habits, professional backgrounds and expression methods of the filling personnel, even for the maintenance content of the same equipment, different description methods may appear. For example, the same transformer may be described as "XX transformer", "XX main transformer" or habitual abbreviations, etc.

[0004] 2. Lack of standardization: Although the power equipment has standardized naming in the power system, because the equipment involved in the maintenance is complex and the operation is complicated, if the equipment selection method is used for data filling in the maintenance system, it will increase the operation complexity and affect the operability of the system, so the current maintenance system often does not realize strict data standardization. Filling personnel may use non-standardized abbreviations, colloquial expressions or personal habits, which makes subsequent data processing and analysis complicated.

[0005] 3. Information silos between systems: Due to the lack of unified naming conventions and data standards, data between different systems is difficult to effectively associate and share, forming information silos.

[0006] 4. Difficulty in business closed loop: Data inconsistency and non-standardization seriously hinder the closed loop of cross-system business processes. For example, it is difficult to accurately track the complete maintenance history of a specific equipment, or to accurately match the maintenance records with the equipment information in the asset management system.

[0007] 5. Limited data analysis: Unstructured and non-standardized data formats greatly limit the analysis value of data. This not only increases the workload of data preprocessing, but also may lead to biased or erroneous analysis results.

[0008] 6. Heavy burden of manual processing: Due to the inconsistency of data, a large amount of manual intervention and data cleaning work is necessary, which not only consumes time and effort, but also easily introduces new errors.

[0009] 7. Insufficient decision support capability: Data quality problems directly affect the effectiveness of decision support systems based on these data, which may lead to improper allocation of resources or unreasonable maintenance strategy formulation.

[0010] The above problems not only affect the efficiency of daily operation, but also pose potential risks to the safe and stable operation of the power grid. With the development of the power system towards intelligence and digitization, the requirements for data quality are becoming higher and higher, and traditional manual reporting and simple data processing methods have been difficult to meet the needs of modern power systems.

[0011] Therefore, there is an urgent need for a method and system that can automatically process non-standardized maintenance data and achieve structured and standardized extraction. Such a method and system should be able to understand and process diverse manual reporting content, extract key information, and convert it into standardized naming and structured format, thereby supporting cross-system data integration and business closed loop, improving data availability and analysis value. SUMMARY

[0012] In view of the deficiencies of the prior art, the present application utilizes the powerful context understanding capability of large language models to provide a large model-based power maintenance data structured extraction method and system to solve the above problems existing in the prior art.

[0013] The technical solution adopted by the present application to achieve the above-mentioned purpose is: a large model-based power maintenance data structured extraction system, the following system is constructed for multi-source complex heterogeneous text, through semantic understanding and multi-level matching mechanism, the extracted information is accurately corresponding to the standardized equipment and plant station name, provided to the maintenance user, the system includes:

[0014] A data acquisition and preprocessing module collects raw power maintenance data, and performs data cleaning and formatting processing;

[0015] A standardized mapping table preparation module loads and parses the standardized naming mapping table of the power system equipment;

[0016] An unstructured text feature extraction module configures and calls a large language model to extract features from the preprocessed text to be analyzed, obtaining structured information of the target field;

[0017] A semantic similarity module uses vector embedding-based semantic similarity calculation to match and find the standardized naming mapping table for the structured information of the target field, and obtains the corresponding preliminary matching result;

[0018] A plant station and equipment name standardization module further calculates the preliminary matching result according to a standardization algorithm, establishes a standardized mapping index table of the current plant station and each level of equipment within the plant station, and realizes data standardization;

[0019] A data integration and output module integrates the standardized mapping index table and encapsulates it into a system service or offline file form for output to the power grid business system.

[0020] The power maintenance data are multi-source heterogeneous text records, Excel tables, and database data.

[0021] The data acquisition preprocessing module comprises:

[0022] A secure transmission unit acquires data files from a remote server using a secure file transfer protocol; and is used to prevent eavesdropping, tampering or forgery during transmission.

[0023] A local cache unit is used for local caching of data files.

[0024] A data parsing unit parses multi-source heterogeneous data into a unified data format; comprising: using a regular expression to filter a plurality of target data meeting the conditions in the data set, and splitting or searching the target data according to a specified character to obtain a group of split target data.

[0025] A data cleaning unit is used to implement data cleaning, including removing irrelevant characters, unifying delimiters, and standardizing formats.

[0026] The standardized mapping table preparation module comprises:

[0027] A multi-level mapping structure unit is used to collect original mapping tables and parse and process them into ID mapping data of corresponding levels to be respectively tabulated into a plant station mapping table and a device mapping table; and is used to adapt to the needs of different scenarios; wherein the original mapping table is a standardized naming mapping table of the existing power system equipment; wherein each standardized naming mapping table stores the standard naming names of all plants and devices under the current system key value without hierarchical levels; the parsing and processing operation is to parse each entry of the original mapping table of the current plant station into a plant station mapping table and a device mapping table according to the level.

[0028] A mapping data loading unit is used to read ID mapping data from a text file.

[0029] A rule unit is used to define that the standardized names in the ID mapping data are consistent with the naming rules of the data to be analyzed; and match the mapping key values of the standardized names in the ID mapping data; wherein the mapping key values are classification identifiers of different source data, and the naming rules are consistent to enable matching of data from different sources.

[0030] The unstructured text feature extraction module comprises:

[0031] A large language model unit is used to use a large language model to iteratively train input data in combination with a parameter unit to obtain an ideal model, and to realize extraction of structured information of a target field; the key target field includes XX transformer, XX substation, XX plant, XX power plant, XX wind farm, and XX wind power plant.

[0032] A parameter unit is configured to define roles, tasks, instruction templates, output data structures, and domain knowledge of the large model; the tasks, instruction templates, output data structures, and domain knowledge are different under different roles; the domain knowledge inputs common device types and maintenance terms of the power system to the model to help the model learn field-specific expressions.

[0033] The semantic similarity module includes:

[0034] A semantic similarity calculation unit is configured to calculate semantic similarity based on vector embedding, and is used for matching and searching the structured information of the power station and the device name in the to-be-extracted text with the standardized names in the parsed power station mapping table and the device mapping table under different system sources.

[0035] A text vector representation unit is configured to represent a given text t as V(t)=f(t); wherein t is the structured information of the power station and the device name in the to-be-processed text or the standardized name in the parsed power station mapping table and the device mapping table under different system sources, and f is an encoding function.

[0036] A similarity calculation unit is configured to calculate similarity and find the most similar item in different ranges.

[0037] The calculation of similarity includes: for a query text q and a text set T={t1, t2,..., tn}, the similarity of each item is calculated according to the following formula:

[0038]

[0039] Wherein the query text q is the structured information of the power station and the device name in the to-be-processed text, and the text set T is a large set of each standardized name in the parsed power station mapping table and the device mapping table under different system sources, wherein · represents dot product, and |V| represents the Euclidean norm of the vector.

[0040] The finding of the most similar item in different ranges includes:

[0041] The most similar item in the set T is selected as the matching result according to the following formula:

[0042] Calculate most_similar(q, T)=argmax t∈T (similarity(q, t));

[0043] Or, arrange in descending order according to the similarity value, and take the first M as the candidate name set c={c1, c2,..., cn} as the matching result output.

[0044] The standardization algorithm of the station and equipment name standardization module is used to process various naming variants and abbreviations; including:

[0045] a) Edit distance calculation, for two strings s1 and s2, calculate the edit distance D(i, j):

[0046]

[0047] Where s1 is the structured information of the station and equipment name in the text to be processed, s2 is any one of the candidate names c={c1, c2,..., cn}; D(i, j) is the semantic similarity distance;

[0048] b) For candidate name c and standard name s, consider the semantic similarity and edit distance, calculate the standardization score:

[0049] score(c, s) = a • similarity(c, s) + (1-a) • (1-normalized_edit_distance(c, s))

[0050] Where a is the weight parameter, similarity(c,s) is the result calculated according to the similarity formula for c and s, and normalized_edit_distance is the normalized edit distance D(i, j);

[0051] c) For candidate name c and standard name set s, the best match is selected as:

[0052] best_match(c, S) = argmax s∈S score(c, s)

[0053] Thus, the unique standardization mapping index table of the target field of all station and equipment names and the standard naming is established.

[0054] The data integration and output module uses the JSON serialization method to format the data in the standardization mapping index table; uses incremental data processing and output mechanism to support efficient processing and real-time output of large-scale data; for realizing data uploading function, safely transmitting the processing result back to the remote server for use by the maintenance system.

[0055] A structured extraction method of power maintenance data based on a large model, which performs the following steps to identify multi-source complex heterogeneous power maintenance data, and matches and searches in the standardization mapping table to obtain unique standard naming for maintenance users, the method comprising the following steps:

[0056] Step S1, collect original power maintenance data, data cleaning and formatting processing;

[0057] Step S2, load the standardized naming mapping table of the power system equipment and parse;

[0058] Step S3, configure and call the large language model, perform feature extraction on the preprocessed text to be analyzed, and obtain the structured information of the target field;

[0059] Step S4, adopt vector embedding-based semantic similarity calculation for matching and searching the structured information of the target field in the standardized naming mapping table to obtain the corresponding preliminary matching result;

[0060] Step S5, further calculate the preliminary matching result according to the standardization algorithm, establish the standardized mapping index table of the current station and the equipment at each level in the station, and realize the standardization of data;

[0061] Step S6, integrate the standardized mapping index table and encapsulate it into the form of system service or offline file to output to the power grid business system.

[0062] Compared with the prior art, the present application has the following beneficial effects and advantages:

[0063] The present application realizes efficient and accurate structured extraction and standardization processing of power maintenance data through the above detailed technical steps and method design. The method can process complex unstructured text input, accurately correspond the extracted information with the standardized equipment and station name through semantic understanding and multi-level matching mechanism. This process greatly improves the automation degree and accuracy of data processing, and lays a solid foundation for subsequent data analysis and business application. BRIEF DESCRIPTION OF DRAWINGS

[0064] Fig. 1 is a system flow diagram of the present application;

[0065] Fig. 2 is a system structure diagram of the present application. DETAILED DESCRIPTION

[0066] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation method of the present application will be further described in detail below. In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the spirit of the present application, therefore the present application is not limited by the specific implementation disclosed below.

[0067] As shown in Figs. 1-2 , a structured extraction method of power maintenance data based on a large model, comprising:

[0068] 1. Data acquisition and preprocessing: Obtain raw power maintenance data from multiple sources, including but not limited to text records, Excel spreadsheets, databases, etc. The acquired data format is as follows:

[0069] B4XXXXX-8XXX-AXXX-BXXX-B5XXXXXXXXXX@@@XX.XX Plant, Unit 5, spot inspection, addressing a major vibration defect in Unit X, X-watt. @@@Shutdown.

[0070] 7DXXXXX-5XXX-9XXX-EXXX-CAXXXXXXXX@@@XXX Factory, Unit 1, standby@@@ is shut down.

[0071] 2. Standardized mapping table preparation: Establish and load the standardized naming mapping table for power system equipment to provide a basis for subsequent data standardization;

[0072] 3. Extraction from unstructured text: Configure and call a large language model to extract structured information of the target field from the preprocessed text to be analyzed;

[0073] 4. Semantic similarity calculation: Implement semantic similarity calculation based on vector embedding to extract the target field (power system equipment name) in the text to be extracted and match it with the standardized name (in the power system equipment standardized naming mapping table) to determine the standardized name content corresponding to the target field.

[0074] 5. Standardization of plant and equipment names: Match the extracted target fields (equipment and plant names) with the standardized mapping table one by one, establish a standardized mapping index table for the current plant and each level of equipment within the station, and achieve data standardization.

[0075] 6. Data integration and output: Integrate the standardized structured data (in the standardized mapping index table) and output it in the form of system services or offline files for use by the power grid business system.

[0076] The specific implementation steps of the technical solution are as follows:

[0077] Step 1: Data acquisition and preprocessing: obtain raw power maintenance data from multiple sources and perform preliminary cleaning and format unification.

[0078] Step 101: Design a secure file transfer mechanism to securely retrieve data files from a remote server. This secure file transfer mechanism uses the commonly used SFTP file transfer mechanism, a secure protocol for file transfer over the SSH protocol. It uses SSH's encryption capabilities to encrypt file transfers, preventing eavesdropping, tampering, or forgery during transmission, thereby protecting the confidentiality and integrity of files.

[0079] Step 102: Implement a local cache strategy for data files to improve processing efficiency and support offline operations.

[0080] Step 103: A general data parsing algorithm is used to process input data in various formats, such as delimited text data. For example, the data parsing algorithm needs to find and parse data in the following format: B4XXXXX-8XXX-AXXX-BXXX-B5XXXXXXXXXX@@@XX.XX Plant, Unit 5, Spot Inspection, Addressing a Large Vibration Defect in Unit A, B Sleeve. @@@ Shutdown.

[0081] You can first use regular expressions to filter out a number of target data that meet the conditions in the data set, and then split or search the target data according to the specified characters. Finally, you will get a set of target data that has been split:

[0082] B4XXXXX-8XXX-AXXX-BXXX-B5XXXXXXXXXX, XX.XX factory, No. 5 machine.

[0083] Step 104: Implement data cleaning functions, including removing irrelevant characters, unifying separators, and format standardization, such as processing XX#1 line into XX No. 1 line.

[0084] Step 2: Prepare the standardized mapping table, establish and load the standardized naming mapping table for power system equipment, and provide a basis for subsequent data standardization.

[0085] Step 201: Design a multi-level ID mapping structure, including an original mapping table, and decomposition processing into ID mapping data: plant station mapping table and equipment mapping table, to meet the needs of different scenarios. Among them, the original mapping table is an existing standardized naming mapping table for power system equipment (for example, data from different systems, and multiple standardized naming mapping tables exist for multiple systems), wherein each standardized naming mapping table stores all the standard naming names of relevant plants and equipment under the current system key value without hierarchy; the post-processing mapping operation is to decompose each entry of the original mapping table of the current plant station into a plant station mapping table and a device mapping table according to the plant station, equipment, etc. hierarchical levels. For example, the first level includes plant 1, plant 2...; the second level includes the power transmission and transformation equipment under each plant...; the third level includes the sensor equipment, switches and other actuators under each power transmission and transformation equipment...

[0086] For example, the original mapping table may be in the following form:

[0087] 99XXXXXXXXXXX@@@XX factory@@@1

[0088] 99XXXXXXXXXXX@ @ @ XX plant @ @ 99XXXXXXXXXX1@ @ 1st machine

[0089] Processed mapping table form:

[0090] {

[0091] "id":"XXXXXXXXX",

[0092] "station":{"confirm_name":"XX plant","confirm_id":"99XXXXXXXXXXX"},

[0093] "devices":[{confirm_name":1st machine","confirm_id":"99XXXXXXXXXX1"}]

[0094] }

[0095] Step 202: Develop a mapping data loading algorithm that can read ID mapping data from a text file.

[0096] Step 203: Implement special character processing functions in standardized names to ensure that standardized naming is consistent with the naming rules of the data to be analyzed.

[0097] Step 204: Design a unified mapping key value processing method to ensure that data from different sources can be consistently matched. Mapping key value is the classification identifier of different source data.

[0098] Step 3: Extract, configure and call large language models in unstructured text to extract structured information from the text to be analyzed.

[0099] Step 301: Design model initialization and preheating mechanism to ensure the model is in the best state before use, reduce large model inference time and improve response speed.

[0100] Step 302: Design a special instruction template to guide the large model to accurately understand and extract key information in power maintenance data. Including:

[0101] 1. Assign a role to the large model. For example: tell the large model the following instructions: you are now a power operation and maintenance engineer, and need to identify the equipment according to the text given to you according to the following requirements.

[0102] 2. Clearly state the task requirements: clearly state the key information (target field) that needs to be extracted from the text, such as device name, etc. Target fields include: XX transformer, XX substation, XX plant, XX power plant, XX wind farm, XX wind farm

[0103] 3. Output format requirements: Specify the JSON structure of the output, including the names and data types of each target field.

[0104] 4. Domain knowledge hints: Input common device types, maintenance terminology, and other key information in the power system to the model to help it understand domain-specific expressions.

[0105] 5. Example illustration: Provide examples of input text (target fields) and expected output in JSON format to help the model understand the task requirements. An example of structured information in JSON format is as follows: {station_name: "power plant or substation name", device: ["device name 1", "device name 2"]}.

[0106] Step 303: Implement the large model API calling interface, including request construction, sending, and response processing.

[0107] Step 304: Design a JSON data extraction algorithm to accurately identify and extract valid structured data from model output.

[0108] Step 4: Semantic similarity calculation, implement semantic similarity calculation based on vector embedding, used for matching the structured information of plant station and device name in the text to be extracted with the standardized names in the plant station mapping table and device mapping table after disassembly from different system sources.

[0109] Step 401: Define the vector representation of the generated text, specifically:

[0110] For a given text t, its vector representation is:

[0111] V(t) = f(t)

[0112] Where t is the structured information of plant station and device name in the text to be processed or the standardized name in the plant station mapping table and device mapping table after disassembly from different system sources, f is the encoding function, for example: model.encode(string).

[0113] Step 402: Design global and local similarity matching algorithms to support finding the most similar items in different ranges, specifically:

[0114] a) For query text q and text set T = {t1, t2,..., tn}, calculate the similarity of each item according to the following formula:

[0115]

[0116] Where q is the structured information of plant station and equipment name in the text to be processed, T is the set of all standardized names in the plant station mapping table and equipment mapping table after disassembly from different system sources, and · represents the dot product, and |V| represents the Euclidean norm of the vector.

[0117] b) An example can select the most similar item in set T as the matching result according to the following formula, and jump to step 6.

[0118] most_similar(q,T)=argmax t∈T (similarity(q,t))

[0119] Another example, in descending order of similarity, and take the first M as the candidate name set c={c1, c2,..., cn}, and then perform steps 403 to 504.

[0120] Step 403: Implement optimization strategies for vector calculation to improve the calculation efficiency when processing large-scale data. The optimization strategies used include: using efficient vector library functions; using parallel and distributed computing to improve calculation efficiency; optimizing memory resource allocation, etc.

[0121] Step 5: Standardization of plant station and equipment name, match the target field of extracted equipment and plant station name with the standardized mapping table to realize the standardization of data.

[0122] The standardization algorithm for plant station name processes various naming variants and abbreviations. It includes:

[0123] a) Edit distance calculation: for two strings s1 and s2, the edit distance D(i, j) is calculated by the following method:

[0124]

[0125] Where s1 is the structured information of plant station and equipment name in the text to be processed, and s2 is any one of the candidate names c={c1, c2,..., cn}; D(i, j) is the semantic similarity distance;

[0126] b) Standardization score calculation: for candidate name c and standard name s, the standardized score considering semantic similarity and edit distance can be expressed as:

[0127] score(c,s)=α·similarity(c,s)+(1-α)·(1-normalized_edit_distance(c,s))

[0128] where a is a weight parameter, similarity(c,s) is the result calculated according to the similarity formula for c and s, and normalized_edit_distance is the normalized edit distance D(i,j).

[0129] c) Best match selection: For a candidate name c and a set of standard names s, the selection of the best match can be represented as:

[0130] best_match(c,S) = argmax s∈S score(c,s)

[0131] Thus, the result of the second best matching method is obtained, and a unique mapping index table of the target field of all plant stations and equipment names and the standard naming is established.

[0132] Step 502: Call the algorithm in step 501 to process the standardized matching of equipment names, which can be flexibly processed according to different scenarios.

[0133] Step 503: Design a context-based mapping table filtering mechanism to narrow the range of equipment matching according to the plant station matching information, and improve the matching efficiency and accuracy.

[0134] Step 504: For some special cases, adjust the calculation weight of the edit distance and the standardized score to adapt to special naming conditions.

[0135] Step 6: Data integration and output, integrate the standardized structured data and output in the specified format.

[0136] Step 601: Design a data synthesis processing algorithm to integrate the information extracted from the large model and the standardized data.

[0137] Step 602: Implement a general JSON serialization method to ensure that all data types (including special numerical types) can be correctly converted to JSON format.

[0138] Step 603: Develop an incremental data processing and output mechanism to support efficient processing and real-time output of large-scale data.

[0139] Step 604: Implement data upload function to safely transmit the processing results back to the remote server for use by the maintenance system.

[0140] {"id":"XXXX-XXXX","station":{"name":"XX.XX Factory","processed_name":"XXXX Factory","confirm_name":"XXXX Factory","confirm_id":"9XXXXXX","similarity":0.8283858299255371},"devices":[{"name":"Unit X","processed_name":"Unit X","confirm_name":"Unit X of XXXX Factory","confirm_id":"9XXXXX6","similarity":0.5086394548416138},{"name":"Unit X","processed_name":"Unit X","confirm_name":"Unit XXXXXX","confirm_id":"9XXXXXXX6","similarity":0.6201760768890381}]}.

[0141] Key technical points:

[0142] 1. Large model application: Using large language models to extract information from unstructured text improves the ability to understand complex expressions.

[0143] 2. Design dedicated instruction templates to guide the large model to accurately understand and extract key information from power maintenance data.

[0144] 3. Semantic similarity calculation: Use pre-trained text embedding models to calculate semantic similarity and achieve high-precision text matching.

[0145] 4. Multi-level mapping mechanism: Through the multi-level mapping structure, the flexibility and accuracy of the standardization process are improved. In set theory, it is expressed as: M = {M_original, M_processed, M_station, M_device}, where each Mi is a mapping function, Mi: X→Y

[0146] 5. Context-aware matching: Considers site information during device name matching to improve matching accuracy.

[0147] 6. Special scenario processing: Implement customized processing logic for specific scenarios. This can be expressed as conditional probability: P(correct_match|special_scenario)>P(correct_match|general_scenario)

[0148] 7. Incremental processing: support efficient processing of large scale data by processing and outputting on the fly.

[0149] It should be noted that the present application is not limited to the embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Any modifications made by those skilled in the art based on the teachings of the present application without departing from the spirit and scope of the present application and the scope of protection of the claims are intended to be within the scope of the present application.

Claims

1. A large model-based structured extraction system for power maintenance data, characterized by, A system is constructed as follows for multi-source complex heterogeneous text, through semantic understanding and multi-level matching mechanism, the extracted information is accurately corresponding to the standardized device and station name, and is provided to the maintenance user, the system comprises: A data acquisition preprocessing module collects original power maintenance data, data cleaning and formatting processing; A standardized mapping table preparation module loads the standardized naming mapping table of the power system device and parses it; An unstructured text feature extraction module configures and calls a large language model to extract features from the preprocessed text to be analyzed, and obtains structured information of the target field; A semantic similarity module uses vector embedding-based semantic similarity calculation to match and search the structured information of the target field in the standardized naming mapping table to obtain the preliminary matching result; A station and device name standardization module further calculates the preliminary matching result according to the standardization algorithm, establishes the standardized mapping index table of the current station and each level of equipment in the station, and realizes the standardization of the data; A data integration and output module integrates the standardized mapping index table and encapsulates it into a system service or an offline file to output to the power grid business system.

2. The method of claim 1, wherein, The power maintenance data is multi-source heterogeneous text records, Excel tables, and database data; The data acquisition preprocessing module includes: A secure transmission unit uses a secure file transfer protocol to obtain data files from a remote server; for preventing eavesdropping, tampering or forgery during transmission; A local cache unit for local caching of data files; A data parsing unit parses multi-source heterogeneous data into a unified data format; including: using regular expressions to filter a number of target data that meet the conditions in the data set, and splitting or finding the target data according to the specified characters to obtain a group of split target data; A data cleaning unit for data cleaning, including removing irrelevant characters, unifying delimiters, and standardizing formats.

3. The system according to claim 1, wherein, The standardized mapping table preparation module includes: A multi-level mapping structure unit for collecting original mapping tables and parsing them into corresponding level ID mapping data to be tabulated as station mapping tables and device mapping tables; for adapting to the needs of different scenarios; wherein the original mapping table is an existing standardized naming mapping table of the power system device; wherein each standardized naming mapping table stores all the standard naming names of the current system key value without hierarchical storage; the parsing operation is to parse each entry of the original mapping table of the current station into a station mapping table and a device mapping table according to the level; A mapping data loading unit for reading ID mapping data from a text file; A rule unit for defining that the standardized names in the ID mapping data are consistent with the naming rules of the data to be analyzed; and matching the mapping key values of the standardized names in the ID mapping data; wherein the mapping key value is a classification identifier of different source data, and the naming rule consistency is used to match different source data.

4. The system according to claim 1, wherein, The unstructured text feature extraction module includes: The large language model unit is configured to adopt a large language model, and iteratively train input data in combination with the parameter unit to obtain an ideal model, so as to realize extraction of structured information of a target field; the target field includes XX transformer, XX substation, XX power plant, XX power plant, XX wind farm and XX wind power plant; The parameter unit is configured to define roles, tasks, instruction templates, output data structures and domain knowledge of the large model; the tasks, instruction templates, output data structures and domain knowledge are different under different roles; the domain knowledge inputs common device types and maintenance terms of a power system to the model, and helps the model learn field-specific expressions.

5. The system according to claim 1, wherein, The semantic similarity module includes: The semantic similarity calculation unit is configured to realize semantic similarity calculation based on vector embedding, and is used for matching and searching of standardized names in a factory station mapping table and a device mapping table parsed under different system sources and structured information of factory stations and device names in to-be-extracted text; The text vector representation unit is configured to represent a given text t as V(t)=f(t); wherein t is structured information of factory stations and device names in to-be-processed text or standardized names in a factory station mapping table and a device mapping table parsed under different system sources, and f is an encoding function; The similarity calculation unit is configured to calculate similarity, and find and match the most similar items in different ranges.

6. The system according to claim 5, wherein, The calculation of similarity includes: for a query text q and a text set T={t1, t2,..., tn}, the similarity of each item is calculated according to the following formula: Wherein the query text q is structured information of factory stations and device names in to-be-processed text, and the text set T is a large set of each standardized name in a factory station mapping table and a device mapping table parsed under different system sources, wherein · represents dot product, and |V| represents Euclidean norm of a vector.

7. The large model-based power maintenance data structured extraction system of claim 5, characterized in that The finding and matching of the most similar items in different ranges include: The most similar item in the set T is selected as a matching result according to the following formula: Compute most_similar(q, T) = argmax t∈T (similarity(q, t)); Alternatively, the first M are taken as a candidate name set c={c1, c2,..., cn} as a matching result output in descending order of similarity values.

8. The large model-based power maintenance data structured extraction system of claim 5, characterized in that The standardization algorithm of the factory station and device name standardization module is used for processing various naming variants and abbreviations; and includes: a) Edit distance calculation: for two strings s1 and s2, the edit distance D(i, j) is calculated: Wherein the string s1 is structured information of factory stations and device names in to-be-processed text, and s2 is any one of candidate names c={c1, c2,..., cn}; D(i, j) is a semantic similarity distance; b) For a candidate name c and a standard name s, the semantic similarity and the edit distance are considered comprehensively to calculate a standardization score: score(c, s)=α·similarity(c, s)+(1-α)·(1-normalized_edit_distance(c, s)) Wherein, a is a weight parameter, similarity(c,s) is the result calculated according to the similarity formula for c and s, and normalized_edit_distance is the normalized edit distance D(i,j); c) The selection of the best match for the candidate name c and the standard name set s is represented as: best match(c, S) = argmax s∈S score(c, s) At this point, the unique standardized mapping index table of the target field and the standard naming of all power stations and equipment names is established.

9. The system according to claim 1, wherein, The data integration and output module adopts the JSON serialization method to format the data in the standardized mapping index table; adopts the incremental data processing and output mechanism to support efficient processing and real-time output of large-scale data; and is used to realize the data uploading function, safely transmit the processing result back to the remote server, and provide the repair system.

10. A method for structured extraction of power maintenance data based on large models, characterized in that, The following steps are performed to identify the multi-source complex heterogeneous power repair data, match and search in the standardized mapping table to obtain the unique standard naming and provide it to the repair user, and the method comprises the following steps: Step S1, collecting original power repair data, data cleaning and formatting processing; Step S2, loading the standardized naming mapping table of the power system equipment and analyzing; Step S3, configuring and calling the large language model, extracting features from the preprocessed text to be analyzed, and obtaining the structured information of the target field; Step S4, using the vector embedding-based semantic similarity calculation to match and search the structured information of the target field in the standardized naming mapping table to obtain the corresponding preliminary matching result; Step S5, further calculating the preliminary matching result according to the standardization algorithm, establishing the standardized mapping index table of the current power station and each level of equipment in the station, and realizing the standardization of the data; Step S6, integrating the standardized mapping index table and packaging it into the form of system service or offline file to output to the power grid business system.

Images