Electrical equipment defect grading knowledge graph construction method
By building a knowledge graph for power equipment defect grading, integrating data source information and using deep learning, the inaccuracy and safety hazards of manual inspection in power equipment defect management are solved, the accuracy and work efficiency of fault diagnosis are improved, and the guarantee of guaranteed fault diagnosis is improved. The equipment is safe and stable.
Patent Information
- Application Number
- CN202410947943.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-16
- Publication Date
- 2024-11-12
Smart Images

Figure CN118940824A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power equipment defect management, and in particular to a method for constructing a knowledge graph for power equipment defect grading. Background Art
[0002] In recent years, the knowledge graph concept and graph model knowledge expression method proposed by Google have similarities with the physical network attributes of the power grid itself and the core business scenarios of intelligent applications. The application of knowledge graphs in the power grid field has many advantages. At present, many scholars have carried out exploratory research and engineering practice work on knowledge graphs in the field of power systems. In the existing technology, Pu Tianjiao et al. systematically studied the application architecture design and related application practices of knowledge graphs in the power field from the dual perspectives of science and engineering, providing effective cross-media organization, management and cognitive capabilities for the power industry chain. In the field of power grid dispatching, Qiao Ji et al. studied and proposed a knowledge graph framework for power grid dispatching fault processing, designed 5 levels and 4 types of domain graphs for knowledge graph construction and application, and gave decomposition ideas and technical solutions; Guo Rong et al. took the power grid fault handling plan text as the object, and further studied the top-down and bottom-up combined power grid fault handling knowledge graph construction method, solving the problem of knowledge extraction in the power field; Zhang Rui et al. proposed based on transfer learning technology. The entity recognition technology of text data for power grid fault handling has studied the construction method of knowledge graph for power grid fault handling, and realized decision support for multiple fault types; Yu Jianming and others have studied the construction method of knowledge graph in the field of intelligent power grid dispatching, and proposed knowledge graph application solutions for application scenarios such as electronic rules, fault handling, switching operation, and dialogue question and answer in response to actual dispatching needs, which has engineering reference and practical significance; Zhang Siyuan and others have studied the production rule system and knowledge model systematic modeling method suitable for the expression of stable regulations in response to the characteristics and complexity of stable operation regulations of power grids, and proposed a digital realization technology route for the knowledge model of stable regulations based on the digital twin model of power grids, and developed a digital application system. In addition, many knowledge graph research works have been carried out in the aspects of auxiliary decision-making for distribution network faults, intelligent recommendation for demand response, and health management of power equipment, which have promoted the deep integration and innovative application of knowledge graphs in power grid business scenarios.
[0003] There are few studies on knowledge graphs in areas such as power equipment defect classification. Under the current power grid demand, the problem of power equipment defects has received enough attention. However, the traditional way of managing power equipment defects mainly relies on manual on-site inspections and verbal communication, which is then recorded in paper forms, and then a rectification plan is formulated and notified to relevant departments. This method has many problems, including safety hazards, low information accuracy, and low efficiency. Therefore, it is very important to conduct research on knowledge graphs for power equipment defects. By constructing corresponding knowledge graphs, power equipment can be managed more effectively, and information such as safety hazards can be conveyed to on-site staff in advance. This is also the direction that power equipment management urgently needs to develop in the future. Establishing a knowledge graph for power equipment defect classification can effectively reduce the inaccuracy and safety risks of manual inspections, improve the accuracy of data recording and analysis, and further ensure the safe and stable operation of power equipment.
[0004] Therefore, it is an urgent problem for technical personnel in this field to propose a method for constructing a knowledge graph for defect classification of power equipment to solve the problems of safety hazards, low information accuracy and low efficiency in the existing technology. Summary of the invention
[0005] In view of this, the present invention provides a method for constructing a knowledge graph for defect classification of power equipment, which can quickly locate and find information about related equipment defects, improve work efficiency, and reduce the difficulty of operation and maintenance.
[0006] In order to achieve the above object, the present invention adopts the following technical solution:
[0007] A method for constructing a knowledge graph for defect classification of electric power equipment comprises the following steps:
[0008] Data acquisition step: acquiring and organizing power equipment defect data and constructing a power equipment defect data set;
[0009] Data preprocessing step: preprocessing the power equipment defect data set to obtain a preprocessed data set;
[0010] Extraction step: perform entity extraction and relationship extraction on the preprocessed data set to obtain power equipment defect entities and power equipment defect relationships;
[0011] Entity relationship construction step: construct the power equipment defect triple through the power equipment defect entity and the power equipment defect relationship;
[0012] Knowledge graph construction steps: store power equipment defect triplets and display them visually to build a power equipment defect knowledge graph.
[0013] In the above method, optionally, the data preprocessing steps include data cleaning, word segmentation and part-of-speech tagging.
[0014] In the above method, optionally, entity extraction in the extraction step includes BIO labeling and entity recognition.
[0015] In the above method, optionally, the BIO annotation includes:
[0016] Data parsing: parse text and corresponding label sequences from the preprocessed dataset;
[0017] Data structuring: convert the parsed text and label sequences into model data structures;
[0018] Entity extraction: According to model requirements, the text sequence and the corresponding label sequence are constructed into the model input format.
[0019] In the above method, optionally, entity recognition uses the BERT-BiGRU-CRF model, which is expressed as follows:
[0020]
[0021] Among them, Q, K, and V represent query vector, key vector, and value vector respectively. k represents the input vector dimension, and T represents the transposed matrix.
[0022] In the above method, optionally, the relation extraction in the extraction step is to construct a relation extraction method based on artificial rules, and the expression is as follows:
[0023] h sent =f(h mask ) = f(ASP-GCN(x)),
[0024] Among them, h mask represents all hidden representations masked, and f:Rd×n→Rd×1 is a maximum pooling function.
[0025] In the above method, optionally, the artificial rule-based relationship extraction method adopts a segmented maximum pooling method to obtain feature information.
[0026] It can be seen from the above technical solutions that, compared with the prior art, the present invention provides a method for constructing a knowledge graph for defect classification of power equipment, which has the following beneficial effects: 1) The knowledge graph of the present invention integrates information from different data sources, including historical maintenance records, sensor data, etc. Through deep learning and pattern recognition, the technology can improve the accuracy of fault diagnosis and quickly determine the root cause of the problem; 2) By constructing a knowledge graph, the present invention can better understand the correlation and regularity between data, so as to conduct in-depth data analysis and mining; 3) Through the knowledge graph technology, the present invention allows operation and maintenance personnel to locate problems more quickly, understand equipment history and maintenance status, thereby reducing the difficulty of operation and maintenance. This is especially important for training new operation and maintenance personnel or dealing with situations where equipment complexity increases; 4) Through the knowledge graph, the present invention can quickly locate and find information on related equipment defects, which provides great help to on-site staff and greatly improves work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0028] Figure 1 A flow chart of a method for constructing a knowledge graph for defect grading of electric power equipment disclosed in the present invention;
[0029] Figure 2 It is the BIO annotation diagram disclosed in the present invention;
[0030] Figure 3 A diagram of a defect entity recognition model for electric power equipment disclosed in the present invention;
[0031] Figure 4 It is a triplet diagram of defects of electric power equipment disclosed in the present invention;
[0032] Figure 5 A knowledge graph of defects in electric power equipment disclosed in the present invention;
[0033] Figure 6 This is a knowledge graph query case diagram disclosed in an embodiment of the present invention. DETAILED DESCRIPTION
[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0035] In this application, relational terms such as first and second, etc. are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of more restrictions, the elements defined by the sentence "comprise one..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.
[0036] Reference Figure 1 As shown, the present invention discloses a method for constructing a knowledge graph for defect grading of electric power equipment, comprising the following steps:
[0037] Data acquisition step: acquiring and organizing power equipment defect data and constructing a power equipment defect data set;
[0038] Data preprocessing step: preprocessing the power equipment defect data set to obtain a preprocessed data set;
[0039] Extraction step: perform entity extraction and relationship extraction on the preprocessed data set to obtain power equipment defect entities and power equipment defect relationships;
[0040] Entity relationship construction step: construct the power equipment defect triple through the power equipment defect entity and the power equipment defect relationship;
[0041] Knowledge graph construction steps: store power equipment defect triplets and display them visually to build a power equipment defect knowledge graph.
[0042] Furthermore, the data preprocessing steps include data cleaning, word segmentation and part-of-speech tagging.
[0043] Furthermore, data cleaning includes: removing noise, such as irrelevant characters, HTML tags, and URLs; normalization, converting to lowercase, unifying numbers, and extracting stems; filtering stop words to remove meaningless words; and 4) spelling correction to ensure accurate information.
[0044] Word segmentation includes: using tools such as jieba to perform precise or fuzzy pattern word segmentation.
[0045] Part-of-speech tagging includes: assigning a part-of-speech label to each word, which helps in subsequent filtering and understanding of semantic relationships.
[0046] Specifically, since defective texts are different from general Chinese texts, it is necessary to analyze the characteristics of defective texts before data analysis in order to adopt more appropriate text preprocessing methods. The text scenarios include:
[0047] 1) Synonym / similar word phenomenon. Defect texts are currently mainly written manually, and the same power equipment defect can be described by multiple words. For example, "oil" and "oil storage cabinet", "disconnector" and "knife switch", "respirator" and "dehumidifier", etc. In this case, similar word processing is required during preprocessing, otherwise the model will identify them as two completely different features, thus affecting the model effect.
[0048] 2) The length of the text varies greatly. Due to the different levels of detail in describing equipment defects, the length of each defect text varies greatly. The short text has about 10 words, and the long text has hundreds of words. Short texts may have vague defect descriptions, and long texts will make it difficult for operation and maintenance personnel to accurately and quickly extract defect information, affecting the efficiency of operation and maintenance.
[0049] 3) Semantic missing. According to the defect terminology standard, the main part of a complete defect description should include: component, component type, location, defect location, and defect condition. However, most actual defect texts are highly colloquial and contain some elements or semantic missing, resulting in vague defect descriptions. For example, "slight oil leakage (no oil droplets are seen but there is oil leakage)" does not specifically specify the subject, that is, the defective component. This phenomenon will directly affect the subsequent equipment defect analysis work.
[0050] Furthermore, entity extraction in the extraction step includes BIO annotation and entity recognition.
[0051] Further, see Figure 2 As shown, BIO annotation includes:
[0052] Data parsing: parse text and corresponding label sequences from the preprocessed dataset;
[0053] Data structuring: convert the parsed text and label sequences into model data structures;
[0054] Entity extraction: According to model requirements, the text sequence and the corresponding label sequence are constructed into the model input format.
[0055] Further, see Figure 3As shown, entity recognition uses the BERT-BiGRU-CRF model, and the expression is as follows:
[0056]
[0057] Among them, Q, K, and V represent query vector, key vector, and value vector respectively. k represents the input vector dimension, and T represents the transposed matrix.
[0058] Specifically, Q, K, and V will generate three new variables for each input vector. The query vector can pay attention to other words when encoding the current word. The key vector and the value vector are a pair of key-value pairs. The key vector is the index, and the value vector is the actual content of the word.
[0059] Furthermore, the relation extraction in the extraction step is to construct a relation extraction method based on artificial rules, and the expression is as follows:
[0060] h sent =f(h mask ) = f(ASP-GCN(x)),
[0061] Among them, h mask represents all hidden representations masked, and f: Rd×n→Rd×1 is a maximum pooling function.
[0062] Furthermore, f: Rd×n→Rd×1 maps n output vectors to 1 sentence vector
[0063] Furthermore, the artificial rule-based relationship extraction method uses a segmented maximum pooling method to obtain feature information.
[0064] Specifically, due to the problem of overlapping relations caused by the abstraction of power equipment defect language, the traditional pooling method cannot obtain more feature information in a refined manner. In order to improve efficiency, the segmented maximum pooling method is adopted. This method divides the sentence into three segments according to the positions of the two target entities, namely, the beginning of the sentence-entity 1, entity 1-entity 2, and entity 2-end of the sentence. Pooling operation is performed on each segment to improve the semantic representation ability. The expression after the segmented maximum pooling operation is as follows:
[0065] d ij =max(c ij ),
[0066] c ij Indicates the three segments after segmentation, namely {c i1 , c i2 , c i3}. Where 1≤i≤n, 1≤j≤3, after calculating the pooling, the result vector d is obtainedi ={d i1 , d i2 , d i3}, and finally, all the results of d are calculated to get the output vector representation, which is expressed as follows:
[0067] s = tanh(d 1:n ),
[0068] s represents the output vector, s∈R 3n .
[0069] Further, see Figure 4 and Figure 5 As shown in the figure, the Protégé ontology construction tool is used to complete the construction of the relationship between the power equipment defect entities, and Neo4j is used to store and visualize the acquired triple knowledge to complete the construction of the power equipment defect knowledge graph.
[0070] In a specific embodiment, a test analysis is performed by taking "What types of defects exist in the intermediate relay?" as an example. The named entity in the question sentence is "intermediate relay" and the intention of the question sentence is "defect type". The test result of the question sentence analysis for "What types of defects exist in the intermediate relay?" is as follows: Figure 6 shown.
[0071] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without creative work.
[0072] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for constructing a knowledge graph for defect grading of electric power equipment, characterized in that: The following steps are involved: Data acquisition step: acquiring and organizing power equipment defect data and constructing a power equipment defect data set; Data preprocessing step: preprocessing the power equipment defect data set to obtain a preprocessed data set; Extraction step: perform entity extraction and relationship extraction on the preprocessed data set to obtain power equipment defect entities and power equipment defect relationships; Entity relationship construction step: construct the power equipment defect triple through the power equipment defect entity and the power equipment defect relationship; Knowledge graph construction steps: store power equipment defect triplets and display them visually to build a power equipment defect knowledge graph.
2. A method for constructing a knowledge graph for defect classification of electric power equipment according to claim 1, characterized in that: The data preprocessing steps include data cleaning, word segmentation and part-of-speech tagging.
3. The method for constructing a knowledge graph for defect classification of electric power equipment according to claim 1, characterized in that: Entity extraction in the extraction step includes BIO annotation and entity recognition.
4. A method for constructing a knowledge graph for defect classification of electric power equipment according to claim 3, characterized in that: BIO labeling includes: Data parsing: parse text and corresponding label sequences from the preprocessed dataset; Data structuring: convert the parsed text and label sequences into model data structures; Entity extraction: According to model requirements, the text sequence and the corresponding label sequence are constructed into the model input format.
5. The method for constructing a knowledge graph for defect classification of electric power equipment according to claim 3, characterized in that: Entity recognition uses the BERT-BiGRU-CRF model, which is expressed as follows: Among them, Q, K, and V represent query vector, key vector, and value vector respectively. k represents the input vector dimension, and T represents the transposed matrix.
6. A method for constructing a knowledge graph for defect classification of electric power equipment according to claim 1, characterized in that: In the extraction step, the relation extraction method is constructed based on artificial rules. The expression is as follows: h sent =f(h mask )=f(ASP-GCN(x)), Among them, h mask represents all hidden representations masked, ASP-GCN(x) is the feature obtained by transforming the attention mechanism with the graph convolutional network, and f:Rd×n→Rd×1 is a maximum pooling function.
7. A method for constructing a knowledge graph for defect classification of electric power equipment according to claim 6, characterized in that: The relationship extraction method based on artificial rules uses the segmented maximum pooling method to obtain feature information.
Citation Information
Patent Citations
Power equipment fault knowledge graph construction method
CN111737496A
Construction method and system of equipment maintenance disassembly and assembly process knowledge graph
CN117194671A
