Power data analysis method and system

Abstract

The invention relates to the technical field of electric power data analysis, in particular to an electric power data analysis method and system. The method comprises a database building step, a data acquisition step, a data preprocessing step, a first extraction step, a second extraction step, a data reprocessing step, a deviation translation step, a similarity calculation step, a similarity judgment step, a replacement step, a correction step, a knowledge ontology construction step, a real-time monitoring step and a result output step. The system comprises a database, a data acquisition module, a data preprocessing module, a first extraction module, a second extraction module, a data reprocessing module, a deviation translation module, a similarity calculation module, a similarity judgment module, a replacement module, a correction module, a knowledge ontology construction module, a real-time monitoring module and a result output module. The method has the effect of improving the accuracy of power data analysis.

Description

Technical Field

[0001] The present application relates to the technical field of power data analysis, and in particular to a power data analysis method and system. Background Art

[0002] At present, in the power system, real-time monitoring and analysis of power data is of great significance for ensuring the stable operation of the power grid and preventing faults. With the rapid development of technologies such as big data, cloud computing, the Internet of Things and artificial intelligence, the power system is facing new opportunities for intelligent upgrading. The application of big data technology in the power system can realize the collection, storage, processing and analysis of massive data, providing strong support for the operation and management of the power system. As an important branch of artificial intelligence, machine learning can automatically extract features and build prediction models through training and learning of historical data, so as to realize real-time monitoring and early warning of the operation status of the power system.

[0003] However, the existing corpora and word embeddings are mostly oriented towards general fields. Due to the gap in text structure and professional vocabulary, some words in the power field may have completely different meanings from words in the general field. Word embeddings oriented towards general fields often cannot meet the parsing needs of the power field, resulting in the inability to accurately parse power data information. Summary of the invention

[0004] In order to improve the accuracy of power data analysis, the present application provides a power data analysis method and system.

[0005] In the first aspect, the present application provides a method for analyzing power data, which adopts the following technical solution: A method for analyzing power data comprises the following steps: Build database: Establish a power system database to store historical power data, fault information and professional related data; Data collection: Collect historical power data and fault information in the power system, collect power system and power equipment information and store it as professional related data. The power equipment information includes the name of the power equipment and the product name data, which can be obtained by the relevant information of the power equipment and power dictionary pages in the professional comprehensive portal website; collect power analysis reports and accident reports and store them as analysis data; Data preprocessing: formatting and standardizing historical power data, fault information and professional related data; First extraction: extract the knowledge related to the operation law and fault characteristics of the power system from the preprocessed data, and record it as the first data; Second extraction: combining the domain professional terms and the first data, segmenting the analysis data to form a first vocabulary, and establishing vocabulary relationships based on the sentence semantics; then combining the first vocabulary with the vocabulary relationship and storing them as the second data; Data reprocessing: filtering the second data; Biased translation: Obtain the professional field information of the analyst, and convert the first word in the second data into professional words or similar written words with a field bias to form a translated word set; Similarity calculation: performing similarity calculation on the words in the translated vocabulary set and the professional words in the first data to obtain a similarity set; Similarity judgment: judge whether the similarity is greater than the set limit value, if so, execute the replacement step, otherwise, execute the correction step; Replacement: replace the translated words according to the first data, and store them as third data in combination with the word relationship; Correction: call out the original sentence of the translated vocabulary and correct the first vocabulary according to the semantics; Knowledge ontology construction: Build a knowledge ontology analysis model based on the first data, third data, power system, historical power data and fault information; Real-time monitoring: obtain power data in real time to form query needs; Output results: According to the query terms in the query requirements, they are input into the knowledge ontology analysis model, the word relationships are retrieved, and the final results are generated.

[0006] By adopting the above technical scheme, the existing technology often focuses on a single processing link of power data, such as data cleaning, feature extraction or fault detection, etc., while the present technical scheme integrates multiple processing links to form a complete power data analysis process, which can process and analyze power data more comprehensively; by setting a domain vocabulary translation step, it can convert common vocabulary into professional vocabulary or synonymous written vocabulary according to the professional field information of the analyst, thereby improving the accuracy and professionalism of the vocabulary. At the same time, through the similarity calculation step, the consistency of the translated vocabulary with the professional vocabulary in the first data is further ensured; and through the intervention of the professional field, the formed knowledge ontology analysis model is targeted, which is convenient for improving the accuracy of power data analysis.

[0007] By constructing a knowledge ontology analysis model, the systematic representation and efficient utilization of power system knowledge are achieved. At the same time, the real-time monitoring step can obtain power data in real time and form query requirements, providing strong support for the dynamic analysis and early warning of power data. According to the query terms in the query requirements, the vocabulary relationships are retrieved to generate results with intelligent and personalized characteristics, which improves the readability and practicality of the results.

[0008] In general, through comprehensive data processing capabilities and domain vocabulary translation and similarity calculation steps, this technical solution can more accurately parse power data and improve the accuracy and efficiency of data parsing; through knowledge ontology construction and real-time monitoring steps, this technical solution can monitor the operating status of the power system in real time, promptly discover and warn of potential faults, thereby enhancing the reliability and safety of the power system; through the processing and analysis of this technical solution, power data can be more effectively utilized and shared, providing richer and more accurate data support for scientific research, engineering design, and operation management in the power industry.

[0009] Optionally, in the knowledge ontology construction step, it also includes determining the propagation path according to the power system construction and generating the SDG directed acyclic graph.

[0010] By adopting the above technical solutions, SDG intuitively displays the knowledge structure and propagation path in the power system in the form of nodes and edges. The nodes and edges in SDG not only represent the association between knowledge, but also imply the propagation path and reasoning rules of knowledge. By analyzing the nodes and edges in SDG, the reasoning and fault location of knowledge in the power system can be realized. When a power system fails, the cause and propagation path of the fault can be quickly located by analyzing the relationship between the fault node and its adjacent nodes in SDG, so as to take corresponding countermeasures.

[0011] Optionally, a repeat determination step is further provided between the second extraction step and the data reprocessing step; Repeat judgment: including character distance judgment step, relationship extraction step, vocabulary repeat judgment step, deletion step, spacing word formation step, combination step and word meaning judgment step; Character distance judgment: obtain the character distance between two adjacent first words, and judge whether it is less than or equal to T. If so, perform word duplication judgment, otherwise, perform the relationship extraction step; Relation extraction: determine whether there is additional information between two adjacent first words. If so, extract the relationship attribute from the additional information and store it in the second data, and then perform data reprocessing steps; Word duplication judgment: judge whether two adjacent first words are the same, if so, execute the deletion step; otherwise, execute the combination step; Deletion: Delete one of the two adjacent first words, and then perform the interval word formation step; Interval word formation: the remaining first word is combined with the interval characters between the previous and next first words to form a new first word, and then the word meaning judgment step is performed; Combining: combining two adjacent first words to form a new first word, and then executing the word meaning judgment step; Word meaning judgment: judge whether the meaning of the first word is complete. If so, form the first word, optimize the second data, and then execute the data reprocessing step. Otherwise, execute the correction step.

[0012] By adopting the above technical solution, by judging the character distance between two adjacent first words, words that are too close can be effectively identified, and then more in-depth word duplication judgment or relationship extraction can be performed. This helps to avoid data redundancy and improve data accuracy and compactness.

[0013] For identical adjacent words, directly deleting them can significantly reduce the duplication of data and improve the efficiency of data processing; for non-identical adjacent words, forming new words by splicing or spacing them can help capture more complex semantic relationships and reduce data fragmentation; when there is additional information between adjacent words, by extracting relationship attributes and storing them in the second data, the semantic relationship between the first words can be enriched, providing more valuable information for subsequent data analysis and processing; by judging the completeness of the newly formed first words, the semantic accuracy and completeness of the first words can be ensured to avoid ambiguity and misunderstanding; when the word meaning judgment is incomplete, executing the correction step can flexibly adjust the first word to ensure the accuracy and availability of the data. This flexibility helps to cope with the complex and changing power data environment.

[0014] After the word meaning judgment is complete and a new first vocabulary is formed, optimizing the second data can further improve the quality and value of the data and provide stronger support for subsequent data analysis and application; most of the sub-steps in the repeated judgment steps can be implemented through automated algorithms, which helps to reduce manual intervention and improve the level of intelligence in data analysis.

[0015] Optionally, an invalidity judgment step is further provided between the character distance judgment step and the relationship extraction step; Invalidity judgment: judge whether there are continuous repeated strings in the first word. If yes, execute the word optimization step. Otherwise, execute the relationship extraction step. Optimize vocabulary: filter the string, optimize the first vocabulary, and then perform the word meaning determination step.

[0016] By adopting the above technical solution, the invalid judgment step can accurately identify the redundant part in the data by checking whether there are consecutive repeated character strings in the first vocabulary; this redundancy may be caused by data entry errors, mistakes in copying and pasting, or incomplete data cleaning; by filtering out these consecutive repeated character strings, the quality of the data can be significantly improved; after identifying the consecutive repeated character strings, the vocabulary optimization step will filter them and adjust the first vocabulary accordingly; this step not only removes redundant information, but also makes the first vocabulary more concise and clear, which helps to improve the efficiency and accuracy of data processing; through the invalid judgment and vocabulary optimization steps, the redundant part in the data can be significantly reduced, thereby reducing the complexity of data processing; this helps to speed up data processing and improve overall efficiency.

[0017] If there are continuous repeated character strings in the data and they are not processed in time, it may cause errors or misleading in subsequent steps such as relationship extraction and word meaning judgment. By adding an invalid judgment step, these problems can be discovered and resolved at an early stage, thus avoiding potential risks in subsequent processing.

[0018] By optimizing the vocabulary step, we can ensure that the first vocabulary in the data remains consistent in format and content; this consistency helps improve the reliability of data analysis, making the analysis results more accurate and reliable; and helps support diverse data analysis needs, such as trend analysis, association rule mining, etc.; the optimized data is clearer and more intuitive when presented in a visual manner, which helps users quickly understand the information and rules behind the data, which helps to improve the breadth of data applications and user experience.

[0019] Optionally, an abnormality determination step is further provided after the result output step; Abnormal determination: including: positioning step, first acquisition step and node replacement step; Positioning: The abnormal data location is recorded as an abnormal node, and the breakpoints before and after the abnormal node are obtained, recorded as the i-th breakpoint and the i+1-th breakpoint; First acquisition: obtain the replacement branch between the i-th breakpoint and the i+1-th breakpoint, recorded as the repair node; Node replacement: Replace abnormal nodes with repair nodes.

[0020] By adopting the above technical solution, through the positioning step, the location of abnormal data in the power data can be accurately identified and marked as an abnormal node. This step is the basis for subsequent abnormal processing and ensures that the abnormal data can be accurately identified and processed.

[0021] While locating the abnormal node, the breakpoints before and after it, namely the i-th breakpoint and the i+1-th breakpoint, are also obtained; this breakpoint information is crucial for the subsequent acquisition of replacement branches (repair nodes), which helps to ensure the accuracy and effectiveness of the replacement operation.

[0022] In the first acquisition step, based on the previous and next breakpoint information, the replacement branch between the i-th breakpoint and the i+1-th breakpoint, that is, the repair node, is efficiently acquired. This step is a prerequisite for the node replacement operation and ensures the feasibility of the replacement operation. By accurately acquiring the replacement branch, it can be ensured that the replacement operation will not introduce new anomalies or errors. At the same time, the accuracy of the replacement branch also helps to improve the quality and reliability of the data after repair.

[0023] In the node replacement step, the abnormal nodes are replaced with repair nodes. This step is the core of exception handling. The replacement operation can eliminate the impact of abnormal data on the overall data quality.

[0024] By replacing abnormal nodes, the quality of power data can be significantly improved, and the normal operation of the power system can be maintained, which is convenient for maintenance during working hours of staff; the repaired data is more accurate, complete and reliable, providing strong support for subsequent data analysis and application.

[0025] Optionally, after the node replacement step, a result re-search step is further provided; Result re-search: Get the power data between the i-th breakpoint and the i+1-th breakpoint, form the query demand again, and then perform the verification and judgment steps; Verification and judgment: judge whether the abnormal data is eliminated. If so, execute the fault location step; otherwise, execute the output result step; Fault location: Determine whether the abnormal node between the i-th breakpoint and the i+1-th breakpoint is indeed abnormal, and log it in the database, and then execute the output result step again.

[0026] By adopting the above technical solution, by re-acquiring the power data between the i-th breakpoint and the i+1-th breakpoint and forming a query demand again, it is possible to ensure a comprehensive check and verification of the replaced data. This step helps to discover new problems or omitted abnormal data that may be introduced by the replacement operation.

[0027] By determining whether the abnormal data has been eliminated, the effectiveness of the replacement operation can be verified; if the abnormal data has been successfully eliminated, it indicates that the replacement operation is successful; if the abnormal data still exists, further processing is required.

[0028] When it is determined that the abnormal data still exists, the fault location step can accurately determine the abnormal node between the i-th breakpoint and the i+1-th breakpoint and confirm that there is indeed an abnormality; this step helps to quickly locate the source of the problem and provide strong support for subsequent processing.

[0029] The results of fault location are recorded in the database as logs, which can facilitate subsequent data analysis and troubleshooting. At the same time, log records can also be used as a basis for tracing data repair and exception handling processes, ensuring the transparency and traceability of the processing process.

[0030] From the location and replacement of abnormal data to the re-search and verification of results, and then to fault location and log recording, a complete exception handling closed loop is formed. This closed-loop process helps ensure that abnormal data is processed comprehensively and accurately, and enhances the accuracy of data analysis and processing.

[0031] By recording and analyzing the logs and data during the exception handling process, we can discover the problems and deficiencies in the data parsing and processing process, and provide strong support for subsequent optimization and improvement; through the result re-search and verification judgment steps, we can ensure that the quality of the repaired data is significantly improved; high-quality data can provide strong support for subsequent data analysis and application, and improve the stability and reliability of data application.

[0032] During the exception handling process, potential security risks can be discovered and handled in a timely manner through accurate fault location and log recording. This helps ensure the security and confidentiality of power data and prevent data leakage or malicious tampering.

[0033] Optionally, a survey step is provided between the verification and judgment step and the result output step; Investigation: According to the SDG directed acyclic graph, determine the predecessor associated nodes of the abnormal node, and then perform the replacement steps one by one; Replace one by one: Get the previous and next breakpoints of the previous associated node, get the repair node, replace them one by one, and execute the result re-search step.

[0034] By adopting the above technical solution, the investigation step determines the predecessor associated node of the abnormal node through SDG (directed acyclic graph); and SDG, as a graphical tool for representing the dependency relationship between system elements, can clearly display the logical connection between the abnormal node and its predecessor node. This step helps to deeply analyze the cause of the abnormal data and provide accurate target positioning for subsequent one-by-one replacement.

[0035] By investigating the front-related nodes, we can further explore the root causes of abnormal data, thereby avoiding only dealing with surface phenomena and ignoring the fundamental problems; this helps to improve the depth and accuracy of exception processing and ensure the long-term effectiveness of data repair.

[0036] After replacing each value one by one, the result re-search step can verify the effectiveness of the replacement operation. This step not only ensures the accuracy of the data, but also improves the efficiency and quality of exception handling through continuous iteration and optimization.

[0037] By investigating and replacing the front-end associated nodes one by one, the robustness of the entire data analysis and processing process can be ensured, which helps to avoid abnormal data from triggering a chain reaction in subsequent processing, leading to more problems; the steps of investigation, one-by-one replacement, and result re-search together constitute a complete abnormality processing closed loop; this closed-loop process not only helps to ensure that abnormal data is processed comprehensively and accurately, but also improves the robustness of data analysis and processing through continuous iteration and optimization; it can ensure that the quality of repaired data is significantly improved. High-quality data can provide strong support for subsequent data analysis and application, and improve the reliability and accuracy of data application; by optimizing the abnormality handling process, it can ensure that data plays a greater role in the decision-making process and improve the intelligence level and operational efficiency of the power system.

[0038] Optionally, after the real-time monitoring step, a demand supplement step is also provided; Supplementary requirements: including: image data acquisition steps, data division steps and real-time extraction steps; Image data acquisition: acquiring real-time images at the location required for query and forming image data; Data division: Divide the image data into multiple image blocks with different precisions; Real-time extraction: extract the relevant information in the image block, enrich the query requirements, then input it into the knowledge ontology analysis model, and execute the output result step.

[0039] By adopting the above technical solution, by obtaining real-time images at the location where the query is required, the actual situation at the location can be intuitively understood. These image data provide important visual references for the analysis of power data, which helps to more accurately understand the meaning and background of the data.

[0040] Dividing the image data into multiple image blocks with different precision and extracting the related information therein can further enrich the content of the query requirements. This processing method not only improves the accuracy and details of the data, but also enhances the correlation between the data, making the power data more complete and accurate.

[0041] By associating power data with image information, the spatial positioning of power data can be achieved. This association method helps to quickly identify the location of abnormal data and provide strong support for subsequent processing.

[0042] By combining image information and power data, abnormal data can be located more accurately and the causes of their occurrence can be identified. This comprehensive processing method not only improves the accuracy of abnormality identification, but also shortens the time for abnormality processing and improves overall efficiency.

[0043] The combination of real-time monitoring steps and demand supplement steps makes the analysis and processing of power data more intelligent; by combining the data obtained through real-time monitoring with image data, comprehensive monitoring and accurate analysis of the power system can be achieved.

[0044] By acquiring and dividing image data, data visualization can be achieved. This display method not only improves the readability and understandability of data, but also provides more possibilities for data application.

[0045] By combining image information and power data, more flexible and diverse data applications can be developed; these applications not only meet the power system's diverse needs for data, but also enhance the application value and practicality of the data.

[0046] Optionally, an abnormality determination step is further provided after the live extraction step; Abnormality judgment: For each image block, determine whether there is surface abnormality information. If so, determine the fault type based on the output result and the surface abnormality information, and execute the output result step.

[0047] By adopting the above technical solution, the abnormality judgment step can accurately identify whether there is surface abnormality information in the image by carefully analyzing each image block; this judgment method not only improves the accuracy of abnormality detection, but also avoids false alarms or omissions caused by ignoring details.

[0048] Once surface abnormality information is detected, the abnormality judgment step will conduct a comprehensive analysis in combination with the output results to accurately determine the type of fault; this determination method helps to formulate targeted follow-up treatment measures and improves the efficiency and accuracy of fault handling; combined with the output results and surface abnormality information, the abnormality judgment step can quickly locate the location of the fault. This positioning method not only shortens the time for fault finding, but also reduces the risk of power outages caused by faults; accurate fault location provides strong support for subsequent processing; maintenance personnel can quickly formulate and implement maintenance plans based on the fault type and location information, thereby improving the efficiency and quality of fault handling.

[0049] The addition of the abnormality judgment step provides more possibilities for the application of data; through the analysis and processing of image data, more diversified data applications can be developed to meet the different needs of the power system; through accurate anomaly detection and fault location, potential problems in the power system can be discovered and dealt with in a timely manner, thereby avoiding or reducing losses caused by faults; this processing method not only improves the practicality of the data, but also enhances the value of the data.

[0050] In a second aspect, the present application provides a power data analysis system, which adopts the following technical solutions: A power data analysis system includes the following modules: Database: used to store historical power data, fault information, professional related data and real-time power data; Data acquisition module: The output end is connected to the input end of the database, and is used to collect historical power data and fault information in the power system, collect power system and power equipment information and store them as professional related data; collect power analysis reports and accident reports, store them as analysis data, and store them in the database; Data preprocessing module: The input end is connected to the output end of the database and is used to format and standardize historical power data, fault information and professional related data; The first extraction module: the input end is connected to the output end of the data preprocessing module, and is used to extract information from the preprocessed data; The second extraction module: the input end is connected to the output end of the first extraction module, and is used to segment the analysis data according to the data extracted by the first extraction module to form second data; Data reprocessing module: the input end is connected to the output end of the second extraction module, and is used to filter the second data; Biased translation module: the input end is connected to the output end of the data reprocessing module, and is used to perform field bias conversion on the first word in the second data according to the professional field information of the analyst; Similarity calculation module: the input end is connected to the output end of the biased translation module, and is used to calculate the similarity between the translated vocabulary and the professional vocabulary in the first data; Similarity judgment module: the input end is connected to the output end of the similarity calculation module, and is used to judge whether the similarity is greater than the set limit value; A replacement module: an input end is connected to an output end of the similarity judgment module, and is used to replace the translated vocabulary according to the first data when the similarity is greater than a set limit value; Correction module: the input end is connected to the output end of the similarity judgment module, and is used to call out the original sentence of the translated vocabulary when the similarity is less than or equal to the set limit value, and correct the first vocabulary according to the semantics; Knowledge ontology construction: The input end is connected with the output end of the correction module, the output end of the replacement module and the output end of the database to build a knowledge ontology analysis model; Real-time monitoring module: The output end is connected to the input end of the knowledge ontology construction module to obtain power data in real time and form query requirements; Output result: The input end is connected to the output end of the knowledge ontology construction module, and is used to retrieve the vocabulary relationship according to the query vocabulary in the query requirement to generate the final result.

[0051] By adopting the above technical solutions, the system realizes the integration and standardization of power data through the data acquisition and preprocessing modules, thereby improving the consistency and accuracy of the data; through the first extraction module and the second extraction module, the system can intelligently extract and segment the power data; the biased translation module can perform field-biased conversion of vocabulary according to the professional field information of the analyst, thereby improving the accuracy and professionalism of the vocabulary.

[0052] The similarity calculation module and judgment module can quickly and accurately calculate the similarity between words and make judgments based on the set boundary values, which helps the system to quickly locate related words and improve the efficiency of data parsing; the system can flexibly select word replacement or correction strategies based on the similarity judgment results, which ensures the accuracy and reliability of the final results.

[0053] Through the knowledge ontology construction module, the system can build a knowledge ontology analysis model for power data; the real-time monitoring module can obtain power data in real time, form query requirements, and combine it with the knowledge ontology analysis model to generate the final result; this enables the system to achieve real-time monitoring and rapid response to power data; overall, the system improves the analysis and application capabilities of power data through intelligent data analysis and processing processes; this helps the power system to achieve more accurate data analysis and decision support, and improve the operating efficiency and safety of the power system.

[0054] In summary, the present application includes at least one of the following beneficial technical effects: 1. By setting the domain vocabulary translation step, common vocabulary can be converted into professional vocabulary or synonymous written vocabulary according to the professional domain information of the analyst, thus improving the accuracy and professionalism of the vocabulary; 2. Through the analysis and processing of image data, more diversified data applications can be developed to meet the different needs of the power system; through accurate anomaly detection and fault location, potential problems in the power system can be discovered and handled in a timely manner, thereby avoiding or reducing losses caused by faults; this processing method not only improves the practicality of data, but also enhances the value of data. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 is a flow chart of the power data parsing method in Example 1 of the present application; Figure 2 is a detailed flow chart of the repeated determination step in Example 2 of the present application; Figure 3 It is a flowchart of the invalidity judgment step and the vocabulary optimization step in Example 2 of the present application; Figure 4 is a flowchart of the abnormality determination step in Example 2 of the present application; Figure 5 It is a flow chart of the requirement supplementation step in Example 2 of the present application. DETAILED DESCRIPTION

[0056] The following combination Figures 1 to 5 This application is described in further detail.

[0057] This embodiment discloses a power data analysis method.

[0058] Example 1: Reference Figure 1 , the power data parsing method comprises the following steps: Build database: Establish a power system database to store historical power data, fault information and professional related data; Data collection: Collect historical power data and fault information in the power system, collect power system and power equipment information and store it as professional related data. The power equipment information includes the name of the power equipment and the product name data, which can be obtained by the relevant information of the power equipment and power dictionary pages in the professional comprehensive portal website; collect power analysis reports and accident reports and store them as analysis data; Specifically, a database management system (such as MySQL, Oracle, etc.) is used to establish a power system database, and data collection tools (such as ETL tools) are used to collect historical power data and fault information from the power system; power equipment information and power dictionary page information are captured from professional comprehensive portals through web crawlers or API interfaces; power analysis reports and accident reports are collected, and appropriately sorted and classified, and the above data is stored in the database.

[0059] Data preprocessing: formatting and standardizing historical power data, fault information and professional related data; Specifically, the collected data often have problems such as noise, outliers, missing values, etc., and need to be preprocessed to improve data quality; the preprocessing steps include data cleaning (such as removing duplicate data, filling missing values), data transformation (such as normalization, standardization), data dimensionality reduction (such as principal component analysis), etc., deleting illegal characters in the corpus, such as spaces, line breaks, letters, punctuation marks, and stop words; and then performing word segmentation.

[0060] First extraction: extract the knowledge related to the operation law and fault characteristics of the power system from the preprocessed data, and record it as the first data; Second extraction: combining the domain professional terms and the first data, segmenting the analysis data to form a first vocabulary, and establishing vocabulary relationships based on the sentence semantics; then combining the first vocabulary with the vocabulary relationship and storing them as the second data; Data reprocessing: filtering the second data; Specifically, data mining or machine learning algorithms are used to extract knowledge related to the operation rules and fault characteristics of the power system from the preprocessed data; the extracted knowledge is stored in a structured manner to form the first data; the analysis data is segmented in combination with the domain professional terms and the first data, and the parts of speech are divided into noun, verb, and noun respectively. Considering that the corpus contains many domain professional terms, this section adopts the Jieba segmentation tool, and the first data and domain professional terms are used as a custom vocabulary for word segmentation to reduce word segmentation errors and accurately divide the text into power domain words; natural language processing technology (such as dependency syntactic analysis) is used to establish semantic relationships between words; the word segmentation results and lexical relationships are stored in an appropriate manner to form the second data, and then the second data is further filtered and screened to remove irrelevant or redundant information.

[0061] Biased translation: Obtain the professional field information of the analyst, and convert the first word in the second data into professional words or similar written words with a field bias to form a translated word set; Specifically, the descriptive language of experts may tend to be colloquial or professional. For colloquial descriptive language or descriptions using non-professional terms, the above-mentioned words are translated into professional words or synonymous written words that are biased towards the current field based on the field information of the experts to which the words originate, thereby forming a translated vocabulary set. Through biased translation, translation errors are reduced, making the translated vocabulary closer to what the experts want to express.

[0062] Similarity calculation: performing similarity calculation on the words in the translated vocabulary set and the professional words in the first data to obtain a similarity set; Similarity judgment: judge whether the similarity is greater than the set limit value, if so, execute the replacement step, otherwise, execute the correction step; Specifically, a text similarity algorithm (such as cosine similarity, Jaccard similarity, edit distance, longest common subsequence, etc.) is used to calculate the similarity between the translated vocabulary set and the professional vocabulary in the first data to obtain a similarity set, and a threshold value is set to determine whether the similarity is high enough to be replaced, wherein the threshold value is set to a constant value, or is determined based on the characteristics of the data set, domain knowledge, and historical data; if the similarity is greater than the threshold value, the professional vocabulary in the first data is used to replace the translated vocabulary; if the similarity is not greater than the threshold value, the original sentence of the translated vocabulary is called out and modified according to the semantics.

[0063] For example, to calculate the similarity, the algorithm formula for calculating the similarity of text content is as follows: ; Where ED is the professional vocabulary text in the first data and translated vocabulary text The edit distance between names, is the length of the professional vocabulary text in the first data, The length of the translated word vocabulary text.

[0064] Replacement: replace the translated words according to the first data, and store them as third data in combination with the word relationship; Correction: call out the original sentence of the translated vocabulary and correct the first vocabulary according to the semantics; Knowledge ontology construction: Based on the first data, third data, power system, historical power data and fault information, a knowledge ontology analysis model is built, and according to the power system construction, the propagation path is determined and the SDG directed acyclic graph is generated; Specifically, according to the first data, the third data, the power system structure, the historical power data and the fault information, the knowledge in the field of the power system is acquired, and related concepts such as generators, transformers, and lines are defined; attributes are defined for each concept, such as the power of the generator, the rated voltage of the transformer, etc.; the relationship between the concepts is clarified, such as the connection relationship between the generator and the line, the power supply relationship between the transformer and the load, etc., and the constraints and rules between the concepts are defined, such as the output power of the generator cannot exceed its rated power, etc., the propagation path and relationship of knowledge in the power system are determined, and the association and transmission relationship between knowledge are represented by graph theory methods (such as directed acyclic graph SDG); the defined concepts, attributes, relationships and constraints are encoded using ontology languages ​​(such as OWL, RDF, etc.), a knowledge ontology analysis model is built, and the encoded knowledge ontology analysis model is stored in a database.

[0065] Real-time monitoring: obtain power data in real time to form query needs; Output results: According to the query terms in the query requirements, they are input into the knowledge ontology analysis model, the word relationships are retrieved, and the final results are generated.

[0066] Specifically, various sensors, such as current sensors, voltage sensors, temperature sensors, etc., are used to collect various data of the power system in real time; the data acquisition equipment is responsible for collecting the data from these sensors and converting them into digital signals that can be used for analysis and processing; the collected data is transmitted to the data processing center via wired or wireless means, and then a query requirement is formed, and then the query requirement is input into the knowledge ontology analysis model, and relevant words are matched according to the query requirement, and then the word relationship is retrieved to generate the final result.

[0067] This embodiment obtains professional vocabulary through historical power data and power system information, and then biases the translation of expert language based on the network analysis report and the professional vocabulary, and then builds a knowledge ontology analysis model, and then performs result query based on real-time query requests; through biased translation, the accuracy and professionalism of the vocabulary are improved, making the model building more directional and facilitating accurate analysis of power data.

[0068] Example 2: Reference Figure 2 , the difference between this embodiment and embodiment 1 is that a repeat determination step is provided between the second extraction step and the data reprocessing step; Repeat judgment: including character distance judgment step, relationship extraction step, vocabulary repeat judgment step, deletion step, spacing word formation step, combination step and word meaning judgment step; Character distance judgment: obtain the character distance between two adjacent first words, and judge whether it is less than or equal to T. If so, perform word duplication judgment, otherwise, perform the relationship extraction step; Relation extraction: determine whether there is additional information between two adjacent first words. If so, extract the relationship attribute from the additional information and store it in the second data, and then perform data reprocessing steps; Word duplication judgment: judge whether two adjacent first words are the same, if so, execute the deletion step; otherwise, execute the combination step; Deletion: Delete one of the two adjacent first words, and then perform the interval word formation step; Interval word formation: the remaining first word is combined with the interval characters between the previous and next first words to form a new first word, and then the word meaning judgment step is performed; Combining: combining two adjacent first words to form a new first word, and then executing the word meaning judgment step; Word meaning judgment: judge whether the meaning of the first word is complete. If so, form the first word, optimize the second data, and then execute the data reprocessing step. Otherwise, execute the correction step.

[0069] Specifically, the character distance between two adjacent first words (ie, word segmentation results) is calculated. The character distance can be defined as the number of characters between the starting positions of the two words (excluding the words themselves).

[0070] A threshold T is set. If the character distance is less than or equal to T, it is considered that the two words may be too close, and it is necessary to further determine whether there is duplication or need to be merged. The value of T can be determined based on wrong characters, incorrect spellings, and historical data. In this embodiment, it is preferably 4 characters, that is, the distance between two Chinese characters. If the character distance judgment result is "yes", then enter the word duplication judgment; directly compare the text contents of the two adjacent first words. If they are exactly the same, it is considered as a word duplication; if the word duplication judgment is "yes", perform the deletion step; delete one of the repeated words to reduce data redundancy; if the word duplication judgment is "no", try to perform interval word grouping; combine the remaining word with the interval characters (such as punctuation marks, spaces, etc.) between the previous and next words to try to form a new and meaningful word; if the interval word grouping is not applicable or a meaningful word cannot be formed, try to directly combine the two adjacent first words; the new word after the combination needs to be judged for word meaning integrity, and the new word formed after the combination or the interval word grouping is judged for word meaning integrity.

[0071] The judgment basis may be the commonness of the vocabulary in the field of power systems, context, etc.; if the meaning of the word is complete, the second data is optimized (ie, the vocabulary list and vocabulary relationship are updated), otherwise a correction step is performed.

[0072] If the character distance judgment result is "no", enter the relationship extraction step.

[0073] Check whether there is additional information (such as prepositions, conjunctions, etc.) between two adjacent first words, and the additional information may imply a certain relationship between the words; extract the relationship attributes from the additional information and store them in the second data.

[0074] Executed when the word meaning is judged as "no" or other corrections are required; the correction steps may include manual correction, re-segmentation, adjustment of word order, addition or deletion of words, etc.

[0075] For example, when a power company processes a power accident report, it needs to perform word segmentation and relationship extraction on the text data in the report. In order to improve the accuracy of word segmentation and relationship extraction, a repeated judgment step is added between the second extraction step and the data reprocessing step.

[0076] Use the word segmentation tool to segment the power accident report and obtain a series of first words; calculate the character distance between two adjacent first words, assuming that the threshold T is set to 4 characters, if the character distance is less than or equal to 4, enter the word duplication judgment step; find that "transformer" and "transformer" are adjacent and the character distance is less than or equal to 4; delete one of the "transformer" words, because the duplicate words have been deleted, there is no need to perform interval word grouping, and also because the duplicate words have been deleted, there is no need to perform splicing.

[0077] By checking the additional information between the remaining words, it is found that there is a relationship between "cause" and "fault", and it is extracted as "fault cause". The extracted relationship is judged to be complete in meaning, and it is judged to be complete in meaning, so the second data is optimized.

[0078] Reference Figure 3 In other embodiments, an invalidity judgment step is provided between the character distance judgment step and the relationship extraction step; Invalidity judgment: judge whether there are continuous repeated strings in the first word. If yes, execute the word optimization step. Otherwise, execute the relationship extraction step. Optimize vocabulary: filter the string, optimize the first vocabulary, and then perform the word meaning determination step.

[0079] Specifically, this step will scan each first word to check whether there are continuous repeated strings in it. Here, "continuous repetition" means that two or more identical strings appear next to each other without other non-repeating characters inserted between them. Once continuous repeated strings are identified, the system will perform the vocabulary optimization step, which will filter out these repeated strings and keep only one, thereby optimizing the first vocabulary list. The optimized vocabulary will be more concise and more accurately reflect the meaning of the original text.

[0080] After the optimized vocabulary is processed, the system will enter the word meaning judgment step, which will evaluate the semantic integrity of the optimized vocabulary to ensure that they can accurately convey the information of the original text. The system will locate and delete consecutive repeated strings in the first vocabulary to ensure that each word is unique and valid; after filtering out repeated strings, the system will update the first vocabulary list to reflect this change.

[0081] For example, a power company is processing a batch of power inspection reports, which contain a large amount of equipment status descriptions and fault information.

[0082] First, the word segmentation tool is used to segment the power inspection report to obtain a series of first words; then, the character distance between two adjacent first words is calculated, and a threshold is set to determine whether they are too close; after the character distance judgment, the system enters the invalid judgment step; in this step, the system scans each first word and finds a continuous repeated character string such as "leading to leading"; the system recognizes that this is invalid information and decides to execute the word optimization step; the system filters out the repeated parts in "leading to leading", retaining only one "leading", and updates the first word list to reflect this change; after the optimized word processing, the system enters the word meaning judgment step; the optimized words are evaluated for word meaning integrity to ensure that they can accurately convey the information of the original text.

[0083] For example, for a word combination such as "transformer failure", the system will determine whether its meaning is complete and perform subsequent relationship extraction based on this; after word meaning judgment, the system will perform relationship extraction to extract useful data and information from the text; these data and information will be used in subsequent tasks such as power data analysis, fault prediction and decision support.

[0084] Reference Figure 4 In other embodiments, an abnormality determination step is further provided after the result output step; Abnormal determination: including: positioning step, first acquisition step and node replacement step; Positioning: The abnormal data location is recorded as an abnormal node, and the breakpoints before and after the abnormal node are obtained, recorded as the i-th breakpoint and the i+1-th breakpoint; First acquisition: obtain the replacement branch between the i-th breakpoint and the i+1-th breakpoint, recorded as the repair node; Node replacement: Replace abnormal nodes with repair nodes.

[0085] Specifically, first, the system will scan each data node in the output result and identify abnormal data through preset rules or algorithms. These abnormal data may appear as unreasonable values, missing values, repeated values, etc. These abnormal data may appear as unreasonable voltage values, current values, etc.; once the abnormal data is identified, the system will mark its position as an abnormal node. At the same time, the system will obtain the breakpoints before and after the abnormal node, that is, the i-th breakpoint and the i+1-th breakpoint. These two breakpoints define the specific position range of the abnormal data in the output result; after determining the position of the abnormal node, the system will start searching for the corresponding pre-set replacement branch, that is, the repair node. This repair node should be able to reasonably replace the abnormal node to restore the integrity and accuracy of the output result; the system will replace the abnormal node with the repair node according to the preset replacement rules or algorithms; after the replacement, the abnormal data in the output result may be corrected, thereby improving the accuracy and reliability of the data, and being able to maintain the accurate operation of the power system and reduce the probability that maintenance personnel must be on site for immediate maintenance.

[0086] After the replacement is completed, the system will update the output results to reflect this change. If the updated output results do not contain any abnormal data, they can be used for subsequent decision analysis, report generation and other tasks.

[0087] In other implementations, after the node replacement step, a result re-search step is also provided; Result re-search: Get the power data between the i-th breakpoint and the i+1-th breakpoint, form the query demand again, and then perform the verification and judgment steps; Verification and judgment: judge whether the abnormal data is eliminated. If so, execute the fault location step; otherwise, execute the output result step; Fault location: Determine whether the abnormal node between the i-th breakpoint and the i+1-th breakpoint is indeed abnormal, and log it in the database, and then execute the output result step again.

[0088] Specifically, after the node replacement step is completed, the system will re-acquire the power data between the i-th breakpoint and the i+1-th breakpoint. These data are new data after exception processing and are used to verify whether the abnormal data has been effectively eliminated.

[0089] Query demand formation: Based on the re-acquired data, the system will form a new query demand. This query demand aims to verify whether the abnormal data has been replaced with reasonable data and whether the entire data sequence maintains consistency and accuracy; the system will perform a series of verification operations to determine whether the abnormal data has been effectively eliminated; these verification operations may include data rationality checks, data consistency checks, etc.; based on the verification results, the system will choose to perform the fault location step or the output result step; if the abnormal data has been eliminated, the system will perform the fault location step to record the process and results of the abnormal processing; if the abnormal data still exists, the system will re-execute the output result step and may need to further analyze the cause and solution of the abnormal data; if the verification judgment step determines that the abnormal data still exists, the system will further confirm the abnormal node between the i-th breakpoint and the i+1-th breakpoint. This step is to ensure the accuracy of the abnormal node and avoid false positives or false negatives; once it is confirmed that the abnormal node is abnormal, the system will form a log record and store it in the database. This log record contains information such as the location of the abnormal node, the type of abnormality, and the processing process, which provides an important reference for subsequent data analysis and troubleshooting.

[0090] In other implementations, a survey step is further provided between the verification and judgment and the result output steps; Investigation: According to the SDG directed acyclic graph, determine the predecessor associated nodes of the abnormal node, and then perform the replacement steps one by one; Replace one by one: Get the previous and next breakpoints of the previous associated node, get the repair node, replace them one by one, and execute the result re-search step.

[0091] Specifically, first, the SDG model is constructed based on the system structure and functional relationship. The nodes in the SDG represent system elements (such as sensors, actuators, etc.), and the directed edges represent the causal relationship between the elements; by monitoring the system operation status, the abnormal nodes are identified, which may be manifested as data anomalies, functional failures, etc.; according to the SDG model, the predecessor-related nodes of the abnormal nodes are found; for each predecessor-related node, its front and back breakpoints (i.e., input and output connection points) are determined, and possible repair nodes (i.e., replacement nodes or repair strategies) are prepared; to ensure that other parts of the system are not affected during the replacement process, the input and output of the replacement node need to be precisely controlled; according to the characteristics of the predecessor-related node and the type of anomaly, the appropriate repair node is selected, which can be a backup node, a repaired original node, or an alternative node with similar functions; the predecessor-related nodes are replaced one by one, and the result re-search step is performed after each replacement, that is, the system status is re-evaluated to check whether the anomaly is eliminated.

[0092] After each replacement operation, verify whether the anomaly has been fixed by monitoring the system status or running specific test cases. If the anomaly still exists, continue to replace the next predecessor-related node; if the anomaly disappears, record the repair strategy and prepare to output the results.

[0093] When the abnormal node is found and successfully repaired, the repair strategy, the replaced node information and the status of the system after recovery are output.

[0094] For example, suppose that in an industrial automation system, sensor S1 detects abnormal data. According to the SDG model, the preceding associated nodes of S1 include controller K1 and power supply P1.

[0095] It is determined that the preceding associated nodes of S1 are K1 and P1, and the repair nodes are a backup controller K1_backup and a stable power supply P1_stable.

[0096] Disconnect K1 and S1, connect K1_backup, and re-search the results. It is found that the exception still exists. Disconnect P1 and S1, connect P1_stable, and re-search the results. It is found that the exception disappears and the system returns to normal.

[0097] The output repair strategy is recorded as replacing power supply P1 with P1_stable, and the replacement node information is recorded: the original power supply P1 is replaced with P1_stable.

[0098] Reference Figure 5 In other embodiments, after the real-time monitoring step, a demand supplement step is also provided; Supplementary requirements: including: image data acquisition steps, data division steps and real-time extraction steps; Image data acquisition: acquiring real-time images at the location required for query and forming image data; Data division: Divide the image data into multiple image blocks with different precisions; Real-time extraction: extract the relevant information in the image block, enrich the query requirements, then input it into the knowledge ontology analysis model, and execute the output result step.

[0099] Specifically, when an abnormality or specific event is monitored in real time, the system will immediately capture real-time images at that location. These image data can be video frames from surveillance cameras, satellite images, pictures taken by drones, etc.

[0100] The captured image data may need to be preprocessed, such as denoising, contrast enhancement, color correction, etc., to improve image quality and the accuracy of subsequent processing; the preprocessed image data will be divided into multiple image blocks with different precisions. These image blocks can be image areas of different resolutions, different sizes or different focal lengths. The purpose of division is to be able to select image data of different precisions for analysis as needed in subsequent processing; the precision of each image block can be set according to actual needs. For example, for areas that require detailed analysis, high-resolution image blocks can be selected, while for areas that only require a general understanding, low-resolution image blocks can be selected.

[0101] In the image block, the system will extract related information related to the query requirements. This related information can be features such as objects, colors, textures, shapes, etc. in the image, or text information such as words and numbers in the image. The purpose of extracting related information is to further enrich the query requirements and improve the accuracy of the output results.

[0102] The extracted related information will be used to enrich the original query requirements. For example, if the original query requirement is to detect abnormal events in a certain area, the extracted related information (such as the shape and color of the abnormal objects) can be added to the query requirements to more accurately identify abnormal events in subsequent processing.

[0103] After enriching the query requirements, the system will prepare to output the results. This output result may be a report containing rich information, an image marked with key information, or a video containing detailed analysis, etc. By enriching the query requirements, the results will be more targeted and accurate.

[0104] For example, a power company is responsible for real-time monitoring and analysis of power data in multiple regions. At a certain moment, the company found that the power consumption data in a certain area increased abnormally. In order to gain a deeper understanding of the power consumption in this area and find out the reasons for the abnormal data, the company decided to adopt an advanced technical solution for demand supplementation and output result preparation, with a special focus on the processing and analysis of power data.

[0105] Real-time monitoring: First, the power company's monitoring system will collect and monitor the power data of each area in real time. When abnormal power consumption data in a certain area is found, the system will immediately trigger the demand replenishment step.

[0106] Image and data acquisition: Real-time image data related to the area (such as video frames from surveillance cameras) and more detailed power data (such as power consumption of each power unit, power load changes, etc.) are obtained, and these data are pre-processed. The pre-processed power data will be divided into multiple time periods or different power units for more detailed analysis. At the same time, similar divisions will be made for image data. In the divided data set, the system will extract related information related to abnormal power consumption; for example, the system can analyze the power consumption trend of each power unit and identify units with abnormally increased power consumption; at the same time, the system can also combine image data to analyze whether there is a possible power failure or illegal power use in the area; the extracted related information will be used to enrich the original query requirements. For example, the original query requirement may be "detect abnormal power consumption data in a certain area", and the enriched query requirement may become "detect abnormal power consumption data in a certain area and analyze the causes of the abnormality (such as abnormally increased power consumption of power units, power failures, illegal power use, etc.)".

[0107] After enriching the query requirements, the system will prepare the output results, which may be a report containing power consumption anomaly analysis, cause identification and recommended measures; it may also be a chart or image with key information (such as abnormal power consumption units, power fault location, etc.); or a video or animation containing detailed analysis (such as power load change trends, power consumption comparison of power units, etc.).

[0108] In other embodiments, an abnormality determination step is further provided after the live extraction step; Abnormality judgment: For each image block, determine whether there is surface abnormality information. If so, determine the fault type based on the output result and the surface abnormality information and execute the output result step.

[0109] Specifically, the system obtains real-time image data and power data from surveillance cameras and power monitoring equipment, and performs preprocessing and feature extraction.

[0110] Abnormal judgment: For each image block, the system detects surface abnormality information. In a specific image block, the system detects color abnormalities, indicating that the device may be overheating. At the same time, the system combines power data analysis and finds that the power load corresponding to the device has increased abnormally, further confirming the possibility of overheating failure.

[0111] The system uses expert systems and knowledge bases, combined with abnormal information and output results, to preliminarily determine that the fault is an abnormal power load caused by equipment overheating; in order to more accurately locate the fault point, the system performs precise image analysis and extracts the edge and contour information of the equipment. At the same time, the system compares the equipment status monitoring data and finds that the temperature data of the equipment has increased abnormally, which is consistent with an overheating fault; according to the system's suggestions, the power company sent technicians to conduct on-site inspections. The technicians visually checked the equipment status and found that the equipment was indeed overheating, and collected more information about the fault; the system ultimately determined that the fault point was the overheated equipment and generated a detailed fault report, which included the fault type, location, possible causes (such as poor heat dissipation, equipment aging, etc.) and recommended repair measures (such as replacing the cooling fan, upgrading the equipment, etc.).

[0112] The embodiment of the present application also discloses a power data analysis system.

[0113] The power data analysis system includes the following modules: Database: used to store historical power data, fault information, professional related data and real-time power data; Data acquisition module: The output end is connected to the input end of the database, and is used to collect historical power data and fault information in the power system, collect power system and power equipment information and store them as professional related data; collect power analysis reports and accident reports, store them as analysis data, and store them in the database; Data preprocessing module: The input end is connected to the output end of the database and is used to format and standardize historical power data, fault information and professional related data; The first extraction module: the input end is connected to the output end of the data preprocessing module, and is used to extract information from the preprocessed data; The second extraction module: the input end is connected to the output end of the first extraction module, and is used to segment the analysis data according to the data extracted by the first extraction module to form second data; Repeat judgment module: the input end is connected to the output end of the second extraction module, and is used to judge whether there is repetition or additional information in the first word, and to group or combine words, and then judge whether the meaning of the word is complete; Data reprocessing module: the input end is connected to the output end of the duplication judgment module, and is used to filter the second data; Biased translation module: the input end is connected to the output end of the data reprocessing module, and is used to perform field bias conversion on the first word in the second data according to the professional field information of the analyst; Similarity calculation module: the input end is connected to the output end of the biased translation module, and is used to calculate the similarity between the translated vocabulary and the professional vocabulary in the first data; Similarity judgment module: the input end is connected to the output end of the similarity calculation module, and is used to judge whether the similarity is greater than the set limit value; A replacement module: an input end is connected to an output end of the similarity judgment module, and is used to replace the translated vocabulary according to the first data when the similarity is greater than a set limit value; Correction module: the input end is connected to the output end of the similarity judgment module, and is used to call out the original sentence of the translated vocabulary when the similarity is less than or equal to the set limit value, and correct the first vocabulary according to the semantics; Knowledge ontology construction: The input end is connected with the output end of the correction module, the output end of the replacement module and the output end of the database to build a knowledge ontology analysis model; Real-time monitoring module: The output end is connected to the input end of the knowledge ontology construction module to obtain power data in real time and form query requirements; Demand supplement module: The input end is connected to the output end of the real-time monitoring module, and the output end is connected to the input end of the knowledge ontology construction module. It is used to obtain the image information of the location where the data belongs, and analyze and enrich the query requirements to facilitate multi-condition queries, making the query results more accurate.

[0114] Output result module: The input end is connected to the output end of the knowledge ontology construction module, and is used to retrieve the vocabulary relationship according to the query vocabulary in the query requirements and generate the final result; Abnormal determination module: The input end is connected to the output end of the output result module, and is used to determine the location of abnormal data, replace it, and verify whether the repair is completed after the replacement.

[0115] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for analyzing power data, characterized in that: The following steps are involved: Build database: Establish a power system database to store historical power data, fault information and professional related data; Data collection: Collect historical power data and fault information in the power system, collect power system and power equipment information and store it as professional related data. The power equipment information includes the name of the power equipment and the product name data, which can be obtained by the relevant information of the power equipment and power dictionary pages in the professional comprehensive portal website; collect power analysis reports and accident reports and store them as analysis data; Data preprocessing: formatting and standardizing historical power data, fault information and professional related data; First extraction: extract the knowledge related to the operation law and fault characteristics of the power system from the preprocessed data, and record it as the first data; Second extraction: combining the domain professional terms and the first data, segmenting the analysis data to form a first vocabulary, and establishing vocabulary relationships based on the sentence semantics; then combining the first vocabulary with the vocabulary relationship and storing them as the second data; Data reprocessing: filtering the second data; Biased translation: Obtain the professional field information of the analyst, and convert the first word in the second data into professional words or similar written words with a field bias to form a translated word set; Similarity calculation: performing similarity calculation on the words in the translated vocabulary set and the professional words in the first data to obtain a similarity set; Similarity judgment: judge whether the similarity is greater than the set limit value, if so, execute the replacement step, otherwise, execute the correction step; Replacement: replace the translated words according to the first data, and store them as third data in combination with the word relationship; Correction: call out the original sentence of the translated vocabulary and correct the first vocabulary according to the semantics; Knowledge ontology construction: Build a knowledge ontology analysis model based on the first data, third data, power system, historical power data and fault information; Real-time monitoring: obtain power data in real time to form query needs; Output results: According to the query terms in the query requirements, they are input into the knowledge ontology analysis model, the word relationships are retrieved, and the final results are generated.

2. The power data analysis method according to claim 1, characterized in that: In the knowledge ontology construction step, it also includes determining the propagation path according to the power system construction and generating the SDG directed acyclic graph.

3. The power data analysis method according to claim 2, characterized in that: A repeat determination step is also provided between the second extraction step and the data reprocessing step; Repeat judgment: including character distance judgment step, relationship extraction step, vocabulary repeat judgment step, deletion step, spacing word formation step, combination step and word meaning judgment step; Character distance judgment: obtain the character distance between two adjacent first words, and judge whether it is less than or equal to T. If so, perform word duplication judgment, otherwise, perform the relationship extraction step; Relation extraction: determine whether there is additional information between two adjacent first words. If so, extract the relationship attribute from the additional information and store it in the second data, and then perform data reprocessing steps; Word duplication judgment: judge whether two adjacent first words are the same, if so, execute the deletion step; otherwise, execute the combination step; Deletion: Delete one of the two adjacent first words, and then perform the interval word formation step; Interval word formation: the remaining first word is combined with the interval characters between the previous and next first words to form a new first word, and then the word meaning judgment step is performed; Combining: combining two adjacent first words to form a new first word, and then executing the word meaning judgment step; Word meaning judgment: judge whether the meaning of the first word is complete. If so, form the first word, optimize the second data, and then execute the data reprocessing step. Otherwise, execute the correction step.

4. The power data analysis method according to claim 3, characterized in that: Between the character distance judgment step and the relationship extraction step, an invalidity judgment step is also provided; Invalidity judgment: judge whether there are continuous repeated strings in the first word. If yes, execute the word optimization step. Otherwise, execute the relationship extraction step. Optimize vocabulary: filter the string, optimize the first vocabulary, and then perform the word meaning determination step.

5. The power data analysis method according to any one of claims 2 to 4, characterized in that: After the result output step, an abnormality determination step is also provided; Abnormal determination: including: positioning step, first acquisition step and node replacement step; Positioning: The abnormal data location is recorded as an abnormal node, and the breakpoints before and after the abnormal node are obtained, recorded as the i-th breakpoint and the i+1-th breakpoint; First acquisition: obtain the replacement branch between the i-th breakpoint and the i+1-th breakpoint, recorded as the repair node; Node replacement: Replace abnormal nodes with repair nodes.

6. The power data analysis method according to claim 5, characterized in that: After the node replacement step, a result re-search step is also provided; Result re-search: Get the power data between the i-th breakpoint and the i+1-th breakpoint, form the query demand again, and then perform the verification and judgment steps; Verification and judgment: judge whether the abnormal data is eliminated. If so, execute the fault location step; otherwise, execute the output result step; Fault location: Determine whether the abnormal node between the i-th breakpoint and the i+1-th breakpoint is indeed abnormal, and log it in the database, and then execute the output result step again.

7. The power data analysis method according to claim 6, characterized in that: Between the verification and judgment and the result output step, there is also a survey step; Investigation: According to the SDG directed acyclic graph, determine the predecessor associated nodes of the abnormal node, and then perform the replacement steps one by one; Replace one by one: Get the previous and next breakpoints of the previous associated node, get the repair node, replace them one by one, and execute the result re-search step.

8. The power data analysis method according to any one of claims 1 to 4, characterized in that: After the real-time monitoring step, a demand supplement step is also provided; Supplementary requirements: including: image data acquisition step, data division step and real-time extraction step; Image data acquisition: acquiring real-time images at the location required for query and forming image data; Data division: Divide the image data into multiple image blocks with different precisions; Real-time extraction: extract the relevant information in the image block, enrich the query requirements, then input it into the knowledge ontology analysis model, and execute the output result step.

9. The power data analysis method according to claim 8, characterized in that: An abnormality judgment step is also provided after the live situation extraction step; Abnormality judgment: For each image block, determine whether there is surface abnormality information. If so, determine the fault type based on the output result and the surface abnormality information, and execute the output result step.

10. A power data analysis system, used in the power data analysis method according to any one of claims 1 to 9, characterized in that: Includes the following modules: Database: used to store historical power data, fault information, professional related data and real-time power data; Data acquisition module: The output end is connected to the input end of the database, which is used to collect historical power data and fault information in the power system, collect power system and power equipment information and store it as professional related data; Collect power analysis reports and accident reports, store them as analysis data, and store them in a database; Data preprocessing module: The input end is connected to the output end of the database and is used to format and standardize historical power data, fault information and professional related data; The first extraction module: the input end is connected to the output end of the data preprocessing module, and is used to extract information from the preprocessed data; The second extraction module: the input end is connected to the output end of the first extraction module, and is used to segment the analysis data according to the data extracted by the first extraction module to form second data; Data reprocessing module: the input end is connected to the output end of the second extraction module, and is used to filter the second data; Biased translation module: the input end is connected to the output end of the data reprocessing module, and is used to perform field bias conversion on the first word in the second data according to the professional field information of the analyst; Similarity calculation module: the input end is connected to the output end of the biased translation module, and is used to calculate the similarity between the translated vocabulary and the professional vocabulary in the first data; Similarity judgment module: the input end is connected to the output end of the similarity calculation module, and is used to judge whether the similarity is greater than the set limit value; A replacement module: an input end is connected to an output end of the similarity judgment module, and is used to replace the translated vocabulary according to the first data when the similarity is greater than a set limit value; Correction module: the input end is connected to the output end of the similarity judgment module, and is used to call out the original sentence of the translated vocabulary when the similarity is less than or equal to the set limit value, and correct the first vocabulary according to the semantics; Knowledge ontology construction: The input end is connected with the output end of the correction module, the output end of the replacement module and the output end of the database to build a knowledge ontology analysis model; Real-time monitoring module: The output end is connected to the input end of the knowledge ontology construction module to obtain power data in real time and form query requirements; Output result: The input end is connected to the output end of the knowledge ontology construction module, and is used to retrieve the vocabulary relationship according to the query vocabulary in the query requirement to generate the final result.

Images