A power quality knowledge processing method and system based on ontology model

Abstract

This invention discloses a power quality knowledge processing method and system based on an ontology model, belonging to the field of power quality monitoring and knowledge management technology. It aims to solve the technical problems of low retrieval efficiency, weak semantic association, and disconnect between models and applications in traditional power quality knowledge management. This invention includes: formally describing a knowledge ontology in the power quality domain; preprocessing standard text and constructing a professional glossary; extracting candidate terms from the text using a fast automatic keyword extraction method, and filtering key domain terms based on word frequency and co-occurrence frequency; mining semantic relationships between terms, defining a set of relational functions, and establishing ontology units; establishing an ontology class system and subclasses including basic measurement parameter classes and power quality event classes; building a multi-dimensional information fusion model framework; constructing and storing the knowledge ontology model; receiving query or analysis requests, and outputting structured knowledge association results. This technical solution effectively improves knowledge retrieval accuracy and collaborative analysis capabilities.

Description

Technical Field

[0001] This invention relates to the field of power quality monitoring, and in particular to a power quality knowledge processing method and system based on ontology models. Background Technology

[0002] With the large-scale integration of distributed power sources and power electronic loads into the power grid, power quality issues have become increasingly complex. Distributed power sources, such as solar photovoltaic power generation and wind power generation, are affected by natural conditions, exhibiting randomness and intermittency, leading to frequent power fluctuations in the power grid. For example, photovoltaic power output drops sharply when obscured by clouds; wind power output also fluctuates significantly when wind speeds are unstable. Power electronic loads, such as frequency converters and rectifiers, generate numerous harmonics during operation, which, when injected into the grid, distort the voltage waveform. Frequency converters, widely used in industrial production, can generate harmonics that interfere with the normal operation of surrounding equipment. Furthermore, the integration of these new power devices can cause voltage fluctuations and three-phase imbalances. For instance, concentrated integration of distributed power sources into a region may cause voltage fluctuations exceeding permissible limits, affecting the normal operation of user equipment; the extensive use of single-phase power electronic loads can cause three-phase current imbalances, increasing line losses and even triggering transformer overloads. These complex power quality issues pose significant challenges to the safe and stable operation of the power grid and the normal electricity consumption of users.

[0003] Traditional power quality knowledge descriptions primarily employ catalog-based retrieval, which becomes inefficient when faced with increasingly complex power quality issues. When searching for solutions to specific power quality problems, it may require navigating through numerous documents, making it difficult to quickly and accurately obtain the necessary information and failing to meet the personalized and diverse needs of power quality management. Furthermore, existing power quality ontology models are limited in variety and lack versatility. Different power systems or application scenarios may require different ontology models to accurately describe power quality knowledge, but existing models often cannot adequately adapt to these differences. For example, in a smart grid environment, the impact of distributed energy access and electricity market transactions on power quality needs to be considered, but traditional ontology models lack effective descriptions of these new factors and cannot comprehensively and accurately express the power quality-related knowledge system, making it difficult to achieve in-depth analysis and effective solutions to power quality problems in practical applications.

[0004] In the various stages of the power grid—generation, transmission, transformation, distribution, and consumption—data sharing is ineffective due to significant systemic barriers. Generation primarily focuses on the operational status of generating equipment and power output; transmission emphasizes the transmission capacity and stability of transmission lines; transformation focuses on voltage transformation and distribution; distribution focuses on the reliability and power quality of the distribution network; and consumption focuses on user demand and load characteristics. Differences in data formats, communication protocols, and management methods across these stages hinder information sharing and interaction. For example, data between power generation and supply companies is often not shared in real time, making it impossible to comprehensively consider various factors in the generation and supply processes during power quality analysis, resulting in a disconnect in management. When power quality issues arise, collaboration between these stages is difficult to achieve, hindering the rapid identification of the root cause and the implementation of effective solutions, severely impacting the overall operational efficiency of the power grid and the improvement of power quality.

[0005] Therefore, there is an urgent need for a technical solution that can structure and semantically associate multi-source heterogeneous knowledge in the field of power quality, and support knowledge services for complex queries and analytical reasoning. Summary of the Invention

[0006] The technical problem to be solved and the technical task proposed by this invention is to improve and refine existing technical solutions, and to provide a power quality knowledge processing method based on an ontology model. This aims to enhance the automation, structuring, and semantic level of knowledge processing in the power quality field, and to solve the problems of low retrieval efficiency, weak semantic connections, and data incompatibility across different stages in traditional knowledge management. To this end, this invention adopts the following technical solution.

[0007] Firstly, an ontology-based method for processing power quality knowledge is provided, which includes the following steps: 1) To formally describe the knowledge ontology in the field of power quality; 2) Preprocess standard texts in the field of power quality and construct a professional glossary by combining professional terminology; 3) Extract candidate terms from the preprocessed text and filter key domain terms based on word frequency and word co-occurrence relationships; 4) Mine the semantic relationships between the key domain terms and establish ontology units that represent these semantic relationships; 5) Establish a knowledge ontology class structure for the power quality domain; 6) Based on the aforementioned knowledge ontology class architecture, build a power quality knowledge ontology model framework that integrates multi-dimensional information; 7) Based on the key domain terms, semantic relationships, ontology units, and ontology model framework, construct and store a knowledge ontology model for the power quality domain; 8) Receive query or analysis requests, call the power quality domain knowledge ontology model for semantic processing, and output knowledge association results.

[0008] This technical solution automates and structures the entire process of knowledge processing in the field of power quality, from text preprocessing to knowledge application. It solves the problems of scattered, disorganized, and difficult-to-efficiently retrieve and analyze traditional power quality knowledge. Through a closed-loop design from text processing to model building and knowledge application, it achieves accurate extraction, orderly organization, and efficient reuse of knowledge, significantly improving the efficiency and standardization of knowledge processing in the field of power quality. The overall process is tailored to the characteristics of knowledge in the field of power quality, accurately capturing core knowledge and providing a unified and rigorous knowledge carrier for subsequent knowledge retrieval and analysis, thus improving the convenience and reliability of knowledge application.

[0009] As a preferred technical means: in step 1), the knowledge ontology in the field of power quality is formally described using the concept of six tuples. The six tuples include a set of concepts, a set of attributes, a set of relations, a set of relation attributes, a classification system, and a set of axioms.

[0010] This technical solution comprehensively and systematically captures knowledge elements in the field of power quality through the division of six major sets, ensuring the integrity and rigor of the formal description, and providing a clear and unified basic framework for subsequent knowledge extraction and ontology construction.

[0011] As a preferred technical means: the construction of the professional vocabulary in step 2) includes: introducing professional terms from standard texts in the field of power quality, classifying and integrating them according to the parts of speech of nouns, verbs and adjectives, expanding the vocabulary database, and generating a professional vocabulary containing power quality professional terms and their word frequency statistics; among which, the noun category includes voltage sag, frequency deviation, harmonics, transient oscillation and flicker coefficient, the verb category includes reactive power compensation, harmonic suppression, flicker value testing and waveform distortion elimination, and the adjective category includes active, passive, load, zero-sequence and non-periodic.

[0012] This technical solution aligns with the specialized characteristics of the power quality field. By integrating parts-of-speech classifications, the professional vocabulary becomes more organized, facilitating accurate identification of domain terms by word segmentation tools and improving segmentation accuracy. The introduction of professional terms from standard texts and the statistical analysis of word frequencies ensure the professionalism and practicality of the vocabulary, providing reliable lexical support for subsequent candidate term extraction and key term selection.

[0013] As a preferred technical approach: Step 3) identifies key domain terms, including: 301) Extraction of candidate terms: Based on the preprocessed text, the text is segmented into several candidate keyword phrases using stop words and punctuation marks; each candidate keyword phrase is treated as a candidate term, and each word contained in each candidate keyword phrase is treated as a candidate word; 302) Calculation of word frequency: The word frequency of any candidate word is equal to the number of times that candidate word appears in the text divided by the sum of the number of times all words appear in the text; 303) Calculation of word weight: The word weight of any candidate word is equal to the sum of the number of times the candidate word co-occurs with all its co-occurring words; wherein, the co-occurring words refer to other words that co-occur with the candidate word in the same candidate keyword phrase, and the number of co-occurrences refers to the number of times the candidate word co-occurs with a certain co-occurring word in the same candidate keyword phrase; 304) Calculation of phrase weight: The weight of any candidate term is equal to the phrase weight of the candidate keyword phrase corresponding to the candidate term, and the phrase weight is equal to the sum of the word weights of all candidate words in the candidate keyword phrase; 305) Screening of key domain terms: Sort all candidate terms from high to low according to phrase weight, and extract candidate terms whose phrase weight is greater than or equal to the threshold as key domain terms.

[0014] This technical solution preserves the integrity of multi-word combination terms by segmenting candidate keyword phrases, avoiding the incorrect segmentation of compound concepts; through hierarchical calculation of word frequency, word weight, and phrase weight, it scientifically evaluates the domain importance of each candidate term, ensuring that the selected key domain terms accurately correspond to the core knowledge of the power quality domain; the preset threshold screening method can be flexibly adjusted according to actual text needs, improving the flexibility and adaptability of term selection.

[0015] As a preferred technical means: the terminology relation function set mentioned in step 4) includes: Superordinate and subordinate functions: used to represent the hierarchical inclusion relationship between terms; where the superordinate term is an abstract generalization of a certain type of thing, and the subordinate term is a specific category with specific attributes within that type of thing. The subordinate term inherits the core attributes of the superordinate term and has its own unique characteristics. Whole-part function: used to express the physical or functional relationship between the things referred to by the term; the whole term is a system, device or collection of concepts with independent functions, and the part terms are indispensable components of the whole term; Attribute functions: used to represent the correspondence between a term and its inherent characteristics, quantification parameters, or descriptive dimensions; attribute terms are precise quantitative or qualitative descriptions of the technical characteristics of the subject term. Causal function: used to represent the cause-and-effect logical relationship between terms. The cause term is the triggering factor that leads to power quality abnormalities, and the result term is the power quality phenomenon or problem that occurs. Based on the set of relational functions, terms and their associated elements are mapped to structured ontology units.

[0016] In this technical solution, each function describes a specific semantic relationship between a domain term and one or more related elements. Partial terms are indispensable components of the overall terminology; without the whole, the part cannot achieve its original function, and the overall function depends on the synergistic effect of the parts. Attribute terms are precise quantitative or qualitative descriptions of the technical characteristics of the subject term; attributes correspond one-to-one with the subject, and the subject's state can be defined through attribute values. This technical solution covers the core semantic relationships between terms in the power quality domain, solving the problems of ambiguous semantic relationships and difficulty in quantitative representation between terms. Through the precise definition of the four functions, the association logic between different types of terms can be clearly and systematically depicted, making the construction of ontology units more targeted and organized. Mapping terms and their related elements to structured ontology units provides modular knowledge units for subsequent ontology class system construction and model framework construction, improving the structure and semantic relevance of the ontology model. These are formalized into computable relational functions. Hierarchical relationships construct the hierarchical skeleton of the ontology class system, whole-part relationships depict the compositional logic between devices and systems, attribute relationships realize the decoupling expression of terms and quantitative indicators, and causal relationships provide reasoning paths for power quality event tracing and impact analysis. This set of relational functions fully covers the semantic dimensions required for the expression of power quality knowledge, enabling ontology units to be stored and retrieved in a standardized manner in the form of "terms-relations-related elements", laying the foundation for a reasonable knowledge structure for subsequent semantic processing.

[0017] As a preferred technical means, the knowledge ontology system mentioned in step 5) specifically includes: The basic measurement parameter class includes at least one of the following subclasses: voltage amplitude, current amplitude, frequency, power, flicker, harmonics, voltage imbalance, and interharmonics; each subclass can further include its subordinate subclasses. Power quality events include at least one of the following subclasses: voltage sag, voltage swell, voltage interruption, flicker, voltage imbalance, frequency deviation, voltage harmonics, current harmonics, and transients.

[0018] This technical solution constructs a dual-pillar ontology system centered on "basic measurement parameters" and "power quality events," clearly decoupling the monitoring data level from the phenomenon / event level. This classification architecture aligns with the business logic of "parameter monitoring—event identification—cause analysis—governance decision-making" in the power quality field, accommodating existing standard terminology while reserving interfaces for subclass extension. By combining class nodes with the relational functions defined in step 4, semantic associations between parameters and events can be achieved, providing an organizational framework for multi-dimensional information fusion.

[0019] As a preferred technical means: the ontology model framework in step 6) is based on the basic measurement parameter class and power quality event class and its subclass system. The term relation function set is used to graphically organize class nodes, term nodes and their semantic relations to form a power quality multi-dimensional information fusion framework characterized by the correlation and integrity of knowledge elements.

[0020] This technical solution graphically integrates class systems and relational function sets to construct an ontology model framework for multi-dimensional information on power quality. The framework uses a graph structure as its underlying representation, with class nodes and terminology nodes interconnected through hierarchical, whole-part, attribute, and causal relationships, forming a knowledge network ranging from basic parameters to complex events, from phenomena to causes, and from indicators to governance. This framework itself does not rely on specific data for filling, possessing excellent scalability and portability, and can serve as a unified knowledge base for various power quality management systems, supporting semantic interoperability between heterogeneous systems. It achieves the organic integration of class systems, terminology, and semantic relationships, solving the problems of scattered information and poor correlation in traditional knowledge models. Through graphical organization, the relationships between knowledge elements are intuitively visible, facilitating subsequent model understanding, invocation, and maintenance. With the correlation and integrity of knowledge elements as its core characteristics, it effectively integrates multi-dimensional information in the power quality field, breaking down barriers between different types of knowledge, and providing clear and efficient framework support for the construction of subsequent knowledge ontology models and knowledge applications.

[0021] As a preferred technical approach: Step 7) involves constructing and storing a knowledge ontology model for the power quality domain, which specifically includes the following sub-steps: 701) Using the input power quality-related standard text as a knowledge source, execute steps 1) to 2) to obtain a formal description framework and a glossary of professional terms; 702) Based on step 3), iteratively filter key domain terms from the preprocessed text until term extraction of all text is completed; 703) Semantically map and instantiate the selected set of key domain terms, the set of relational functions defined in step 4), and the ontology class system established in step 5; 704) Persistently store the instantiated model using a graph database or ontology storage language to form a power quality domain knowledge ontology model that can be called.

[0022] This technical solution improves the comprehensiveness of term extraction by iteratively screening key domain terms and avoids the omission of core terms; semantic mapping and instantiation filling realize the organic integration of terms, relational functions, and class systems, enabling the ontology model to have computable and reusable characteristics; persistent storage of graph databases or ontology storage languages ​​ensures the stability and callability of model storage.

[0023] Secondly, an ontology-based power quality knowledge processing system is provided, employing the aforementioned ontology-based power quality knowledge processing method. The system includes: The formal description module is used to formally describe knowledge ontology in the field of power quality; The text preprocessing and vocabulary building module is used to preprocess standard texts in the field of power quality and build a professional vocabulary list by combining professional terms. The terminology extraction and filtering module is used to extract candidate terms from preprocessed text and filter key domain terms based on word frequency and word co-occurrence relationships. The semantic relationship modeling module is used to mine the semantic relationships between the key domain terms and establish ontology units that represent the semantic relationships. The ontology class system construction module is used to establish the knowledge ontology class system architecture in the power quality domain; The model framework building module is used to build a power quality knowledge ontology model framework that integrates multi-dimensional information based on the knowledge ontology class architecture. The model generation and storage module is used to construct and store a knowledge ontology model of the power quality domain based on the key domain terms, semantic relationships, ontology units and ontology model framework. The knowledge service interface module is used to receive query or analysis requests, call the knowledge ontology model in the power quality domain for semantic processing, and output knowledge association results.

[0024] The text preprocessing and vocabulary construction module extracts the main text, performs word segmentation using a word segmentation tool, and introduces professional terms from standard texts in the power quality field, constructing a professional vocabulary list categorized by part of speech. The terminology extraction and screening module calculates the weight of each candidate term based on word frequency and co-occurrence frequency, sorts them from high to low weight, and filters out key domain terms. The ontology class system construction module includes two core classes: basic measurement parameters and power quality events, and defines corresponding subclasses for each class. The model generation and storage module integrates the formal description module, text preprocessing and vocabulary construction module, terminology extraction and screening module, semantic relationship modeling module, and ontology class system. The system's construction module and model framework building module process the results, instantiating and generating a computable and storable knowledge ontology model for the power quality domain, and persistently storing it. The functions of each module in this technical solution correspond one-to-one with the steps of the aforementioned knowledge processing method, enabling the implementation of all process flows and ensuring the standardization and efficiency of knowledge processing. The model generation and storage module integrates the results of each module, ensuring the integrity and consistency of the ontology model. The knowledge service interface module enables convenient access to knowledge queries and analysis, quickly outputting structured knowledge association results, improving the efficiency of knowledge application, and providing reliable system support for knowledge services in the power quality domain.

[0025] As a preferred technical means, the term extraction and screening module further includes: The candidate term extraction unit is used to segment the preprocessed text into several candidate keyword phrases using stop words and punctuation marks, treat each candidate keyword phrase as a candidate term, and treat each word contained in each candidate keyword phrase as a candidate word. The word frequency calculation unit is used to calculate the word frequency of any candidate word, wherein the word frequency is equal to the number of times the candidate word appears in the text divided by the sum of the number of times all words appear in the text; The word weight calculation unit is used to calculate the word weight of any candidate word, wherein the word weight is equal to the sum of the number of times the candidate word co-occurs with all its co-occurring words; The phrase weight calculation unit is used to calculate the weight of any candidate term. The weight is equal to the phrase weight of the candidate keyword phrase corresponding to the candidate term. The phrase weight is equal to the sum of the word weights of all candidate words in the candidate keyword phrase. The key term filtering unit is used to sort all candidate terms from high to low according to their phrase weights, and extract candidate terms whose phrase weights are greater than or equal to the threshold as key domain terms by setting a preset threshold.

[0026] In this technical solution, the co-occurring word refers to other words that appear together with the candidate word in the same candidate keyword phrase, and the co-occurrence count refers to the number of times the candidate word and a certain co-occurring word appear together in the same candidate keyword phrase. This technical solution subdivides the term extraction and screening module into multiple functional units, realizing a detailed division of labor for term extraction, weight calculation, and key screening. The function of each unit corresponds one-to-one with the sub-steps of key term screening in the aforementioned method, ensuring the standardization and normalization of the term extraction and screening process. The design of subdivided units makes the module function more flexible, facilitates the optimization and adjustment of individual units according to actual needs, and facilitates the maintenance and expansion of the module.

[0027] Beneficial effects: 1. It achieves automated construction from unstructured standard text to structured knowledge ontology, laying the foundation for terminology recognition through word segmentation and the construction of a professional vocabulary; it employs a fast automatic keyword extraction method to automatically filter key domain terms based on word frequency and co-occurrence frequency; and it automatically mines semantic relationships and establishes ontology units based on terminology extraction. The entire process requires no manual intervention, significantly improving the efficiency, repeatability, and scalability of knowledge modeling, and providing a feasible technical path for large-scale, continuous knowledge engineering in the field of power quality.

[0028] 2. By defining a set of terminology relation functions and establishing an ontology class system, scattered terms are organized into a knowledge network with clear semantic relation constraints. Semantic processing is then performed on this basis to identify semantic elements such as concepts, attributes, and relationships in query requests, returning structured knowledge results with causal chains, hierarchical affiliations, and attribute associations. This elevates knowledge retrieval from "string matching" to the level of "semantic understanding," significantly improving the accuracy, recall, and interpretability of knowledge discovery.

[0029] 3. By establishing a dual-pillar ontology system centered on basic measurement parameters and power quality events, an ontology model framework capable of integrating multi-dimensional information is built upon this foundation. This allows basic parameters such as voltage, current, frequency, and power, along with power quality events such as sags, harmonics, and flicker, to be uniformly expressed and correlated within the same knowledge framework, laying the foundation for achieving seamless knowledge integration across the entire chain of "phenomenon—cause—impact—governance."

[0030] 4. By constructing a knowledge ontology model for the power quality domain, a unified semantic description framework is provided for heterogeneous data across various stages. The knowledge service interface can receive query and analysis requests from all stages of the power grid, including generation, transmission, transformation, distribution, and consumption, and perform cross-node relationship tracing and path reasoning based on the semantic relationships defined in the model. This achieves semantic alignment and collaborative analysis of cross-stage knowledge in the field of power quality knowledge management, providing key technical support for tracing the source, assigning responsibility, and comprehensively managing power quality issues across the entire chain.

[0031] 5. The formal description of this technical solution provides a general knowledge representation framework; the ontology class system reserves subclass extension interfaces; the constructed model exists in a computable and storable form and can be persisted through graph databases or ontology storage languages; this enables the model to absorb new terms, new relationships, and new categories at low cost, and to be ported and reused across systems and platforms, effectively addressing the dynamic evolution needs of the knowledge system in the field of power quality. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the six-tuple of the main body of the present invention.

[0033] Figure 2 This is a block diagram of the basic measurement parameter class structure of the present invention.

[0034] Figure 3 This is a block diagram of the power quality event class structure of the present invention.

[0035] Figure 4 This is a schematic diagram of the basic ontology model framework of the present invention.

[0036] Figure 5 This is a flowchart of the power quality knowledge ontology model construction process of the present invention.

[0037] Figure 6 This is a flowchart of the present invention. Detailed Implementation

[0038] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings.

[0039] Example 1: This embodiment discloses a method for constructing a knowledge ontology model based on the field of power quality, including the following steps: S1: Provide a formal description of the knowledge ontology in the field of power quality.

[0040] In this step, the concept of six tuples is used to formally describe the ontology, thereby capturing knowledge in the field of power quality, providing a common understanding of knowledge in the field of power quality, identifying commonly accepted vocabulary in the field, and giving clear definitions of these vocabulary (terms) and the relationships between them from different levels of formal models.

[0041] The formal description of an ontology, represented by a six-tuple, is as follows.

[0042] .

[0043] In the formula: , , , , , .

[0044] C represents a collection of concepts or classes; R represents the set of attributes for each class; R represents the set of relations. H represents the set of attributes for each relation; H represents the classification system of the class; X represents the set of axioms. It is a series of objects of a class, and can be generated by... The attributes of the logo are used to describe it; yes and The binary relation, and can be... Described by attributes; of yes The parent class; Yes , The values ​​of the attributes and the restrictions on the relationships between R objects. Figure 1 This is a schematic diagram of the six-tuple of the ontology.

[0045] S2: Preprocess standard texts in the field of power quality and construct a professional glossary by combining professional terminology.

[0046] In this step, word segmentation software is used to segment the text and construct a professional vocabulary list. The text is preprocessed before word frequency calculation. By extracting the main text of the power quality standard document and using word segmentation and annotation processing, it is split into candidate nouns and verbs with independent descriptive meanings, respectively, to obtain nouns and their word frequency sets, as well as a set of marker verbs used to distinguish the relationship between terms.

[0047] Because Chinese text contains rich semantic information, preprocessing the input text, such as word segmentation and part-of-speech tagging, is a prerequisite for accurately extracting domain-specific terms. Different word segmentation methods can yield different semantics for a single sentence. However, when processing large amounts of information, relying solely on general dictionaries cannot identify long candidate words (nouns with three or more characters), leading to overly fine word segmentation. The obtained word frequency data cannot accurately reflect the importance and domain relevance of candidate words, directly affecting the term extraction results. Therefore, it is necessary to introduce domain-specific vocabulary during word segmentation. The national standard "Electric Power Quality Terminology" defines the basic terminology in the field of electric power quality and provides professional vocabulary resources. Classifying electric power quality terms by part of speech yields the electric power quality professional vocabulary list shown in Table 1.

[0048] Table 1. Glossary of Power Quality Terms

[0049] In text preprocessing, the vocabulary in Table 1 expands the word segmentation lexicon. The characters in the input text are matched with words in the lexicon, and the resulting text units are used as candidate words, with their frequency counted. After preprocessing, the candidate words and their frequency lists are exported for subsequent term extraction.

[0050] S3: Preprocess standard texts in the field of power quality and construct a professional glossary by combining professional terminology.

[0051] In this step, the Rapid Automatic Keyword Extraction (RAKE) method is used to calculate the weight of each candidate term based on word frequency and word co-occurrence frequency, and key domain terms are selected.

[0052] Terms used for ontology construction are extracted from the source files to obtain a categorized terminology library, including general nouns, verbs, and specialized terms. First, term frequency (TF) is calculated. Then, using the Fast Automatic Keyword Extraction (RAKE) method, which represents the independence of each candidate word, the text is split into candidate words based on TF and co-occurrence frequency. Their weights are calculated and ranked, and the candidate words are sorted from highest to lowest according to this weight, thus obtaining the key domain terms in the analyzed text.

[0053] Term frequency (TF) is the quotient of the number of times a word appears in a text and the sum of the number of times all words in the text appear. The formula is as follows:

[0054] In the formula: Let be the word frequency of the j-th candidate word in the i-th text; This represents the number of times the j-th candidate word appears in the i-th text. Let be the total number of occurrences of all candidate words in the i-th text. Based on the calculated word frequencies, sort the words and set a threshold to extract the conceptual entities.

[0055] The Rapid Automatic Keyword Extraction (RAKE) method is a keyword extraction technique that uses stop words and punctuation marks to segment text and obtain candidate keyword phrases. First, it calculates the term frequency (TF), the number of times each word in the candidate words appears; then, it calculates the co-occurrence frequency, the number of times words in the candidate words appear together; finally, it calculates the word weights, obtaining the sum of the term frequency and co-occurrence frequency. The phrase weight is the sum of the weights of the constituent words. The word weights are calculated as follows:

[0056] In the formula: The word weight of candidate word t; The frequency of candidate word t; For words The set of words that co-occur. for and The number of times they co-occur.

[0057] The phrase weight is calculated as follows:

[0058] In the formula: The phrase weight of candidate keyword phrase p; The candidate word t is the lexical composition of the candidate keyword phrase p. The reason for using the RAKE method is that it is simple and fast to calculate, and under the premise of certain preprocessing of the text and importing power quality domain terminology, the word frequency information used by the algorithm is accurate, so that the results are more consistent with the domain terminology.

[0059] S4: Mine the semantic relationships between the key domain terms and establish ontology units that represent the semantic relationships.

[0060] In this step, a set of term relation functions is defined to represent the relationships between terms, and an ontology unit representing a single semantic relationship is established based on this.

[0061] Based on text preprocessing and term extraction, it is necessary to fully explore the semantic relationships between power quality terms, determine the term relationship function set, and then establish ontology units.

[0062] By summarizing, the main relationships between power quality terms can be identified, including hierarchical relationships, whole-part relationships, attribute relationships, and causal relationships. To formally represent the relationships between these terms, the following relational functions are defined. Where A is a domain term, and B represents a set of multiple elements such as domain terms, feature words, and mathematical expressions.

[0063] Constructing a form like The terminology and its relational expressions.

[0064] (1) Upper and lower dimension functions The hierarchical relationship refers to the class-specific inclusion relationship between terms. The superordinate term (parent class) is an abstract generalization of a certain type of thing, while the subordinate term (child class) is a specific category within that type of thing with specific attributes. It satisfies the logical rule that "the child class inherits the core attributes of the parent class and possesses its own unique characteristics," and is the foundation for constructing an ontology hierarchy. Here, A is the superordinate term of B, and B is the subordinate term of A. For example, A = power quality disturbance. If it is an element in B, then It could be a voltage dip, voltage swell, voltage interruption, harmonics, interharmonics, etc.

[0065] (2) Global and partial functions The whole-part relationship refers to a physical or functional connection between the things referred to by the terms. The whole term is a system / device / concept set with independent functions, and the part terms are its indispensable constituent units, satisfying the logical requirement that "the part cannot achieve its original function without the whole, and the function of the whole depends on the synergistic effect of the parts." Here, B is a component of A, and the relational markers are "composed of," "includes," etc. Unlike the relationship represented by hierarchical functions, A needs to combine several elements from B to form a relatively complete entity in terms of structure and function. For example, A = power system, B = {generation side, transmission lines, transformers, distribution network, user side}.

[0066] (3) Attribute functions Attribute relationships refer to the precise quantitative or qualitative description of specific terms and their inherent characteristics, quantitative parameters, or descriptive dimensions, satisfying the logical rule of "one-to-one correspondence between attributes and subjects, and the state of the subject can be defined through attribute values." Here, B is used to explain the characteristics of A or its quantitative parameters and dimensional descriptions. For example, if A = harmonics, then B = {harmonic order, total harmonic distortion (THD), harmonic current amplitude}.

[0067] (4) Causal function Causality refers to a logical connection between terms where "the former causes the latter." The causal term is the triggering factor leading to abnormal power quality (such as equipment characteristics, external interference, operational behavior, etc.), and the consequent term is the power quality phenomenon or problem directly or indirectly caused by that factor. The causal relationship must satisfy the condition that "the causal relationship can be verified experimentally or theoretically." This relationship typically appears in statements analyzing power quality phenomena and changes in physical quantities, with the sentence structure "A is the cause of B" or "B is caused by A." A is used to analyze the main cause of phenomenon B, and a causal relationship exists between A and B. For example, if A = nonlinear load, then B = {harmonics, waveform distortion, total harmonic distortion exceeding the standard}.

[0068] Based on the established set of relational functions for power quality terms, ontology units can be constructed, which in simple form are mappings from A to B. For example, for terms in attribute relations, an ontology unit can be constructed using branches for "terms" and "feature sets," and its relational functions are... This ontology unit can be interpreted as a precise quantitative or qualitative description of the technical features of a term by elements in the feature set.

[0069] Table 2 Body Unit Types

[0070] S5: Establish a knowledge ontology class architecture for the power quality domain.

[0071] In this step, the system includes basic measurement parameter classes and power quality event classes, and defines corresponding subclass systems for each class. The basic measurement parameter class contains the basic parameters monitored by the monitoring points, and the power quality event class contains possible power quality problems. The construction of the domain ontology mainly involves extracting relevant concepts from the power quality domain, establishing a conceptual system and a set of relationships between concepts in the power quality domain, and describing the knowledge system in this domain using formal language, thereby establishing a domain ontology model.

[0072] Therefore, in summary, the power quality ontology mainly includes two ontology classes: basic measurement parameter class and power quality event class. The basic measurement parameter class contains the basic parameters monitored by the monitoring points, while the power quality event class contains possible power quality problems.

[0073] The basic measurement parameter categories include voltage amplitude, current amplitude, frequency, power, flicker, harmonics, voltage imbalance, and interharmonics. Each subcategory also contains further subcategories, such as... Figure 2 As shown. Power quality events include voltage sags, voltage swells, voltage interruptions, flicker, voltage imbalance, frequency deviations, voltage harmonics, current harmonics, transients, and other power quality events, such as... Figure 3 As shown.

[0074] S6: Based on the aforementioned knowledge ontology class architecture, build a power quality knowledge ontology model framework that integrates multi-dimensional information.

[0075] In this step, an ontology model framework is built based on the defined power quality knowledge ontology class. Because power grid construction and operation suffer from fragmentation, and power quality ontology serves as an important knowledge base, it must be able to reorganize power quality phenomena, indicators, and other knowledge in a reasonable way, transforming Chinese texts such as standard documents into a clearly structured knowledge network, thus linking power quality technical information from different stages. To reflect the relevance and integrity of knowledge elements, the embodiment constructs a framework such as... Figure 4 The ontology model shown is used to integrate multi-dimensional information on power quality.

[0076] S7: Based on the key domain terms, semantic relationships, ontology units and ontology model framework, construct and store the power quality domain knowledge ontology model.

[0077] In this process, based on the previous text preprocessing, terminology screening, and ontology model framework construction, the overall steps for building the power quality knowledge ontology model can be obtained: First, the input power quality-related standard text is formally described as an ontology. Then, based on the constructed professional vocabulary, the text is preprocessed before word frequency calculation. Next, the RAKE method is used to screen terms in the text based on word frequency and co-occurrence frequency to obtain candidate words. By calculating word weights, it is determined whether the term weight exceeds a threshold; if it does, the term is extracted. If it does not exceed the threshold, the next term is evaluated. After multiple iterations of screening, a set of selected terms is obtained. Combined with the constructed terminology relation function set, ontology class system, and model framework, instantiation and filling are completed to form a power quality domain knowledge ontology model. Figure 5 This is a flowchart illustrating the steps involved in building a power quality knowledge ontology model.

[0078] S8: Receive query or analysis requests for the power quality domain, call the power quality domain knowledge ontology model for semantic processing, and output structured knowledge association results.

[0079] In this step, semantic processing and output of knowledge association results specifically include the following methods: Semantic retrieval: Receives query statements containing power quality phenomena, parameters, or device names, performs semantic matching in the knowledge ontology model, and returns term nodes and their relationship paths that are directly or indirectly related to the query content. Relationship tracing: Receives analysis requests for specific power quality events or measurement parameters, and traces forward or backward along the causal relationships, hierarchical relationships, or whole-part relationships defined in the knowledge ontology model to generate a knowledge chain containing causes, effects, and constituent elements; Reasoning and analysis: Based on the attribute functions and axiomatic constraints defined in the knowledge ontology model, the consistency verification or status determination of the input monitoring data or event description is performed, and diagnostic conclusions or governance suggestions are output. Results presentation: The semantic processing results are output as a knowledge subgraph, list, or structured document containing term nodes, relation edges, and attribute values.

[0080] This embodiment improves existing knowledge description methods by constructing a knowledge ontology model and leveraging information technology. The ontology model organizes and represents power quality knowledge in a structured and semantic manner, making the relationships between knowledge points clearer and more explicit.

[0081] Example 2 This embodiment provides a power quality knowledge processing system based on an ontology model; the power quality knowledge processing system includes: The formal description module is used to formally describe the knowledge ontology of the power quality domain using the six-tuple concept, in order to capture the structured knowledge system of this domain. The text preprocessing and vocabulary building module is used to preprocess standard texts in the field of power quality, extract the main text, perform word segmentation using word segmentation tools, introduce professional terms from standard texts in the field of power quality, and build a professional vocabulary list according to parts of speech. The terminology extraction and filtering module is used to extract candidate terms based on preprocessed text, and to use a fast automatic keyword extraction method to calculate the weight of each candidate term based on word frequency and word co-occurrence frequency, and to sort them from high to low weight to filter out key domain terms. The semantic relationship modeling module is used to mine the semantic relationships between the key domain terms, define a set of term relationship functions, and establish an ontology unit representing a single semantic relationship based on this. The ontology class system construction module is used to establish a knowledge ontology class system in the field of power quality. The system includes two core classes: basic measurement parameter class and power quality event class, and defines corresponding subclass systems for each class. The model framework building module is used to build a power quality knowledge ontology model framework that can integrate multi-dimensional information based on the basic measurement parameter class and power quality event class. The model generation and storage module is used to integrate the processing results of the formal description module, text preprocessing and vocabulary construction module, term extraction and screening module, semantic relationship modeling module, ontology class system construction module and model framework construction module, instantiate and generate a computable and storable power quality domain knowledge ontology model, and perform persistent storage. The knowledge service interface module is used to receive query or analysis requests for the power quality domain, call the stored power quality domain knowledge ontology model for semantic processing, and output structured knowledge association results.

[0082] The system's functional modules can be divided according to business processes into a front-end knowledge processing layer, a mid-end model building layer, and a back-end knowledge service layer. The front-end knowledge processing layer includes a formal description module, a text preprocessing and vocabulary building module, and a terminology extraction and filtering module, responsible for knowledge extraction and modeling. The mid-end model building layer includes a semantic relationship modeling module, an ontology class system construction module, a model framework building module, and a model generation and storage module, responsible for model framework building and instantiation storage. The back-end knowledge service layer is a knowledge service interface module, responsible for knowledge service output. Modules can collaborate through standardized input / output interfaces, supporting both fully automated sequential operation and allowing independent upgrades or replacements for each module. The system encapsulates methods into directly deployable software entities, providing knowledge enhancement capabilities for business systems such as power quality monitoring, analysis, and governance.

[0083] In specific implementation, the working principle, control process and technical effects of the power quality knowledge processing system based on ontology model provided in this embodiment of the invention are the same as those of the power quality knowledge processing method based on ontology model in the foregoing embodiment, and will not be repeated here.

[0084] The above specific embodiments have demonstrated the substantial features and progress of the present invention. Equivalent modifications can be made to them according to actual usage needs, based on the guidance of the present invention, and all such modifications are within the scope of protection of this solution.

Claims

1. A method for power quality knowledge processing based on ontology model, characterized in that Includes the following steps: 1) To formally describe the knowledge ontology in the field of power quality; 2) Preprocess standard texts in the field of power quality and construct a professional glossary by combining professional terminology; 3) Extract candidate terms from the preprocessed text and filter key domain terms based on word frequency and co-occurrence relationships; 4) Mine the semantic relationships between the key domain terms and establish ontology units that represent these semantic relationships; 5) Establish a knowledge ontology class structure for the power quality domain; 6) Based on the aforementioned knowledge ontology class architecture, build a power quality knowledge ontology model framework that integrates multi-dimensional information; 7) Based on the key domain terms, semantic relationships, ontology units, and ontology model framework, construct and store a knowledge ontology model for the power quality domain; 8) Receive query or analysis requests, call the power quality domain knowledge ontology model for semantic processing, and output knowledge association results.

2. The method of claim 1, wherein the method is based on an ontology model. In step 1), the concept of six tuples is used to formally describe the knowledge ontology in the field of power quality. The six tuples include a set of concepts, a set of attributes, a set of relations, a set of relation attributes, a classification system, and a set of axioms.

3. The method of claim 1, wherein the method further comprises: Step 2) describes the construction of a professional vocabulary list, which includes: introducing professional terms from standard texts in the field of power quality, classifying and integrating them according to the parts of speech of nouns, verbs, and adjectives, expanding the vocabulary database, and generating a professional vocabulary list containing power quality professional terms and their word frequency statistics; among them, the noun category includes voltage sag, frequency deviation, harmonics, transient oscillation, and flicker coefficient; the verb category includes reactive power compensation, harmonic suppression, flicker value testing, and waveform distortion elimination; and the adjective category includes active, passive, load-related, zero-sequence, and non-periodic.

4. The method of claim 1, wherein the method is based on an ontology model. Step 3) Select key domain terms, including: 301) Extraction of candidate terms: Based on the preprocessed text, the text is segmented into several candidate keyword phrases using stop words and punctuation marks; each candidate keyword phrase is treated as a candidate term, and each word contained in each candidate keyword phrase is treated as a candidate word; 302) Calculation of word frequency: The word frequency of any candidate word is equal to the number of times that candidate word appears in the text divided by the sum of the number of times all words appear in the text; 303) Calculation of word weight: The word weight of any candidate word is equal to the sum of the number of times the candidate word co-occurs with all its co-occurring words; wherein, the co-occurring words refer to other words that co-occur with the candidate word in the same candidate keyword phrase, and the number of co-occurrences refers to the number of times the candidate word co-occurs with a certain co-occurring word in the same candidate keyword phrase; 304) Calculation of phrase weight: The weight of any candidate term is equal to the phrase weight of the candidate keyword phrase corresponding to the candidate term, and the phrase weight is equal to the sum of the word weights of all candidate words in the candidate keyword phrase; 305) Screening of key domain terms: Sort all candidate terms from high to low according to phrase weight, and extract candidate terms whose phrase weight is greater than or equal to the threshold as key domain terms.

5. The method of claim 1, wherein the method further comprises: The terminology relation function set mentioned in step 4) includes: Superordinate and subordinate functions: used to represent the hierarchical inclusion relationship between terms; where the superordinate term is an abstract generalization of a certain type of thing, and the subordinate term is a specific category with specific attributes within that type of thing. The subordinate term inherits the core attributes of the superordinate term and has its own unique characteristics. Whole-part function: used to express the physical or functional relationship between the things referred to by the term; the whole term is a system, device or collection of concepts with independent functions, and the part terms are indispensable components of the whole term; Attribute functions: used to represent the correspondence between a term and its inherent characteristics, quantification parameters, or descriptive dimensions; attribute terms are precise quantitative or qualitative descriptions of the technical characteristics of the subject term. Causal function: used to represent the cause-and-effect logical relationship between terms. The cause term is the triggering factor that leads to power quality abnormalities, and the result term is the power quality phenomenon or problem that occurs. Based on the set of relational functions, terms and their associated elements are mapped to structured ontology units.

6. The method of claim 1, wherein the method is based on an ontology model. The knowledge ontology system described in step 5) specifically includes: The basic measurement parameter class includes at least one of the following subclasses: voltage amplitude, current amplitude, frequency, power, flicker, harmonics, voltage imbalance, and interharmonics; each subclass can further include its subordinate subclasses. Power quality events include at least one of the following subclasses: voltage sag, voltage swell, voltage interruption, flicker, voltage imbalance, frequency deviation, voltage harmonics, current harmonics, and transients.

7. The power quality knowledge processing method based on ontology model according to claim 1, characterized in that: The ontology model framework in step 6) is based on the basic measurement parameter class and power quality event class and their subclass system. The term relation function set is used to graphically organize class nodes, term nodes and their semantic relationships to form a multi-dimensional power quality information fusion framework characterized by the correlation and integrity of knowledge elements.

8. The power quality knowledge processing method based on ontology model according to claim 1, characterized in that: Step 7) involves constructing and storing a knowledge ontology model for the power quality domain, which specifically includes the following sub-steps: 701) Using the input power quality-related standard text as a knowledge source, execute steps 1) to 2) to obtain a formal description framework and a glossary of professional terms; 702) Based on step 3), iteratively filter key domain terms from the preprocessed text until term extraction of all text is completed; 703) Semantically map and instantiate the selected set of key domain terms, the set of relational functions defined in step 4), and the ontology class system established in step 5; 704) Persistently store the instantiated model using a graph database or ontology storage language to form a power quality domain knowledge ontology model that can be called.

9. A power quality knowledge processing system based on an ontology model, characterized in that: The system employs a power quality knowledge processing method based on an ontology model as described in any one of claims 1-8, wherein the system comprises: The formal description module is used to formally describe knowledge ontology in the field of power quality; The text preprocessing and vocabulary building module is used to preprocess standard texts in the field of power quality and build a professional vocabulary list by combining professional terms. The terminology extraction and filtering module is used to extract candidate terms from preprocessed text and filter key domain terms based on word frequency and word co-occurrence relationships. The semantic relationship modeling module is used to mine the semantic relationships between the key domain terms and establish ontology units that represent the semantic relationships. The ontology class system construction module is used to establish the knowledge ontology class system architecture in the power quality domain; The model framework building module is used to build a power quality knowledge ontology model framework that integrates multi-dimensional information based on the knowledge ontology class architecture. The model generation and storage module is used to construct and store a knowledge ontology model of the power quality domain based on the key domain terms, semantic relationships, ontology units and ontology model framework. The knowledge service interface module is used to receive query or analysis requests, call the knowledge ontology model in the power quality domain for semantic processing, and output knowledge association results.

10. A power quality knowledge processing system based on an ontology model according to claim 9, characterized in that: The terminology extraction and filtering module further includes: The candidate term extraction unit is used to segment the preprocessed text into several candidate keyword phrases using stop words and punctuation marks, treat each candidate keyword phrase as a candidate term, and treat each word contained in each candidate keyword phrase as a candidate word. The word frequency calculation unit is used to calculate the word frequency of any candidate word, wherein the word frequency is equal to the number of times the candidate word appears in the text divided by the sum of the number of times all words appear in the text; The word weight calculation unit is used to calculate the word weight of any candidate word, wherein the word weight is equal to the sum of the number of times the candidate word co-occurs with all its co-occurring words; The phrase weight calculation unit is used to calculate the weight of any candidate term. The weight is equal to the phrase weight of the candidate keyword phrase corresponding to the candidate term. The phrase weight is equal to the sum of the word weights of all candidate words in the candidate keyword phrase. The key term filtering unit is used to sort all candidate terms from high to low according to their phrase weights, and extract candidate terms whose phrase weights are greater than or equal to the threshold as key domain terms by setting a preset threshold.

Images