Power equipment data management method and system and storage medium

Abstract

The invention discloses a power equipment data management method and system and a storage medium, and the method comprises the steps: carrying out the sentence coding of the obtained initial text data of each power equipment of a target power system through employing a BERT model, and obtaining each semantic feature vector; performing sequence labeling on the semantic feature vector by using a conditional random field to obtain each labeling result; performing dependency syntactic analysis on the semantic feature vector to obtain a dependency type matrix and an adjacent matrix; performing feature extraction on a first fusion result of the semantic feature vector, the dependency type matrix and the adjacent matrix by adopting a graph convolutional network model; processing a feature extraction result by adopting a classifier to obtain each relation extraction result; and storing a second fusion result of all the labeling results and all the relation extraction results by adopting a graph database to obtain a power equipment knowledge base. According to the power equipment data management method and system provided by the embodiment of the invention, the management efficiency of the power equipment data is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a power equipment data management method, system and storage medium. Background Art

[0002] Power equipment data plays a vital role in the power industry. Power equipment data covers the basic information, operating status, maintenance records and other aspects of power equipment. By analyzing historical power equipment data and failure cases, companies can more accurately formulate equipment maintenance plans to ensure that equipment operates in the best condition and reduce downtime caused by equipment failure. Therefore, the management of power equipment data is of great significance to the management, monitoring, maintenance and optimization of equipment.

[0003] However, the existing power equipment data management is decentralized and lacks integrated processing, resulting in low management efficiency of power equipment data, which is not conducive to the subsequent industrialization, intelligentization and large-scale development of power equipment.

[0004] It can be seen that how to build a unified power equipment data management platform to improve the management efficiency of power equipment data has become a technical problem that technical personnel in this field need to solve urgently. Summary of the invention

[0005] The present invention provides a power equipment data management method, system and storage medium to solve the technical problem that the existing power equipment data management is decentralized and lacks integrated processing, resulting in low management efficiency of the power equipment data.

[0006] In order to solve the above technical problems, an embodiment of the present invention provides a method for managing power equipment data.

[0007] The BERT model is used to perform sentence encoding on the initial text data of each power device in the target power system to obtain each semantic feature vector; Performing sequence labeling on all the semantic feature vectors using a conditional random field to obtain various labeling results; Perform dependency syntactic analysis on each of the semantic feature vectors to obtain a dependency type matrix and an adjacency matrix; use a graph convolutional network model to perform feature extraction on a first fusion result of the semantic feature vector, the dependency type matrix, and the adjacency matrix; use a classifier to process the result of the feature extraction to obtain each relationship extraction result; A graph database is used to store the second fusion result of all the annotation results and all the relationship extraction results to obtain a power equipment knowledge base.

[0008] As one of the preferred solutions, the BERT model is used to perform sentence encoding on the acquired initial text data of each power device in the target power system to obtain various semantic feature vectors, including: Using a forward long short-term memory network to perform a first encoding on each of the initial text data, to obtain a first hidden layer representation of each of the initial text data; wherein the first hidden layer representation reflects the semantics and contextual relationship of the corresponding initial text data from front to back; Performing a second encoding on each of the initial text data using a backward long short-term memory network to obtain a second hidden layer representation of each of the initial text data; wherein the second hidden layer representation reflects the semantics and contextual relationship of the corresponding initial text data from back to front; The vector obtained by concatenating the first hidden layer representation of each of the initial text data and the corresponding second hidden layer representation is used as the hidden layer feature vector of each of the initial text data; The BERT model is used to perform sentence encoding on the hidden layer feature vector to obtain various semantic feature vectors.

[0009] As one of the preferred solutions, the BERT model is used to perform sentence encoding on the acquired initial text data of each electric device in the target electric power system to obtain each semantic feature vector, and further includes: Adjusting the first weight of the BERT model based on the degree of association between all word vectors of each of the initial text data; The word vector includes the first hidden layer representation and the second hidden layer representation.

[0010] As one of the preferred solutions, the dependency type matrix is ​​used to represent the dependency relationship type between words in a sentence; The adjacency matrix is ​​used to represent the connection relationship between words in a sentence.

[0011] As one of the preferred solutions, the method of using a graph convolutional network model to extract features from the first fusion result of the semantic feature vector, the dependency type matrix, and the adjacency matrix also includes: Using an attention mechanism to fuse the dependency type matrix and the adjacency matrix; The second weight of the graph convolutional network model is adjusted according to the result of the fusion processing.

[0012] As one of the preferred solutions, each of the initial text data at least includes the construction process, technical parameters, selection principles, personnel configuration and equipment use specifications of the equipment.

[0013] As one of the preferred solutions, the power equipment knowledge base includes nodes, attributes and relationships; The connection between the node and the attribute is one-to-many; The connection mode between the nodes and the relationships is one-to-many.

[0014] Another embodiment of the present invention provides an electric power equipment data management system, comprising: A data encoding module is used to use a BERT model to perform sentence encoding on the acquired initial text data of each power device in the target power system to obtain each semantic feature vector; An entity labeling module, used to perform sequence labeling on all the semantic feature vectors using a conditional random field to obtain various labeling results; A relationship extraction module is used to perform dependency syntactic analysis on each of the semantic feature vectors to obtain a dependency type matrix and an adjacency matrix; use a graph convolutional network model to perform feature extraction on a first fusion result of the semantic feature vector, the dependency type matrix and the adjacency matrix; use a classifier to process the result of the feature extraction to obtain each relationship extraction result; The database construction module is used to use a graph database to store the second fusion result of all the annotation results and all the relationship extraction results to obtain a power equipment knowledge base.

[0015] As one of the preferred solutions, the entity labeling module and the relationship extraction module adopt a parameter sharing strategy so that the entity labeling module and the relationship extraction module can work together under the same network framework.

[0016] Yet another embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein when a device where the computer-readable storage medium is located executes the computer program, an electric power equipment data management method as described above is implemented.

[0017] Compared with the prior art, the embodiments of the present invention have the following advantages: The BERT model is used to perform sentence encoding on the initial text data of each power equipment in the target power system to obtain various semantic feature vectors. The first weight of the BERT model is adjusted based on the degree of association between all word vectors of each initial text data, which helps to obtain important features in the initial text data; all semantic feature vectors are sequence labeled using conditional random fields to obtain various labeling results; dependency syntactic analysis is performed on each semantic feature vector to obtain a dependency type matrix and an adjacency matrix, and a graph convolutional network model is used to extract features from the first fusion result of the semantic feature vector, dependency type matrix and adjacency matrix, which not only considers the dependency relationship of context words in the initial text data, but also incorporates the relationship type into the context features. In order to measure different The weight of the contextual features is used to fuse the dependency type matrix and the adjacency matrix with the attention mechanism. The second weight of the graph convolutional network model is adjusted according to the result of the fusion processing. Different weights are assigned to the contextual features to replace the weights of the traditional standard graph convolutional network model. This helps to filter out noise from syntactic knowledge and highlight the value of key information, thereby improving the model's ability to recognize important information. The classifier is used to process the results of feature extraction to obtain various relationship extraction results; the graph database is used to store the second fusion results of all annotation results and all relationship extraction results to obtain a power equipment knowledge base. The graph database is used to store the power equipment data in an integrated manner, making full use of the complex relationships between power equipment and improving the management efficiency of power equipment data. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flowchart of a method for managing power equipment data in one embodiment of the present invention; Figure 2 A structural block diagram of a power equipment data management system in one embodiment of the present invention; Figure 3 A schematic diagram of obtaining a knowledge base of electric power equipment in one embodiment of the present invention. DETAILED DESCRIPTION

[0019] 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. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. 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.

[0020] In the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second", "third", etc. may explicitly or implicitly include one or more of the feature. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0021] In the description of the present application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. The terms "vertical", "horizontal", "left", "right", "upper", "lower" and similar expressions used herein are only for illustrative purposes, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0022] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by those skilled in the art. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood by specific circumstances.

[0023] An embodiment of the present invention provides a method for managing data of an electric power device. For details, see Figure 1~Figure 3 , Figure 1 FIG. 1 is a flow chart of a method for managing power equipment data in one embodiment of the present invention. Figure 2 FIG. 1 is a structural block diagram of a power equipment data management system in one embodiment of the present invention. Figure 3 Shown is a schematic diagram of acquiring a knowledge base of electric power equipment in one embodiment of the present invention.

[0024] A flowchart of a power equipment data management method in one embodiment of the present invention includes the following steps S1 to S4, which are specifically as follows: Step S1: Use the BERT model to perform sentence encoding on the acquired initial text data of each power device in the target power system to obtain each semantic feature vector.

[0025] It should be noted that the information related to power equipment in the power system is massive and complex, and there are many related documents. The existing power equipment data management is scattered and lacks integrated processing. The management efficiency of power equipment data is low, which leads to difficulties for technicians in the process of consulting and analysis, which is not conducive to the subsequent industrialization, intelligence and large-scale development of power equipment. Therefore, it is necessary to build a unified power equipment data management platform to improve the management efficiency of power equipment data. Named entity recognition is an important task in natural language processing. It aims to identify named entities with specific meanings, such as names, address names, and organization names. Relation extraction refers to extracting the relationship between entities from text, which can help us understand the connection between entities in the text. By fusing the named entity annotation results and relationship extraction results of the initial text data of the power equipment, the complex relationship between the power equipment can be fully utilized to improve the management efficiency of the power equipment data. However, before the named entity annotation of the initial text data, it is necessary to use a pre-trained model to extract features from the initial text data of the power equipment to lay the foundation for the subsequent entity and relationship extraction.

[0026] Specifically, a forward long short-term memory network is used to perform a first encoding on each initial text data to obtain a first hidden layer representation of each initial text data; wherein the first hidden layer representation reflects the semantics and contextual relationship of the corresponding initial text data from front to back; a backward long short-term memory network is used to perform a second encoding on each initial text data to obtain a second hidden layer representation of each initial text data; wherein the second hidden layer representation reflects the semantics and contextual relationship of the corresponding initial text data from back to front.

[0027] The vector obtained by concatenating the first hidden layer representation and the corresponding second hidden layer representation of each initial text data is used as the hidden layer feature vector of each initial text data. The BERT model is used to perform sentence encoding on the hidden layer feature vector to obtain each semantic feature vector.

[0028] The first weight of the BERT model is adjusted based on the degree of association between all word vectors of each initial text data; wherein the word vector includes a first hidden layer representation and a second hidden layer representation.

[0029] It should be noted that each initial text data includes at least the construction process, technical parameters, selection principles, personnel configuration and equipment usage specifications of the equipment. The BERT model is a pre-trained language model based on the Transformer architecture. Through bidirectional context modeling and pre-training-fine-tuning paradigm, it significantly improves the performance of natural language processing tasks. The bidirectional long short-term memory network model includes a forward long short-term memory network and a backward long short-term memory network, which is good at capturing local dependencies in sequences. The BERT model and the bidirectional long short-term memory network model are well-known technologies and will not be described in detail in this embodiment.

[0030] Step S2: Use conditional random fields to perform sequence labeling on all semantic feature vectors to obtain various labeling results.

[0031] It should be noted that named entity recognition aims to identify specific types of entities from text, which helps to extract important entity information, such as equipment name, fault type, location, etc., so as to provide support for subsequent data management, analysis and decision-making. In order to efficiently manage power equipment data, it is necessary to first extract the named entity annotation results of the initial text data of the power equipment.

[0032] Specifically, all semantic feature vectors are sequence labeled using conditional random fields to obtain various labeling results.

[0033] It should be noted that the annotation result reflects the named entity annotation result of the initial text data. The conditional random field is a probabilistic graph model for annotation tasks, which is widely used in named entity recognition, part-of-speech tagging, word segmentation and other tasks in natural language processing. The conditional random field technology is a well-known technology and will not be described in detail in this embodiment.

[0034] Step S3: Perform dependency syntactic analysis on each semantic feature vector to obtain a dependency type matrix and an adjacency matrix; use a graph convolutional network model to perform feature extraction on the first fusion result of the semantic feature vector, the dependency type matrix and the adjacency matrix; use a classifier to process the feature extraction results to obtain various relationship extraction results.

[0035] Specifically, dependency syntactic analysis is performed on each semantic feature vector to obtain a dependency type matrix and an adjacency matrix. A graph convolutional network model is used to perform feature extraction on the first fusion result of the semantic feature vector, the dependency type matrix and the adjacency matrix. A classifier is used to process the feature extraction results to obtain various relationship extraction results.

[0036] It should be noted that the dependency type matrix is ​​used to represent the dependency relationship type between words in a sentence; the adjacency matrix is ​​used to represent the connection relationship between words in a sentence. The relationship extraction result reflects the relationship between power equipment. The dependency syntactic analysis aims to analyze the grammatical relationship between words in a sentence and construct a dependency structure tree of the sentence. The core is to determine the dependency relationship between each word in the sentence and other words; the graph convolutional network model is a deep learning model for processing graph structured data. The graph structure consists of nodes and edges. Nodes represent entities, and edges represent the relationship between entities. The graph convolutional network model and dependency syntactic analysis are both well-known technologies, and this embodiment will not be described in detail here.

[0037] It should be further explained that by using a graph convolutional network model to perform feature extraction on the first fusion result of the semantic feature vector, dependency type matrix and adjacency matrix, not only the dependency relationship of the context words in the initial text data is considered, but also the relationship type is included in the context features. Different weights are assigned to the context features to replace the weights of the traditional standard graph convolutional network model, which helps to filter out noise from syntactic knowledge and highlight the value of key information, thereby improving the model's ability to recognize important information.

[0038] Specifically, the dependency type matrix and the adjacency matrix are fused using an attention mechanism; and the second weight of the graph convolutional network model is adjusted according to the result of the fusion process.

[0039] Step S4: using a graph database to store the second fusion result of all the annotation results and all the relationship extraction results to obtain a power equipment knowledge base.

[0040] It should be noted that traditional entity recognition and relationship extraction methods often find it difficult to simultaneously process multiple entities and multiple relationships in a sentence. In order to improve the management efficiency of power equipment data, it is necessary to integrate the named entity labeling results and relationship extraction results of the initial text data of the power equipment.

[0041] Specifically, a graph database is used to store the second fusion result of all the annotation results and all the relationship extraction results to obtain a power equipment knowledge base.

[0042] It should be noted that the power equipment knowledge base includes nodes, attributes and relationships; the connection between nodes and attributes is one-to-many; the connection between nodes and relationships is one-to-many. A graph database is a database specifically used to store and query graph structure data. Graph databases are well-known technologies and will not be described in detail in this embodiment.

[0043] Compared with the prior art, the power equipment data management method provided by the embodiment of the present invention has the following beneficial effects: The BERT model is used to perform sentence encoding on the initial text data of each power equipment in the target power system to obtain various semantic feature vectors. The first weight of the BERT model is adjusted based on the degree of association between all word vectors of each initial text data, which helps to obtain important features in the initial text data; all semantic feature vectors are sequence labeled using conditional random fields to obtain various labeling results; dependency syntactic analysis is performed on each semantic feature vector to obtain a dependency type matrix and an adjacency matrix, and a graph convolutional network model is used to extract features from the first fusion result of the semantic feature vector, dependency type matrix and adjacency matrix, which not only considers the dependency relationship of context words in the initial text data, but also incorporates the relationship type into the context features. In order to measure different The weight of the contextual features is used to fuse the dependency type matrix and the adjacency matrix with the attention mechanism. The second weight of the graph convolutional network model is adjusted according to the result of the fusion processing. Different weights are assigned to the contextual features to replace the weights of the traditional standard graph convolutional network model. This helps to filter out noise from syntactic knowledge and highlight the value of key information, thereby improving the model's ability to recognize important information. The classifier is used to process the results of feature extraction to obtain various relationship extraction results; the graph database is used to store the second fusion results of all annotation results and all relationship extraction results to obtain a power equipment knowledge base. The graph database is used to store the power equipment data in an integrated manner, making full use of the complex relationships between power equipment and improving the management efficiency of power equipment data.

[0044] Another embodiment of the present invention provides an electric power equipment data management system, comprising: A data encoding module 11 is used to use a BERT model to perform sentence encoding on the acquired initial text data of each power device in the target power system to obtain each semantic feature vector; The entity labeling module 12 is used to perform sequence labeling on all semantic feature vectors using a conditional random field to obtain various labeling results; The relationship extraction module 13 is used to perform dependency syntactic analysis on each semantic feature vector to obtain a dependency type matrix and an adjacency matrix; use a graph convolutional network model to perform feature extraction on the first fusion result of the semantic feature vector, the dependency type matrix and the adjacency matrix; use a classifier to process the result of feature extraction to obtain each relationship extraction result; The database construction module 14 is used to use a graph database to store the second fusion result of all the annotation results and all the relationship extraction results to obtain a power equipment knowledge base.

[0045] It should be noted that the entity labeling module and the relationship extraction module adopt a parameter sharing strategy so that the entity labeling module and the relationship extraction module can work together under the same network framework.

[0046] Yet another embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein when a device where the computer-readable storage medium is located executes the computer program, an electric power equipment data management method as described above is implemented.

[0047] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for managing power equipment data, characterized in that: The method comprises: The BERT model is used to perform sentence encoding on the initial text data of each power device in the target power system to obtain each semantic feature vector; Performing sequence labeling on all the semantic feature vectors using a conditional random field to obtain various labeling results; Perform dependency syntactic analysis on each of the semantic feature vectors to obtain a dependency type matrix and an adjacency matrix; use a graph convolutional network model to perform feature extraction on a first fusion result of the semantic feature vector, the dependency type matrix, and the adjacency matrix; use a classifier to process the result of the feature extraction to obtain each relationship extraction result; A graph database is used to store the second fusion result of all the annotation results and all the relationship extraction results to obtain a power equipment knowledge base.

2. A method for managing power equipment data according to claim 1, characterized in that: The BERT model is used to perform sentence encoding on the acquired initial text data of each power device in the target power system to obtain various semantic feature vectors, including: Using a forward long short-term memory network to perform a first encoding on each of the initial text data, to obtain a first hidden layer representation of each of the initial text data; wherein the first hidden layer representation reflects the semantics and contextual relationship of the corresponding initial text data from front to back; Performing a second encoding on each of the initial text data using a backward long short-term memory network to obtain a second hidden layer representation of each of the initial text data; wherein the second hidden layer representation reflects the semantics and contextual relationship of the corresponding initial text data from back to front; The vector obtained by concatenating the first hidden layer representation of each of the initial text data and the corresponding second hidden layer representation is used as the hidden layer feature vector of each of the initial text data; The BERT model is used to perform sentence encoding on the hidden layer feature vector to obtain various semantic feature vectors.

3. A method for managing power equipment data according to claim 2, characterized in that: The BERT model is used to perform sentence encoding on the acquired initial text data of each electric device in the target electric power system to obtain each semantic feature vector, and further includes: Adjusting the first weight of the BERT model based on the degree of association between all word vectors of each of the initial text data; The word vector includes the first hidden layer representation and the second hidden layer representation.

4. A method for managing power equipment data according to claim 1, characterized in that: The dependency type matrix is ​​used to represent the dependency relationship type between words in a sentence; The adjacency matrix is ​​used to represent the connection relationship between words in a sentence.

5. A method for managing power equipment data according to claim 1, characterized in that: The step of using a graph convolutional network model to perform feature extraction on a first fusion result of the semantic feature vector, the dependency type matrix, and the adjacency matrix further includes: Using an attention mechanism to fuse the dependency type matrix and the adjacency matrix; The second weight of the graph convolutional network model is adjusted according to the result of the fusion processing.

6. A method for managing power equipment data according to claim 1, characterized in that: Each of the initial text data at least includes the construction process, technical parameters, selection principles, personnel configuration and equipment use specifications of the equipment.

7. A method for managing power equipment data according to claim 1, characterized in that: The electric power equipment knowledge base includes nodes, attributes and relationships; The connection between the node and the attribute is one-to-many; The connection mode between the nodes and the relationships is one-to-many.

8. A power equipment data management system, characterized in that: The system comprises: A data encoding module is used to use a BERT model to perform sentence encoding on the acquired initial text data of each power device in the target power system to obtain each semantic feature vector; An entity labeling module, used to perform sequence labeling on all the semantic feature vectors using a conditional random field to obtain various labeling results; A relationship extraction module is used to perform dependency syntactic analysis on each of the semantic feature vectors to obtain a dependency type matrix and an adjacency matrix; use a graph convolutional network model to perform feature extraction on a first fusion result of the semantic feature vector, the dependency type matrix and the adjacency matrix; use a classifier to process the result of the feature extraction to obtain each relationship extraction result; The database construction module is used to use a graph database to store the second fusion result of all the annotation results and all the relationship extraction results to obtain a power equipment knowledge base.

9. The power equipment data management system according to claim 8, characterized in that: The entity labeling module and the relationship extraction module adopt a parameter sharing strategy so that the entity labeling module and the relationship extraction module can work together under the same network framework.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein when the device where the computer-readable storage medium is located executes the computer program, the power equipment data management method according to any one of claims 1 to 7 is implemented.

Images