Knowledge base-based power operation and maintenance scheduling automation operation entity extraction method and system

Through the knowledge base-based power operation and maintenance scheduling automation operation entity extraction method, combined with multimodal data extraction, similarity calculation and LDA algorithm, a high-quality power scheduling automation text knowledge base is built, and the poor physical extraction accuracy in the field of power automation is solved. The problem has achieved higher extraction accuracy and potential law analysis.

CN118917586AActive Publication Date: 2024-11-08STATE GRID FUJIAN ELECTRIC POWER CO LTD +1
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202410938467.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-13
Publication Date
2024-11-08
Estimated Expiration
2044-07-13

AI Technical Summary

Technical Problem

The prior art cannot effectively meet the needs of fault diagnosis and data analysis in large-scale power scheduling automation operations, especially in the field of power automation, which is poor in the extraction accuracy of entity.

Method used

Using the knowledge base-based power operation and maintenance scheduling automated operation entity extraction method and system, including multimodal data extraction module, text extraction module based on similarity calculation, and entity labeling module based on LDA algorithm, it is carried out through the BERT-BiLSTM-CRF model. Entity extraction to build a high-quality power scheduling automated operation text knowledge base.

Benefits of technology

The accuracy of entity extraction in the field of power automation is improved, and the potential laws of multimodal power scheduling automation data can be better analyzed, and the potential entity characteristics in the text can be deeply explored.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118917586A_ABST
    Figure CN118917586A_ABST
Patent Text Reader

Abstract

The invention provides a knowledge base-based power operation and maintenance scheduling automation operation entity extraction method and system, which are used for extracting power scheduling automation operation entities and mining a potential rule of generating multi-modal data in the field of scheduling automation. Comprising a multi-modal data extraction module, a text extraction module based on similarity calculation and an entity labeling module based on an LDA algorithm. The text extraction module comprises a multi-modal data extraction module, and is used for preliminarily extracting text information by using different toolkits according to multi-modal data characteristics, and then obtaining high-quality text data by comprehensively considering similarity calculation methods such as editing distance and cosine similarity; the entity labeling module is used for extracting key entities in the text by using an LDA algorithm and labeling and constructing a power dispatching automation operation text knowledge base; according to the method, the potential rule of the multi-modal power dispatching automation data can be better analyzed, the potential entity features in the power dispatching automation operation text are deeply mined and marked, and then the entity extraction accuracy of the power automation field is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of power grid dispatching automation, and in particular to a method and system for extracting power operation and maintenance dispatching automation operation entities based on a knowledge base. Background Art

[0002] In the long-term process of automated dispatching of power networks, a lot of structured and unstructured data has accumulated. Nowadays, traditional manual data analysis methods cannot meet the fault diagnosis and data analysis needs of large-scale power dispatching automation operations under the background of big data. The use of knowledge graph technology can effectively solve the problem of knowledge expression in the field of power automation. Entity extraction is a key link in building knowledge graphs. In view of the poor accuracy of entity extraction in the field of power automation, it is particularly important to design a knowledge base to assist entity extraction. The knowledge base can effectively improve the accuracy of entity extraction. Summary of the invention

[0003] The present invention proposes a method and system for extracting entities from power operation and maintenance dispatching automation operations based on a knowledge base, which can better analyze the potential laws of multimodal power dispatching automation data, deeply explore and annotate potential entity features in power dispatching automation operation texts, and thus improve the accuracy of entity extraction in the field of power automation.

[0004] The present invention adopts the following technical solutions.

[0005] The knowledge base-based electric power dispatching automation entity extraction system is used for electric power dispatching automation operation entity extraction and mining the potential rules of multimodal data generated in the field of electric power dispatching automation. It includes a multimodal data extraction module, a text extraction module based on similarity calculation and an entity annotation module based on the LDA algorithm. The text extraction module includes a multimodal data extraction module, which uses different toolkits to preliminarily extract text information according to the characteristics of multimodal data, and then comprehensively considers similarity calculation methods such as edit distance and cosine similarity to obtain high-quality text data. The entity annotation module uses the LDA algorithm to extract key entities in the text and annotate them to build a text knowledge base for electric power dispatching automation operations.

[0006] The method for extracting entities of power dispatching automation based on knowledge base adopts the power dispatching automation entity extraction system based on knowledge base, including the following steps:

[0007] Step S1, extracting text features from multimodal power dispatch automation data;

[0008] Step S2, reducing text redundancy by calculating text similarity to obtain high-quality text data;

[0009] Step S3: Design an entity extraction module based on LDA to build a knowledge base by accurately annotating entities;

[0010] Step S4: Based on the knowledge base, use the BERT-BiLSTM-CRF model to extract entities from power dispatch automation data and obtain entities in the text.

[0011] In step S1, text feature extraction is performed on multimodal power dispatch automation data, which specifically includes the following steps:

[0012] Step S1A1: For data of text types such as word and txt, use the python-docx library to extract and classify text according to the structural format of the text;

[0013] Step S1A2: For data of PDF type such as ledger information and transportation manual, use OCR technology to extract text information from PDF files;

[0014] Step S1A3: For data in the Excel table type of work tickets and safety regulations, use the pandas library to extract key text;

[0015] Step S1A4: reconstruct different types of text using a text segmentation method with a threshold, segment text with a length greater than an upper threshold, and remove text with a length less than a lower threshold;

[0016] Step S1A5: After steps S1A1, S1A2, S1A3, and S1A4, the initially extracted power dispatching automation text data is obtained.

[0017] In step S2, the text redundancy is reduced by calculating the text similarity to obtain high-quality text data. The specific steps include:

[0018] Step S2A1, setting a set of weight parameters λ=(λ1, λ2, λ3), which respectively represent the weights of edit distance, cosine similarity, and TF-IDF; setting a threshold Ω, which represents the threshold of text similarity;

[0019] Step S2A2, manually extract a small number of samples of different types and establish a sample knowledge base D;

[0020] Step S2A3, by calculating the edit distance between the text data and the knowledge base D, the edit distance similarity S1 is obtained; Step S2A4, by calculating the cosine similarity between the text data and the knowledge base D, the cosine similarity S2 is obtained; Step S2A5, by calculating the TF-IDF value of the text data and the knowledge base D, the TF-IDF similarity S3 is obtained; Step S2A6, by weighted calculation of the similarity, the text similarity score S is obtained;

[0021] Step S2A7, comparing the similarity score S with the threshold, if it is greater than the threshold Ω, it means deleting the text; if it is less than or equal to the threshold, the text is added to the knowledge base D;

[0022] Step S2A8: Repeat steps S2A3-S2A7 to build a complete entity text knowledge base D.

[0023] In step S2A3, the specific calculation of the edit distance and the edit distance similarity is as follows:

[0024]

[0025]

[0026] In the formula, is the indicator function, when a i =b j When the value is 0, the value is 1 at other times; a,b (i,j) represents the edit distance between the first i characters of a and the first j characters of b; |a|, |b| represent the length of text a and text b, and similarity represents the edit distance similarity;

[0027] In step S2A4, the specific calculation of cosine similarity is as follows:

[0028]

[0029] In the formula, A i , B i Represents the values ​​of each dimension of vector A and B, It means the inner product of vectors A and B. represents the product of A and B modulo;

[0030] In step S2A5, the specific calculation of TF-IDF is as follows:

[0031]

[0032]

[0033] TF-IDF=TF·IDF

[0034] Where n i,j Indicates the term t i In the document d j The number of occurrences in TF i,j Indicates the term t i In the document d j The frequency of occurrence in. |D| represents the number of all documents, |j:t i ∈d j | indicates that the term is included iThe number of documents; in step S2A6, the weighted calculation of similarity is specifically calculated as follows:

[0035] S=λ1S1+λ2S2+λ3S3

[0036] Where λ1, λ2, λ3 are weight parameters set in S2A1, and S1, S2, S3 are similarities calculated in steps S2A3, S2A4, and S2A5.

[0037] In step S3, an entity extraction module based on LDA is designed to accurately annotate entities and realize knowledge base construction, which specifically includes the following steps:

[0038] Step S3A1: Use the LDA algorithm to extract text keywords from the knowledge base D.

[0039] Step S3A2: label the entities extracted using the LDA algorithm according to the actual situation and construct an entity extraction knowledge base D.

[0040] In step S3A1, the specific calculation of the LDA algorithm is as follows:

[0041]

[0042] Where P(W,Z) represents the file probability of a certain combination generated by LDA; P(θ j ,α) represents the probability distribution of the main body composition of each file; represents the composition probability distribution of each topic; P(Z j,t |θ j ) represents the probability multinomial distribution of different topics for a certain file; Represents the probability multinomial distribution of words for a topic.

[0043] In step S4, a BERT-BiLSTM-CRF entity extraction model is constructed, and the knowledge base is used to correct the labels when predicting entity labels, which specifically includes the following steps:

[0044] Step S4A1: Build a BERT-BiLSTM-CRF model to extract entities from text and perform label prediction;

[0045] Step S4A2: Use the knowledge base to correct the incorrect labels of the entities extracted in S4A1.

[0046] The BERT-BiLSTM-CRF model includes the BERT model, the BiLSTM model, and the CRF model;

[0047] In step S4A1, the specific formula of the BERT model is as follows:

[0048]

[0049] V=BERT([CLS],T1,...,T n ,[SEP])=(e1,...,e n )

[0050] Where, E={E1,E2,...,E n} represents the initial text sequence, T={T1,T2,...,T n} represents the text vector obtained after the initial text sequence is encoded by the transformer encoder.

[0051] In step S4A1, the specific formula of the BiLSTM model is as follows:

[0052] i t =σ(W ix X t +W ih h t-1 +b i )

[0053] f t =σ(W fx X t +W fh h t-1 +b f )

[0054] o t =σ(W ox X t +W oh h t-1 +b o )

[0055]

[0056]

[0057] h t =o t Θtanh(C t )

[0058]

[0059] In the formula, f t represents the state of the forget gate at time t, o t represents the state of the output gate at time t, represents the cell state at time t, h t represents the hidden layer state at time t;

[0060] In step S4A1, the specific formula of the CRF model is as follows:

[0061]

[0062] In the formula, N represents the number of training samples, C represents the regularization parameter, and y (i) represents the true label sequence of the i-th sample, x (i) represents the input sequence of the i-th sample, p(y (i) |x (i) ) represents the conditional probability distribution in the CRF model.

[0063] Compared with the prior art, the present invention has the following beneficial effects: better analysis of the potential rules of multimodal power dispatching automation data, in-depth mining and annotation of potential entity features in power dispatching automation texts, and thus improving the accuracy of entity extraction in the field of power automation. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0065] Attached Figure 1 It is a schematic diagram of the principle framework of the present invention for extracting entities in power dispatch automation;

[0066] Attached Figure 2 It is a schematic diagram of the principle framework of constructing a knowledge base in the present invention. DETAILED DESCRIPTION

[0067] As shown in the figure, the knowledge base-based power dispatching automation entity extraction system is used for power dispatching automation operation entity extraction and mining of potential rules of multimodal data generated in the field of power dispatching automation, including a multimodal data extraction module, a text extraction module based on similarity calculation and an entity annotation module based on the LDA algorithm; the text extraction module includes a multimodal data extraction module, which uses different toolkits to preliminarily extract text information according to the characteristics of multimodal data, and then comprehensively considers the edit distance and cosine similarity calculation methods to obtain high-quality text data; the entity annotation module uses the LDA algorithm to extract key entities in the text and annotate to build a power dispatching automation operation text knowledge base.

[0068] The knowledge-based power dispatching automation entity extraction method adopts the knowledge-based power dispatching automation entity extraction system. The overall framework of the method is shown in the attached Figure 1As shown. Specifically, it includes a multimodal data extraction module, a text extraction module, and an entity extraction module based on LDA. In view of the characteristics of multimodal data, different toolkits are used to achieve preliminary text extraction; thresholds are set and a variety of similarity calculation methods are used to reduce text redundancy and obtain high-quality text data; the LDA algorithm is used to extract text entities, and then the entities are accurately labeled to realize knowledge base construction;

[0069] The steps include:

[0070] Step S1, extracting text features from multimodal power dispatch automation data;

[0071] Step S2, reducing text redundancy by calculating text similarity to obtain high-quality text data;

[0072] Step S3: Design an entity extraction module based on LDA to build a knowledge base by accurately annotating entities;

[0073] Step S4: Based on the knowledge base, use the BERT-BiLSTM-CRF model to extract entities from power dispatch automation data and obtain entities in the text.

[0074] In step S1, text feature extraction is performed on multimodal power dispatch automation data, which specifically includes the following steps:

[0075] Step S1A1: For data of text types such as word and txt, use the python-docx library to extract and classify text according to the structural format of the text;

[0076] Step S1A2: For data of PDF type such as ledger information and transportation manual, use OCR technology to extract text information from PDF files;

[0077] Step S1A3: For data in the Excel table type of work tickets and safety regulations, use the pandas library to extract key text;

[0078] Step S1A4: reconstruct different types of text using a text segmentation method with a threshold, segment text with a length greater than an upper threshold, and remove text with a length less than a lower threshold;

[0079] Step S1A5: After steps S1A1, S1A2, S1A3, and S1A4, the initially extracted power dispatching automation text data is obtained.

[0080] In step S2, the text redundancy is reduced by calculating the text similarity to obtain high-quality text data. The specific steps include:

[0081] Step S2A1, setting a set of weight parameters λ=(λ1, λ2, λ3), which respectively represent the weights of edit distance, cosine similarity, and TF-IDF; setting a threshold Ω, which represents the threshold of text similarity;

[0082] Step S2A2, manually extract a small number of samples of different types and establish a sample knowledge base D;

[0083] Step S2A3, by calculating the edit distance between the text data and the knowledge base D, the edit distance similarity S1 is obtained; Step S2A4, by calculating the cosine similarity between the text data and the knowledge base D, the cosine similarity S2 is obtained; Step S2A5, by calculating the TF-IDF value of the text data and the knowledge base D, the TF-IDF similarity S3 is obtained; Step S2A6, by weighted calculation of the similarity, the text similarity score S is obtained;

[0084] Step S2A7, comparing the similarity score S with the threshold, if it is greater than the threshold Ω, it means deleting the text; if it is less than or equal to the threshold, the text is added to the knowledge base D;

[0085] Step S2A8: Repeat steps S2A3-S2A7 to build a complete entity text knowledge base D.

[0086] In step S2A3, the specific calculation of the edit distance and the edit distance similarity is as follows:

[0087]

[0088] In the formula, 1 (ai≠bj) is the indicator function, when a i =b j When the value is 0, the value is 1 at other times; a,b (i,j) represents the edit distance between the first i characters of a and the first j characters of b; |a|, |b| represent the length of text a and text b, and similarity represents the edit distance similarity;

[0089] In step S2A4, the specific calculation of cosine similarity is as follows:

[0090]

[0091] In the formula, A i , B i Represents the values ​​of each dimension of vector A and B, It means the inner product of vectors A and B. represents the product of A and B modulo;

[0092] In step S2A5, the specific calculation of TF-IDF is as follows:

[0093]

[0094] TF-IDF=TF·IDF

[0095] Where n i,j Indicates the term t i In the document d j The number of occurrences in TF i,j Indicates the term t i In the document d j The frequency of occurrence in. |D| represents the number of all documents, |j:t i ∈d j | indicates that the term is included i The number of documents;

[0096] In step S2A6, the weighted calculation of similarity is specifically calculated as follows:

[0097] S=λ1S1+λ2S2+λ3S3

[0098] Where λ1, λ2, λ3 are weight parameters set in S2A1, and S1, S2, S3 are similarities calculated in steps S2A3, S2A4, and S2A5.

[0099] In step S3, an entity extraction module based on LDA is designed to accurately annotate entities and realize knowledge base construction, which specifically includes the following steps:

[0100] Step S3A1: Use the LDA algorithm to extract text keywords from the knowledge base D.

[0101] Step S3A2: label the entities extracted using the LDA algorithm according to the actual situation and construct an entity extraction knowledge base D.

[0102] In step S3A1, the specific calculation of the LDA algorithm is as follows:

[0103]

[0104] Where P(W,Z) represents the file probability of a certain combination generated by LDA; P(θ j ,α) represents the probability distribution of the main body composition of each file; represents the composition probability distribution of each topic; P(Z j,t |θ j ) represents the probability multinomial distribution of different topics for a certain file; Represents the probability multinomial distribution of words for a topic.

[0105] In step S4, a BERT-BiLSTM-CRF entity extraction model is constructed, and the knowledge base is used to correct the labels when predicting entity labels, which specifically includes the following steps:

[0106] Step S4A1: Build a BERT-BiLSTM-CRF model to extract entities from text and perform label prediction;

[0107] Step S4A2: Use the knowledge base to correct the incorrect labels of the entities extracted in S4A1.

[0108] The BERT-BiLSTM-CRF model includes the BERT model, the BiLSTM model, and the CRF model;

[0109] In step S4A1, the specific formula of the BERT model is as follows:

[0110]

[0111] V=BERT([CLS],T1,...,T n ,[SEP])=(e1,...,e n )

[0112] Where, E={E1,E2,...,E n} represents the initial text sequence, T={T1,T2,...,T n} represents the text vector obtained after the initial text sequence is encoded by the transformer encoder.

[0113] In step S4A1, the specific formula of the BiLSTM model is as follows:

[0114] i t =σ(W ix X t +W ih h t-1 +b i )

[0115] f t =σ(W fx X t +W fh h t-1 +b f )

[0116] o t =σ(W ox X t +W oh h t-1 +b o )

[0117]

[0118] h t =o t Θtanh(C t )

[0119]

[0120] In the formula, f t represents the state of the forget gate at time t, o t represents the state of the output gate at time t, represents the cell state at time t, h t represents the hidden layer state at time t;

[0121] In step S4A1, the specific formula of the CRF model is as follows:

[0122]

[0123] In the formula, N represents the number of training samples, C represents the regularization parameter, and y (i) represents the true label sequence of the i-th sample, x (i) represents the input sequence of the i-th sample, p(y (i) |x (i) ) represents the conditional probability distribution in the CRF model.

[0124] This embodiment includes the following functional modules: a multimodal data extraction module, a text extraction module based on similarity calculation, an entity annotation module based on the LDA algorithm, and an entity extraction module based on a knowledge base and a BERT-BiLSTM-CRF model;

[0125] Among them, for the multimodal data of power dispatching automation, the traditional manual extraction and analysis method cannot cope with the massive data today. Using different toolkits to extract text data according to the characteristics of different modal data can reduce labor costs; the text extraction module based on similarity calculation can reduce text redundancy, explore the potential rules of text in power dispatching automation data, and obtain high-quality text; the entity extraction method of power dispatching automation operation text based on LDA algorithm can extract key fields in the text, so as to use sequence annotation to realize the construction of power dispatching automation operation entity extraction knowledge base and improve the accuracy of entity extraction.

[0126] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. The knowledge-based power dispatching automation entity extraction system is used to extract power dispatching automation operation entities and mine the potential laws of multimodal data generated in the field of power dispatching automation. It is characterized by: It includes a multimodal data extraction module, a text extraction module based on similarity calculation and an entity annotation module based on the LDA algorithm; the text extraction module includes a multimodal data extraction module, which uses different toolkits to preliminarily extract text information according to the characteristics of multimodal data, and then comprehensively considers the edit distance and cosine similarity calculation methods to obtain high-quality text data; the entity annotation module uses the LDA algorithm to extract key entities in the text and annotate them to build a text knowledge base for power dispatching automation operations.

2. A method for extracting entities from power dispatching automation based on a knowledge base, using a power dispatching automation entity extraction system based on a knowledge base, characterized in that: The steps include: Step S1, extracting text features from multimodal power dispatch automation data; Step S2, reducing text redundancy by calculating text similarity to obtain high-quality text data; Step S3: Design an entity extraction module based on LDA to build a knowledge base by accurately annotating entities; Step S4: Based on the knowledge base, use the BERT-BiLSTM-CRF model to extract entities from power dispatch automation data and obtain entities in the text.

3. The method for extracting entities from power dispatch automation based on a knowledge base according to claim 2 is characterized in that: In step S1, text feature extraction is performed on multi-modal power dispatch automation data, which specifically includes the following steps: Step S1A1: For data of word and txt text types, use the python-docx library to extract and classify text according to the structural format of the text; Step S1A2: For the PDF data of ledger information and operation manual, use OCR technology to extract text information from the PDF file; Step S1A3: For data in the Excel table type of work tickets and safety regulations, use the pandas library to extract key text; Step S1A4: reconstruct different types of text using a text segmentation method with a threshold, segment text with a length greater than an upper threshold, and remove text with a length less than a lower threshold; Step S1A5: After steps S1A1, S1A2, S1A3, and S1A4, the initially extracted power dispatching automation text data is obtained.

4. The method for extracting entities from power dispatch automation based on a knowledge base according to claim 2 is characterized in that: In step S2, the text redundancy is reduced by calculating the text similarity to obtain high-quality text data. The specific steps include: Step S2A1, setting a set of weight parameters λ=(λ1, λ2, λ3), which respectively represent the weights of edit distance, cosine similarity, and TF-IDF; setting a threshold Ω, which represents the threshold of text similarity; Step S2A2, manually extract a small number of samples of different types and establish a sample knowledge base D; Step S2A3, by calculating the edit distance between the text data and the knowledge base D, the edit distance similarity S1 is obtained; Step S2A4, by calculating the cosine similarity between the text data and the knowledge base D, the cosine similarity S2 is obtained; Step S2A5, by calculating the TF-IDF value of the text data and the knowledge base D, the TF-IDF similarity S3 is obtained; Step S2A6, by weighted calculation of the similarity, the text similarity score S is obtained; Step S2A7, comparing the similarity score S with the threshold, if it is greater than the threshold Ω, it means deleting the text; if it is less than or equal to the threshold, the text is added to the knowledge base D; Step S2A8: Repeat steps S2A3-S2A7 to build a complete entity text knowledge base D.

5. The method for extracting entities from power dispatch automation based on a knowledge base according to claim 4 is characterized in that: In step S2A3, the specific calculation of the edit distance and the edit distance similarity is as follows: In the formula, is the indicator function, when a i =b j When the value is 0, the value is 1 at other times; a,b (i,j) represents the edit distance between the first i characters of a and the first j characters of b; |a|, |b| represent the length of text a and text b, and similarity represents the edit distance similarity; In step S2A4, the specific calculation of cosine similarity is as follows: In the formula, A i , B i Represents the values ​​of each dimension of vector A and B, It means the inner product of vectors A and B. represents the product of A and B modulo; In step S2A5, the specific calculation of TF-IDF is as follows: TF-IDF=TF·IDF Where n i,j Indicates the term t i In the document d j The number of occurrences in TF i,j Indicates the term t i In the document d j The frequency of occurrence in. |D| represents the number of all documents, |j:t i ∈d j | indicates that the term is included i The number of documents; In step S2A6, the weighted calculation of similarity is specifically calculated as follows: S=λ1S1+λ2S2+λ3S3 Where λ1, λ2, λ3 are weight parameters set in S2A1, and S1, S2, S3 are similarities calculated in steps S2A3, S2A4, and S2A5.

6. The method for extracting entities of electric power dispatching automation based on knowledge base according to claim 2 is characterized in that: In step S3, an entity extraction module based on LDA is designed to accurately annotate entities and realize knowledge base construction, which specifically includes the following steps: Step S3A1: Use the LDA algorithm to extract text keywords from the knowledge base D. Step S3A2: label the entities extracted using the LDA algorithm according to the actual situation and construct an entity extraction knowledge base D.

7. The method for extracting entities from power dispatch automation based on a knowledge base according to claim 6 is characterized in that: In step S3A1, the specific calculation of the LDA algorithm is as follows: Where P(W,Z) represents the file probability of a certain combination generated by LDA; P(θ j ,α) represents the probability distribution of the main body composition of each file; represents the composition probability distribution of each topic; P(Z j,t |θ j ) represents the probability multinomial distribution of different topics for a certain file; Represents the probability multinomial distribution of words for a topic.

8. The method for extracting entities of electric power dispatching automation based on knowledge base according to claim 1 is characterized in that: In step S4, a BERT-BiLSTM-CRF entity extraction model is constructed, and the knowledge base is used to correct the labels when predicting entity labels, which specifically includes the following steps: Step S4A1: Build a BERT-BiLSTM-CRF model to extract entities from text and perform label prediction; Step S4A2: Use the knowledge base to correct the incorrect labels of the entities extracted in S4A1.

9. The method for extracting entities from power dispatch automation based on a knowledge base according to claim 8, characterized in that: The BERT-BiLSTM-CRF model includes the BERT model, the BiLSTM model, and the CRF model; In step S4A1, the specific formula of the BERT model is as follows: V=BERT([CLS],T1,...,T n ,[SEP])=(e1,...,e n ) Where, E={E1,E2,...,E n } represents the initial text sequence, T={T1,T2,...,T n } represents the text vector obtained after the initial text sequence is encoded by the transformer encoder. In step S4A1, the specific formula of the BiLSTM model is as follows: i t =σ(W ix X t +W ih h t-1 +b i ) f t =σ(W fx X t +W fh h t-1 +b f ) o t =σ(W ox X t +W oh h t-1 +b o ) h t =o t Θtanh(C t ) In the formula, f t represents the state of the forget gate at time t, o t represents the state of the output gate at time t, represents the cell state at time t, h t represents the hidden layer state at time t; In step S4A1, the specific formula of the CRF model is as follows: In the formula, N represents the number of training samples, C represents the regularization parameter, and y (i) represents the true label sequence of the i-th sample, x (i) represents the input sequence of the i-th sample, p(y (i) |x (i) ) represents the conditional probability distribution in the CRF model.

Citation Information

Patent Citations

  • Domain entity disambiguation method for fusing word vectors and topic model

    CN107861939A

  • A field entity attribute relation extraction method based on distance supervision

    CN109408642A

  • Function point extraction processing method and system

    CN116775954A

  • Power grid fault prediction method based on multi-modal knowledge hybrid reasoning

    CN116910633A

  • Electrolyte for lithium secondary battery and lithium secondary battery comprising same

    KR1020250058418A