Semantic analysis and integration method and system for power grid standard document

Abstract

The invention discloses a semantic analysis integration method and system for a power grid standard document, and relates to the technical field of power grid standard document processing, and the method comprises the steps: collecting and preprocessing a power grid system document, and employing a first model to recognize a power entity in the power grid system document; a second model is adopted to mine the relationship between the power entities, knowledge fusion is carried out, and a knowledge graph is constructed; and performing statistical analysis and classification on the user work orders in the power grid system document, and reinforcing the existing relationship in the knowledge graph by using a classification result to complete analysis and integration of the power grid system document. By using the combined structure of the bidirectional long-short-term memory network and the conditional random field, the method can effectively capture the dependency relationship between the context information and the tag, and the accuracy of power entity identification is greatly improved. And meanwhile, specific entity embedded information in the power field is introduced, so that the understanding capability of the model on the power terminology is further enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid standard document processing, and in particular to a semantic analysis and integration method and system for power grid standard documents. Background Art

[0002] Semantic analysis and integration of power grid standard documents is a key link in the digital transformation of power systems, and is of great significance for improving the efficiency and reliability of power grid operations. With the rapid development of smart grid technology, the number and complexity of power system documents have grown exponentially, and traditional manual processing methods can no longer meet current needs. In recent years, artificial intelligence technologies such as natural language processing and knowledge graphs have made significant progress in the field of text analysis, providing new technical support for the intelligent processing of power grid standard documents. However, due to the professionalism and particularity of the power field, general semantic analysis methods often have problems such as insufficient adaptability and lack of domain knowledge when applied to power grid documents, making it difficult to accurately identify and extract power professional terms, equipment entities and their complex relationships.

[0003] Existing power grid document analysis methods mainly focus on basic tasks such as keyword extraction and text classification, lacking the mining of deep semantics of document content and the integration of cross-document knowledge. These methods usually use rule matching or shallow machine learning algorithms, which are difficult to handle implicit semantics and complex contextual relationships in power grid documents. In addition, existing technologies often separate document analysis from knowledge application, and fail to fully utilize user feedback and actual application scenarios to optimize and improve the knowledge system. This leads to a gap between the analysis results and actual needs, making it difficult to provide accurate and comprehensive knowledge support for power grid operations and decision-making. At the same time, due to the lack of an effective knowledge fusion mechanism, power grid document information from different sources and in different formats is difficult to form a unified knowledge system, affecting the consistency and reusability of knowledge. Summary of the invention

[0004] In view of the problems existing in the existing semantic analysis integration technology, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is that it is difficult to process the implicit semantics and complex contextual relationships in power grid documents.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides a semantic analysis and integration method for power grid standard documents, which includes:

[0008] Collecting and preprocessing power grid system documents, and using a first model to identify power entities in the power grid system documents;

[0009] The second model is used to mine the relationship between the power entities, and knowledge fusion is performed to construct a knowledge graph;

[0010] Statistical analysis and classification are performed on the user work orders in the power grid system documents, and the classification results are used to strengthen the existing relationships in the knowledge graph to complete the analysis and integration of the power grid system documents.

[0011] As a preferred solution of the semantic analysis and integration method of the power grid standard document of the present invention, wherein: the first model identifies the power entities in the power grid system document including:

[0012] Use neural networks to capture the context of input sequences;

[0013] The captured information is linearly transformed and compressed into a fixed-dimensional vector;

[0014] Introduce specific entity embedded information and synthesize;

[0015] Performing feature fusion on the dimension vector and entity embedding information to obtain entity probability distribution;

[0016] The entity probability distribution is shown as follows:

[0017]

[0018] Among them, PEIM(x) is the probability distribution of the output, x is the input text sequence, T is the sequence length, and h t is the hidden state of Bi-LSTM at time step t, N is the number of entity types, E(x i ) is the embedding representation of the i-th entity, and λ is the contribution hyperparameter.

[0019] As a preferred solution of the semantic analysis and integration method of the power grid standard document of the present invention, the second model mines the relationship between the power entities, as shown in the following formula:

[0020]

[0021] Among them, SAM is the output of the syntactic analysis model; x is the input text sequence; T is the sequence length; H is the output of the hidden Markov model; C is the output of the conditional random field model, b is the observation probability; λ is the feature weight; and f is the feature function.

[0022] As a preferred solution of the semantic analysis and integration method of the power grid standard document of the present invention, wherein: the knowledge fusion includes entity disambiguation and coreference resolution;

[0023] The entity disambiguation includes:

[0024] For polysemous words, the context of the target word is analyzed, including surrounding modifiers, related attributes or technical parameters. If specific technical indicators or location information appear near the target word, it is judged to refer to a specific device or entity. On the contrary, if the target word is paired with an operational verb or appears in an operation flow description, it refers to an action or process.

[0025] As a preferred solution of the semantic analysis and integration method of the power grid standard document of the present invention, wherein: the coreference resolution includes:

[0026] When encountering a pronoun, look for the most recently mentioned compatible entity and infer the specific entity name based on the context. When processing a descriptive phrase, extract the key features and attributes in the descriptive phrase and search for previously identified entities. If all known features of an entity completely match the descriptive phrase, then this entity is the description object. If there is no complete match, check which entities have the most matching features and no obvious contradictory attributes. The one with the most matching features is the description object. If no entity matches, the related entity closest to the descriptive phrase is identified as the description object. If no suitable match is found, evaluate whether the descriptive phrase contains enough new information to establish a completely new entity. If not, mark the descriptive phrase, otherwise establish a completely new entity.

[0027] As a preferred solution of the semantic analysis and integration method of the power grid standard document of the present invention, the statistical analysis and classification of the user work order includes:

[0028] Count the keyword frequency in the text of the work order

[0029] Determine statistics for numerical data in work orders;

[0030] Use the third model to derive specific work order classification;

[0031] The third model is shown as follows:

[0032]

[0033] Among them, FIF(x) is the work order type, t is a keyword or keyword combination, d is a document, D is a document set, and f t,d is the number of occurrences of keyword t in document d, ∑ t′∈d f t′,d f t,d The total number of all keywords in document d, |D| is the total number of documents in document set D, {d∈D:t∈d} is the number of documents containing keyword t, α is the adjustment parameter, and μ is the mean; x iis the ith data point; n is the number of data points; σ2 is the variance; γ1 is the skewness; γ2 is the kurtosis; β is the adjustment parameter; T is the keyword set; CSF(x) is the comprehensive statistical feature; w i is the weight of each feature.

[0034] As a preferred solution of the semantic analysis and integration method of the power grid standard document of the present invention, wherein: the keyword includes a language representation composed of two or more words.

[0035] In a second aspect, an embodiment of the present invention provides a semantic analysis and integration system for power grid standard documents, which includes:

[0036] Data acquisition module, used to collect and pre-process power grid system documents;

[0037] A knowledge graph construction module, used to use a first model to identify power entities in the power grid system document, use a second model to mine the relationship between the power entities, and perform knowledge fusion to construct a knowledge graph;

[0038] The classification and integration module is used to perform statistical analysis and classification on the user work orders in the power grid system documents, and use the classification results to strengthen the existing relationships in the knowledge graph to complete the analysis and integration of the power grid system documents.

[0039] In a third aspect, an embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the processor executes the computer program, it implements any step of the above-mentioned semantic analysis and integration method of power grid standard documents.

[0040] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the above-mentioned method for semantic analysis and integration of power grid standard documents is implemented.

[0041] The beneficial effect of the present invention is that by using the combined structure of bidirectional long short-term memory network (Bi-LSTM) and conditional random field (CRF), the method can effectively capture the dependency relationship between context information and labels, greatly improving the accuracy of power entity recognition. At the same time, the introduction of entity embedding information specific to the power field further enhances the model's ability to understand power professional terms. Combining the syntactic analysis method of conditional random field and hidden Markov model, the specific language patterns and complex dependencies in power grid documents are processed. By developing a dynamic update mechanism, the system can analyze the newly processed work orders in real time, extract new keywords, entity relationships and other information, and integrate this information into the knowledge graph through a multi-stage verification process. Through entity disambiguation and co-reference resolution technology, the system can effectively solve the problem of entities with the same name referring to different concepts, and identify different expressions but referring to the same entity. Using the work order classification results to strengthen the existing relationships in the knowledge graph, the system can automatically discover and establish new associations, such as the "common fault risk" relationship. This method enables the knowledge graph to be continuously optimized and improved as the actual application scenario changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0043] Figure 1 Flowchart of the integrated approach for semantic analysis of power grid standard documents.

[0044] Figure 2 Schematic diagram of the process of building a knowledge base for the semantic analysis and integration method of power grid standard documents.

[0045] Figure 3 Schematic diagram of the flow of text classification integrating methods for semantic analysis of power grid standard documents. DETAILED DESCRIPTION

[0046] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0047] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0048] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0049] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0050] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0051] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0052] Example 1

[0053] Reference Figure 1 to Figure 3 , which is the first embodiment of the present invention, and which provides a semantic analysis and integration method for power grid standard documents, comprising:

[0054] Build a knowledge base, annotate relevant documents of the power grid system based on information such as operation tickets and user work order descriptions, and then combine the annotated corpus and deep learning methods to train the power entity recognition model and syntactic analysis model. Through the power entity recognition model, the power grid system related entities can be recognized. By training the syntactic analysis model, the syntactic structure between power entities and other power feature words can be extracted and identified, and the relationship between power entities can be mined, structured knowledge information can be obtained, and the domain knowledge base can be constructed using knowledge fusion engineering.

[0055] Specifically, various types of power grid system-related documents are collected, including but not limited to operation tickets, user work orders, equipment maintenance records, fault reports, etc. These documents come from different systems and departments, and the formats are inconsistent. Therefore, data cleaning and standardization are required to unify the format, remove noise data, and correct erroneous information.

[0056] The preprocessed documents are subjected to preliminary text analysis, including word segmentation and part-of-speech tagging. Experts in the power field manually annotate the key information in the documents and identify important power entities (such as equipment name, operating steps, fault type, etc.) and the relationships between entities.

[0057] The labeled data set is used to train the power entity recognition model, which adopts a combination of a bidirectional long short-term memory network (Bi-LSTM) and a conditional random field (CRF). Bi-LSTM can capture contextual information, while CRF can learn the dependency relationship between labels and improve the accuracy of entity recognition.

[0058] For example, in one feasible embodiment, as follows:

[0059] Use Bi-LSTM to capture the contextual information of the input sequence:

[0060] h t =BiLSTM(x t )

[0061] Furthermore, the output of Bi-LSTM is linearly transformed and averaged, and then normalized to reduce the impact of sequence length on the result, thereby compressing the sequence information into a fixed-dimensional vector:

[0062]

[0063] Then, we introduce entity embedding information specific to the power sector and integrate the information of multiple entities by averaging:

[0064]

[0065] Finally, feature fusion is performed to balance the importance of the two types of information and map their output to a specific interval, so that the model output is an interpretable parameter value.

[0066] PEIM(x)=σ(PEIM 1 (x)+λPEIM 2 (x)

[0067] Expands to:

[0068]

[0069] Among them, PEIM(x) is the output probability distribution of the power entity recognition model, x is the input text sequence, T is the sequence length, and h t is the hidden state of Bi-LSTM at time step t, N is the number of entity types, E(x i ) is the embedding representation of the i-th entity, and λ is the contribution hyperparameter.

[0070] Furthermore, based on the labeled data, a syntactic analysis model is trained. This model combines conditional random fields and hidden Markov models (HMM). First, HMM is used for preliminary part-of-speech tagging, and then CRF is used for more detailed dependency analysis. This model can reveal the grammatical relationship between the various components in the sentence, laying the foundation for subsequent entity relationship extraction.

[0071] For example, in a feasible embodiment, the syntax analysis model is as follows:

[0072] First, HMM is used for preliminary part-of-speech tagging. In power grid documents, certain word order patterns often appear, such as "equipment-action-state". HMM can effectively capture this sequence pattern. For example, the sequence "transformer-start-normal operation" frequently appears in the operation ticket, but the initial state probability of the original HMM is less affected by the middle state of the sequence, so this part is removed and only the state transition probability and observation probability are retained, as shown in the following formula:

[0073]

[0074] Among them, a is the state transition probability; b is the observation probability, and x is the input text sequence.

[0075] Furthermore, there are complex dependencies in power grid documents, such as the connection relationship between devices and the sequence of operation steps. CRF can capture these long-distance dependencies. For example, in a fault report, the relationship between "line tripping" and "protection action" can be modeled by the characteristic function of CRF:

[0076] C(x t -x t-1 )=∑ k λ k f k (yt-1 ,y t ,x,t)

[0077] Among them, λ is the feature weight; f is the feature function;

[0078] Furthermore, word order (such as the order of operation steps), dependency (such as the connection between devices) and professional terminology are considered at the same time, and the above two are added together. Since it is necessary to select the best one from multiple possible syntactic structures, various possible syntactic analysis results are converted into probability distribution, so as to select the most reasonable structure, as shown in the following formula:

[0079]

[0080] Among them, SAM is the output of the syntactic analysis model; x is the input text sequence; T is the sequence length; H is the output of the hidden Markov model; C is the output of the conditional random field model, b is the observation probability; λ is the feature weight; and f is the feature function.

[0081] The document is processed using the trained entity recognition model and syntax analysis model. After the power entities are identified, the relationship between the entities is mined based on the syntax structure and semantic rules. For example, the relationship between "transformer" and "fault" or the sequence relationship between "operation steps" can be identified.

[0082] After acquiring a large number of entities and relations, knowledge fusion is required. This process includes entity disambiguation and coreference resolution. Entity disambiguation aims to solve the problem that entities with the same name refer to different concepts, while coreference resolution is to identify different expressions that refer to the same entity.

[0083] Preferably, for polysemous words, the context of the target word is analyzed, including surrounding modifiers, related attributes or technical parameters. If specific technical indicators or location information appear near the target word, it is determined to refer to a specific device or entity. On the contrary, if the target word is paired with an operational verb or appears in an operational flow description, it refers to an action or process. Secondly, the results of part-of-speech tagging and dependency syntax analysis are used to determine the meaning of the target word based on its grammatical role in the sentence.

[0084] For coreference resolution, we build an entity mention chain in the document. When we encounter a pronoun or ambiguous description, we review the previous content and guess which device or concept it refers to based on the context and known information. In this process, we consider many factors, such as whether the meaning of the words matches, whether the features described are consistent, whether the positions in the article are close, and whether they fit the overall context. If we are not sure, we will mark it out for manual inspection.

[0085] Furthermore, based on the integrated knowledge, a multi-dimensional knowledge graph is constructed.

[0086] Furthermore, in order to keep the knowledge base up to date and comprehensive, a dynamic update mechanism was developed. This mechanism analyzes newly processed work orders in real time and extracts new keywords, entity relationships, and other information. Then, through a multi-stage verification process, this new information is integrated into the knowledge graph.

[0087] Specifically, whenever a new work order enters the system, the analysis process is immediately triggered. This analysis process first applies the aforementioned professional word segmentation and entity recognition models to convert the work order text into structured information. Next, the system performs an information extraction process to identify key information in the work order. This includes new professional terms, new relationships between entities, newly emerging attribute values, etc. The system not only focuses on known information types, but also identifies potential new categories of information.

[0088] The extracted information then goes through a multi-stage verification process. The first stage is an automated consistency check. The system compares the newly extracted information with the information already in the knowledge graph. It checks whether the new information conflicts or is inconsistent with existing knowledge. If a potential conflict is found, the system flags the information for further verification.

[0089] The second stage is credibility assessment. The system considers multiple factors to assess the credibility of new information, including the reliability of the information source, the confidence of the extraction model, the degree of consistency with existing knowledge, etc. Based on these factors, the system assigns a credibility score to each new piece of information.

[0090] The third stage is human-machine collaborative verification. For information with high credibility, the system directly integrates it into the knowledge graph. For information with medium credibility or potential conflicts, the system will submit it to customer service personnel for review. This process is carried out through a specially designed interface, where customer service personnel can view new information, relevant context and the system's recommendations, and then make a decision to accept, reject or modify it.

[0091] New information that passes verification will be formally integrated into the knowledge graph. This process is not just a simple addition. The system will update the attributes of related entities, establish new relationship connections, and reorganize some knowledge structures to adapt to new information.

[0092] Furthermore, the frequency, mean, and variance of the keywords on the user work order are counted to integrate and classify the text, and the relevant information in the knowledge graph is summarized and classified according to the semantics expressed by the key information on the user work order.

[0093] Specifically, we first calculate the frequency of keywords in the text, TF-IDF value and other statistical features, and calculate the mean, variance, skewness, kurtosis and other statistics for the numerical data. These statistics are then input into a multi-class classification model to obtain the specific classification of the user ticket.

[0094] Preferably, in the daily operation and maintenance of the power system, common types of problems include power outages due to unpaid bills, equipment failures, line maintenance, etc. These problems are usually determined by multiple text and numerical features. In order to improve the accuracy and effectiveness of the classification model, the keywords in the text are analyzed in detail and comprehensively evaluated in combination with the numerical features.

[0095] Furthermore, based on the traditional TF-IDF algorithm, the influence of keyword frequency on the results is adjusted, and the weight of high-frequency words is reduced, so as to more accurately reflect the importance of keywords, as follows:

[0096]

[0097] Among them, t is a keyword or keyword combination, d is a document, D is a document collection, and f t,d is the number of occurrences of keyword t in document d, ∑ t′∈d f t′,d f t,d The total number of all keywords in document d, |D| is the total number of documents in document set D, {d∈D:t∈d} is the number of documents containing keyword t, and α is an adjustment parameter used to adjust the influence of keyword frequency on the results.

[0098] Furthermore, the mean, variance, skewness and kurtosis in the document are calculated, and the TF-IDF algorithm is combined with the statistical features to form comprehensive statistical features:

[0099]

[0100] Among them, μ is the mean; x i is the ith data point; n is the number of data points; σ2 is the variance; γ1 is the skewness; γ2 is the kurtosis; β is the adjustment parameter; T is the keyword set; CSF(x) is the comprehensive statistical feature;

[0101] Finally, the comprehensive statistical features are input into the multi-class classification model:

[0102]

[0103] Among them, w i is the weight of each feature.

[0104] It should be noted that keywords or keyword combinations are language expressions consisting of two or more words, and are a way of expressing something.

[0105] Keyword collocations include: noun phrases, verb phrases and some fixed phrases, such as power outage due to arrears, no power when the circuit breaker is closed, circuit breaker tripped, etc. Phrases (collocations) give new meanings to words and play a very important role in their specific usage contexts. For example, both power outage due to arrears and power outage contain the word "power outage", but the former refers to a special event caused by a power outage due to arrears, while the latter refers to the broad concept of power outage.

[0106] Furthermore, the classification results are used to strengthen existing relationships. For example, if multiple work orders are classified as the same type of problem and involve equipment in the same area, the "common failure risk" relationship between these devices is added to the knowledge graph.

[0107] Furthermore, this embodiment also provides a semantic analysis and integration system for power grid standard documents, including:

[0108] Data acquisition module, used to collect and pre-process power grid system documents;

[0109] A knowledge graph construction module, used to use a first model to identify power entities in the power grid system document, use a second model to mine the relationship between the power entities, and perform knowledge fusion to construct a knowledge graph;

[0110] The classification and integration module is used to perform statistical analysis and classification on the user work orders in the power grid system documents, and use the classification results to strengthen the existing relationships in the knowledge graph to complete the analysis and integration of the power grid system documents.

[0111] This embodiment also provides a computer device, which is suitable for the semantic analysis and integration method of power grid standard documents, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the semantic analysis and integration method of power grid standard documents proposed in the above embodiment.

[0112] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0113] This embodiment further provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for semantic analysis and integration of power grid standard documents proposed in the above embodiment is implemented.

[0114] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0115] Example 2

[0116] This is the second embodiment of the present invention, which provides a semantic analysis and integration method for power grid standard documents. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0117] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A semantic analysis and integration method for power grid standard documents, characterized by: include, Collecting and preprocessing power grid system documents, and using a first model to identify power entities in the power grid system documents; The second model is used to mine the relationship between the power entities, and knowledge fusion is performed to construct a knowledge graph; Statistical analysis and classification are performed on the user work orders in the power grid system documents, and the classification results are used to strengthen the existing relationships in the knowledge graph to complete the analysis and integration of the power grid system documents.

2. The semantic analysis and integration method of power grid standard documents according to claim 1, characterized in that: The first model identifies the power entities in the power grid system document including: Use neural networks to capture the context of input sequences; The captured information is linearly transformed and compressed into a fixed-dimensional vector; Introduce specific entity embedded information and synthesize; Performing feature fusion on the dimension vector and entity embedding information to obtain entity probability distribution; The entity probability distribution is shown as follows: Among them, PEIM(x) is the probability distribution of the output, x is the input text sequence, T is the sequence length, and h t is the hidden state of Bi-LSTM at time step t, N is the number of entity types, E(x i ) is the embedding representation of the i-th entity, and λ is the contribution hyperparameter.

3. The semantic analysis and integration method of power grid standard documents according to claim 2, characterized in that: The second model mines the relationship between the power entities, as shown in the following formula: Among them, SAM is the output of the syntactic analysis model; x is the input text sequence; T is the sequence length; H is the output of the hidden Markov model; C is the output of the conditional random field model, b is the observation probability; λ is the feature weight; and f is the feature function.

4. The semantic analysis and integration method of power grid standard documents according to claim 3, characterized in that: The knowledge fusion includes entity disambiguation and coreference resolution; The entity disambiguation includes: For polysemous words, the context of the target word is analyzed, including surrounding modifiers, related attributes or technical parameters. If specific technical indicators or location information appear near the target word, it is judged to refer to a specific device or entity. On the contrary, if the target word is paired with an operational verb or appears in an operation flow description, it refers to an action or process.

5. The semantic analysis and integration method of power grid standard documents according to claim 4, characterized in that: The coreference resolution includes: When encountering a pronoun, look for the most recently mentioned compatible entity and infer the specific entity name based on the context. When processing a descriptive phrase, extract the key features and attributes in the descriptive phrase and search for previously identified entities. If all known features of an entity completely match the descriptive phrase, then this entity is the description object. If there is no complete match, check which entities have the most matching features and no obvious contradictory attributes. The one with the most matching features is the description object. If no entity matches, the related entity closest to the descriptive phrase is identified as the description object. If no suitable match is found, evaluate whether the descriptive phrase contains enough new information to establish a completely new entity. If not, mark the descriptive phrase, otherwise establish a completely new entity.

6. The semantic analysis and integration method of power grid standard documents according to claim 5, characterized in that: Statistical analysis and classification of the user work orders include: Count the keyword frequency in the text of the work order Determine statistics for numerical data in work orders; Use the third model to derive specific work order classification; The third model is shown as follows: Among them, FIF(x) is the work order type, t is a keyword or keyword combination, d is a document, D is a document set, and f t,d is the number of occurrences of keyword t in document d, ∑ t′∈d f t′,d f t,d The total number of all keywords in document d, |D| is the total number of documents in document set D, {d∈D:t∈d} is the number of documents containing keyword t, α is the adjustment parameter, and μ is the mean; x i is the ith data point; n is the number of data points; σ2 is the variance; γ1 is the skewness; γ2 is the kurtosis; β is the adjustment parameter; T is the keyword set; CSF(x) is the comprehensive statistical feature; w i is the weight of each feature.

7. The semantic analysis and integration method of power grid standard documents according to claim 6, characterized in that: The keyword includes a language representation consisting of two or more words.

8. A semantic analysis and integration system for power grid standard documents, based on the semantic analysis and integration method for power grid standard documents according to any one of claims 1 to 7, characterized in that: include, Data acquisition module, used to collect and pre-process power grid system documents; A knowledge graph construction module, used to use a first model to identify power entities in the power grid system document, use a second model to mine the relationship between the power entities, and perform knowledge fusion to construct a knowledge graph; The classification and integration module is used to perform statistical analysis and classification on the user work orders in the power grid system documents, and use the classification results to strengthen the existing relationships in the knowledge graph to complete the analysis and integration of the power grid system documents.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for semantic analysis and integration of power grid standard documents according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for semantic analysis and integration of power grid standard documents described in any one of claims 1 to 7 are implemented.

Images