Nuclear power information recommendation method, system and computer device based on knowledge graph

Abstract

The present invention discloses a nuclear power information recommendation method, system and computer device based on a knowledge graph, which relates to the fields of nuclear power human factors engineering and control system design. The method includes: obtaining data in a nuclear power plant system, and constructing a nuclear power knowledge graph based on the data in the nuclear power plant system, where the nuclear power knowledge graph includes multiple nuclear power entities and the relationships between the nuclear power entities. Establishing a vector space according to the nuclear power knowledge graph, and obtaining a first similarity matrix between nuclear power entities according to the vector space. Obtaining the historical operation information of an operator, and obtaining a set of historical operation sequences according to the historical operation information of the operator. Obtaining a second similarity matrix between nuclear power entities according to the set of historical operation sequences. Generating a fused similarity matrix according to the first similarity matrix and the second similarity matrix. And, in response to the real-time operation of the operator, obtaining recommended information according to the fused similarity matrix and displaying the recommended information.

Description

Technical Field

[0001] The present invention relates to the field of nuclear power human factors engineering and control system design, and particularly to a nuclear power information recommendation method, system and computer device based on a knowledge graph. Background Art

[0002] There are a large number of systems and devices in a nuclear power plant. The DCS (Distributed Control System) control system collects a large amount of power plant process parameter information and presents it to the operator through the man-machine interface in the main control room. In a digital control room, the man-machine interface is mainly implemented in the form of screens, alarms, procedures and trends on a digital operator workstation. The screen and alarm information in the workstation are independent of each other, and the operator relies on procedures and his own knowledge to call various types of information.

[0003] However, in accident conditions, a large amount of information that appears simultaneously is likely to cause the operator's instantaneous cognitive load to be too large, thereby reducing his judgment and control efficiency of the power plant state. In addition, in the context of the application of nuclear power plant unit automatic start-stop, intelligent diagnosis and intelligent control technologies, a very high level of automation will cause some information to be opaque to the operator, and the operator's ability to master and control power plant information will be greatly reduced. Summary of the Invention

[0004] In view of the above deficiencies of the prior art, the present invention provides a nuclear power information recommendation method, system and computer device based on a knowledge graph. By processing and outputting information streams through intelligent algorithms driven by big data such as knowledge graphs, semantic recognition, and relevance ranking, the information that the operator currently needs most is presented on the screen, enabling the operator to perform monitoring tasks more efficiently.

[0005] On the one hand, the present invention provides a nuclear power information recommendation method based on a knowledge graph, including: obtaining data in a nuclear power plant system, and constructing a nuclear power knowledge graph according to the data in the nuclear power plant system. The nuclear power knowledge graph includes multiple nuclear power entities and the relationships between the nuclear power entities. Establish a vector space according to the nuclear power knowledge graph, and obtain a first similarity matrix between nuclear power entities according to the vector space. Obtain the historical operation information of the operator, and obtain a set of historical operation sequences according to the historical operation information of the operator. Obtain a second similarity matrix between nuclear power entities according to the set of historical operation sequences. Generate a fusion similarity matrix according to the first similarity matrix and the second similarity matrix. And, in response to the operator's real-time operation, obtain recommended information according to the fusion similarity matrix, and display the recommended information. The set of historical operation sequences includes multiple historical operation sequences, and the historical operation sequence is a sequence of nuclear power entities formed according to the order in which the operator clicks on the nuclear power entities. The recommended information includes a list of nuclear power entities recommended to the operator.

[0006] Specifically, obtain the data in the nuclear power plant system and construct a nuclear power knowledge graph based on the data in the nuclear power plant system, including: obtaining the data in the nuclear power plant system, and based on the data in the nuclear power plant system, obtaining the structural data and unstructured data in the nuclear power plant system. Construct a first nuclear power knowledge graph based on the structural data in the nuclear power plant system. For the unstructured data in the nuclear power plant system, use a preset deep learning algorithm to extract nuclear power knowledge and obtain a second nuclear power knowledge graph. And, fuse the first nuclear power knowledge graph and the second nuclear power knowledge graph to obtain a nuclear power knowledge graph.

[0007] Specifically, constructing a first nuclear power knowledge graph based on the structural data in the nuclear power plant system includes: based on the structural data in the nuclear power plant system, obtaining the definition of the data model of the first nuclear power knowledge graph, and the definition of the data model of the first nuclear power knowledge graph includes the definition of the domain, ontology, first nuclear power entities, and relationships in the first nuclear power knowledge graph. And, based on the definition of the data model of the first nuclear power knowledge graph, construct the first nuclear power knowledge graph based on a database.

[0008] Specifically, using a preset deep learning algorithm to extract nuclear power knowledge from the unstructured data in the nuclear power plant system to obtain a second nuclear power knowledge graph includes: based on a neural network algorithm, identify nuclear power entities from the text data of the unstructured data in the nuclear power plant system as the second nuclear power entities, and use the second nuclear power entities as the nodes in the second nuclear power knowledge graph. Divide the text of the unstructured data in the nuclear power plant system into multiple text segments according to sentences, and in multiple text segments, extract the relationships between the second nuclear power entities one by one through a relationship extraction network. And, construct the second nuclear power knowledge graph based on the relationships between the second nuclear power entities and the second nuclear power entities.

[0009] Specifically, fusing the first nuclear power knowledge graph and the second nuclear power knowledge graph to obtain a nuclear power knowledge graph includes: performing reference resolution and entity disambiguation on the knowledge in the first nuclear power knowledge graph and the second nuclear power knowledge graph, screening and merging the knowledge with similar names to obtain a nuclear power knowledge graph.

[0010] Specifically, establish a vector space based on the nuclear power knowledge graph and obtain a first similarity matrix between nuclear power entities according to the vector space, including: obtaining nuclear power entity vectors and relationship vectors according to the relationships between nuclear power entities and nuclear power entities, and establish a vector space based on the nuclear power entity vectors and relationship vectors. And, obtain a first similarity matrix between nuclear power entities according to the distances of nuclear power entity vectors in the vector space.

[0011] Specifically, obtaining the second similarity matrix between nuclear power entities according to the set of historical operation sequences includes: obtaining the behavior matrix between multiple historical operation sequences and nuclear power entities according to the set of historical operation sequences; and obtaining the second similarity matrix between nuclear power entities through cosine distance according to the behavior matrix.

[0012] Specifically, in response to the real-time operation of the operator, obtaining recommendation information according to the fusion similarity matrix includes: for any nuclear power entity, obtaining the reference entity set; obtaining the weight of each nuclear power entity in the reference entity set according to the fusion similarity matrix; for any nuclear power entity, obtaining the predicted correlation degree between any nuclear power entity and the real-time operation entity set of the operator according to the weight of each nuclear power entity in the reference entity set and the fusion similarity matrix; and for all nuclear power entities, sorting them in descending order according to the value of the predicted correlation degree with the real-time operation entity set of the operator to generate recommendation information. The reference entity set is the intersection of the similar entity set and the real-time operation entity set of the operator. The similar entity set is the set of nuclear power entities similar to any nuclear power entity in the nuclear power knowledge graph. The real-time operation entity set of the operator includes the set of nuclear power entities arranged according to the real-time click order of the operator.

[0013] Specifically, obtaining the weight of each nuclear power entity in the reference entity set according to the fusion similarity matrix includes: obtaining the similarity between each nuclear power entity in the reference entity set and any nuclear power entity, and obtaining the sum of the similarities between any nuclear power entity and all other entities in the nuclear power knowledge graph; calculating the quotient of the similarity between each nuclear power entity in the reference entity set and any nuclear power entity and the sum of the similarities between any nuclear power entity and all other entities in the nuclear power knowledge graph as the weight of each nuclear power entity in the reference entity set. All other entities are nuclear power entities in the nuclear power knowledge graph other than any nuclear power entity.

[0014] Specifically, for any nuclear power entity, obtaining the predicted correlation degree between any nuclear power entity and the real-time operation entity set of the operator according to the weight of each nuclear power entity in the reference entity set and the fusion similarity matrix includes: obtaining the attention degree of each nuclear power entity in the reference entity set according to the fusion similarity matrix; and obtaining the weighting of each nuclear power entity in the reference entity set. The weighting of each nuclear power entity in the reference entity set is the product of the weight of each nuclear power entity in the reference entity set and the attention degree of each nuclear power entity in the reference entity set; calculating the sum of the weightings of each nuclear power entity in the reference entity set as the predicted correlation degree between the real-time operation entity set of the operator and any nuclear power entity.

[0015] Second aspect, the present invention provides a nuclear power information recommendation system based on a knowledge graph, including a data acquisition unit and a data processing unit. The data acquisition unit is used to acquire data in the nuclear power plant system and construct a nuclear power knowledge graph according to the data in the nuclear power plant system. The nuclear power knowledge graph includes multiple nuclear power entities and the relationships between the nuclear power entities. The data processing unit is connected to the data acquisition unit and is used to: establish a vector space according to the nuclear power knowledge graph, and obtain a first similarity matrix between the nuclear power entities according to the vector space. Obtain the historical operation information of the operator, and obtain a set of historical operation sequences according to the historical operation information of the operator. Obtain a second similarity matrix between the nuclear power entities according to the set of historical operation sequences. Generate a fusion similarity matrix according to the first similarity matrix and the second similarity matrix. And, in response to the real-time operation of the operator, obtain recommendation information according to the fusion similarity matrix and display the recommendation information. The recommendation information includes a list of nuclear power entities recommended to the operator. The set of historical operation sequences includes multiple historical operation sequences, and the historical operation sequence is a sequence of nuclear power entities formed according to the click order of the operator on the nuclear power entities.

[0016] Third aspect, the present invention provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned nuclear power information recommendation method based on a knowledge graph are implemented.

[0017] Fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the processor executes the steps of the above-mentioned nuclear power information recommendation method based on a knowledge graph.

[0018] The beneficial effects of the present invention include: A nuclear power information recommendation method, system and computer device provided by the present invention can synthesize a large amount of scattered power plant operation knowledge into a large-scale semantic information network by constructing a knowledge graph, so as to organize the complex and scattered data in the operation process of the nuclear power plant and abstract it into a nuclear power knowledge graph. Through collaborative filtering algorithms, deep learning algorithms and recommendation algorithms, based on this nuclear power knowledge graph and the historical operation information of the operator, the information most needed by the operator is presented to the operator in real time. Description of the Drawings

[0019] Figure 1 It is a flowchart of a nuclear power information recommendation method based on a knowledge graph in an embodiment of the present invention;

[0020] Figure 2 It is a flowchart of another nuclear power information recommendation method based on a knowledge graph in an embodiment of the present invention;

[0021] Figure 3Flowchart of yet another nuclear power information recommendation method based on a knowledge graph in an embodiment of the present invention;

[0022] Figure 4 Flowchart of yet another nuclear power information recommendation method based on a knowledge graph in an embodiment of the present invention;

[0023] Figure 5 Flowchart of yet another nuclear power information recommendation method based on a knowledge graph in an embodiment of the present invention;

[0024] Figure 6 Flowchart of yet another nuclear power information recommendation method based on a knowledge graph in an embodiment of the present invention;

[0025] Figure 7 Flowchart of yet another nuclear power information recommendation method based on a knowledge graph in an embodiment of the present invention;

[0026] Figure 8 Flowchart of yet another nuclear power information recommendation method based on a knowledge graph in an embodiment of the present invention;

[0027] Figure 9 Flowchart of yet another nuclear power information recommendation method based on a knowledge graph in an embodiment of the present invention;

[0028] Figure 10 Flowchart of yet another nuclear power information recommendation method based on a knowledge graph in an embodiment of the present invention;

[0029] Figure 11 Schematic diagram of recommended information and display interface in an embodiment of the present invention;

[0030] Figure 12 Block diagram of the structure of a nuclear power information recommendation system based on a knowledge graph in an embodiment of the present invention;

[0031] Figure 13 Block diagram of the structure of a computer device in an embodiment of the present invention. Detailed implementation manners

[0032] To enable those skilled in the art to better understand the technical solutions of the present invention, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0033] It can be understood that the specific embodiments and drawings described herein are only for explaining the present invention and are not intended to limit the present invention.

[0034] It can be understood that, without conflict, the various embodiments in the present invention and the various features in the embodiments may be combined with each other.

[0035] It can be understood that, for the convenience of description, only the parts related to the present invention are shown in the drawings of the present invention, while the parts unrelated to the present invention are not shown in the drawings.

[0036] It can be understood that each unit and module involved in the embodiments of the present invention may correspond to only one entity structure, or may be composed of multiple entity structures. Alternatively, multiple units and modules may also be integrated into one entity structure.

[0037] It can be understood that, without conflict, the functions and steps marked in the flowcharts and block diagrams of the present invention may occur in a sequence different from that marked in the drawings.

[0038] It can be understood that in the flowcharts and block diagrams of the present invention, the possible system architectures, functions, and operations of the systems, devices, equipment, and methods according to the embodiments of the present invention are shown. Among them, each block in the flowchart or block diagram may represent a unit, module, program segment, or code, which contains executable instructions for implementing the specified function. Moreover, each block or combination of blocks in the block diagram and flowchart may be implemented by a hardware-based system for implementing the specified function, or may be implemented by a combination of hardware and computer instructions.

[0039] It can be understood that the units and modules involved in the embodiments of the present invention may be implemented in software or in hardware. For example, the units and modules may be located in a processor.

[0040] With the advent of the big data era, technologies for information mining and intelligent presentation of data have emerged and developed rapidly. Intelligent information presentation (recommendation / push) technology is currently commonly used in the field of daily life. It filters and recommends content that users need or are interested in for users based on factors such as collaborative filtering algorithms, deep learning algorithms, and the semantic connection between user historical operations and item semantics. And the knowledge graph, as a semantic network, can provide semantic association information between users and entities for the intelligent information presentation system. The knowledge graph stores knowledge and the semantic relationships between knowledge in the form of a graph, abstracting knowledge into the triple form of "entity-relationship-entity". By constructing a knowledge graph, a large amount of scattered power plant operation knowledge can be integrated into a large semantic information network, so as to perform complex tasks such as knowledge reasoning and content recommendation.

[0041] The knowledge graph constructs the nuclear power plant operation knowledge system from independent information such as the screens, alarms, and procedures required by the operator. By means of expert knowledge and big data, a relationship network of this information is constructed, and then the most needed information is presented to the operator in real time according to the plant conditions, fault tracing, and the operator's operation habits. The intelligent information presentation technology is displayed in the form of an information list, and the list content includes information such as the links of screens, procedures, parameters, equipment, as well as the fault tracing results and equipment status. The operator can select and open the information that he needs to monitor according to the requirements.

[0042] Studying how to organize the complex and scattered knowledge in the power plant operation, abstract it into a knowledge graph and establish an operator knowledge base to achieve intelligent information presentation is crucial for the development of power plant intelligent technology and the enhancement of the operator's control ability.

[0043] An embodiment of the present invention provides a nuclear power information recommendation method based on a knowledge graph. By constructing a knowledge graph, a large number of scattered power plant operation knowledge can be integrated into a large semantic information network to organize the complex and scattered data in the nuclear power plant operation process and abstract it into a nuclear power knowledge graph. Through collaborative filtering algorithms, deep learning algorithms, and recommendation algorithms, based on the nuclear power knowledge graph and the historical operation information of the operator, the information most needed by the operator is presented to the operator in real time.

[0044] As Figure 1 shown, an embodiment of the present invention provides a nuclear power information recommendation method based on a knowledge graph, including step 101 to step 106.

[0045] Step 101, obtain the data in the nuclear power plant system, and construct a nuclear power knowledge graph according to the data in the nuclear power plant system.

[0046] It can be understood that the nuclear power knowledge graph includes multiple nuclear power entities and the relationships between nuclear power entities. Exemplarily, the data in the nuclear power plant system includes the data in the HPR1000 RCS, RCV, and RHR systems. For example, the data in the nuclear power plant system can be data such as nuclear power engineering documents, design documents, and procedure information. In order to define the unique knowledge graph structure in the nuclear power field, it is necessary to analyze and organize a large amount of document data to establish a nuclear power plant knowledge graph data model for continuing to construct the lower-level relationships.

[0047] In some embodiments, as Figure 2 shown, the implementation manner of step 101 may include step 201 to step 204.

[0048] Step 201, obtain the data in the nuclear power plant system, and according to the data in the nuclear power plant system, obtain the structural data and non-structural data in the nuclear power plant system.

[0049] Step 202: Construct a first nuclear power knowledge graph based on the structural data in the nuclear power plant system.

[0050] In some embodiments, the data model can be defined in a top-down manner, that is, starting from the top-level concepts and gradually refining them downward to form a clearly structured taxonomic hierarchy, build a knowledge graph framework, and then add entities one by one to the knowledge graph framework to form a relationship network.

[0051] For example, when defining a data model in a top-down manner, you can first abstract the concept of "domain". The concept of domain overrides all models, and the definition needs to be as specific as possible, and the domains must be independent and non-overlapping. In addition, the concept of domain needs to be able to be further abstracted downwards to facilitate further sorting. After the domain is divided, it is also necessary to sort out concepts such as "ontology", "entity", and "relationship" in turn. "Ontology" is used to describe the essence of things. It is a definition of real entity types, entity attributes, and the relationships between entities. "Entity" is a thing in the objective world and is the basic unit that constitutes the knowledge graph, such as a device or a picture. "Relationship" refers to the connection between two or more "entities". The knowledge graph connects countless scattered "entity" nodes through "relationships" and describes the semantic relationship between "entities".

[0052] After the data model is defined, the first nuclear power knowledge graph framework can be constructed using the structural data in the nuclear power plant system. Figure 3 As shown, the specific implementation method of step 202 may include steps 301 and 302.

[0053] Step 301: Obtain the definition of the data model of the first nuclear power knowledge graph based on the structural data in the nuclear power plant system.

[0054] It can be understood that the definition of the data model of the first nuclear power knowledge graph includes the definition of the domain in the first nuclear power knowledge graph, the definition of the ontology, the definition of the second nuclear power entity and the definition of the relationship.

[0055] Step 302: According to the definition of the data model of the first nuclear power knowledge graph, construct the first nuclear power knowledge graph based on the database.

[0056] It can be understood that the specific method of step 302 can be: according to the definition of the data model of the first nuclear power knowledge graph, the structural data in the nuclear power plant system is organized into several "first nuclear power entity-relationship-first nuclear power entity" triplets, and the first nuclear power knowledge graph is constructed based on the database.

[0057] Exemplarily, the method for constructing the first nuclear power knowledge graph based on a database can be as follows: Use the graph database Neo4j for data storage and the construction of the knowledge network. Neo4j is a NoSQL database, and its internal data structure is in the form of a graph, consisting of nodes, edges, and attributes. The directed graph created based on Neo4j is the first nuclear power knowledge graph.

[0058] After establishing the first nuclear power knowledge graph using structured data such as tables and databases, it is also necessary to use unstructured data, such as text files like engineering documents and procedures, to expand and supplement a large amount of content and relationships in the first nuclear power knowledge graph. The processing of unstructured data can use methods of entity extraction and relationship extraction based on deep learning. The above embodiments illustrate how to establish the first nuclear power knowledge graph, and the following will give an exemplary illustration of how to expand and supplement a large amount of content and relationships in the first nuclear power knowledge graph.

[0059] Step 203: According to the unstructured data in the nuclear power plant system, adopt a preset deep learning algorithm to extract nuclear power knowledge and obtain the second nuclear power knowledge graph.

[0060] It can be understood that the extraction of nuclear power knowledge includes entity extraction and relationship extraction. The purpose of entity extraction is to automatically identify entities from the text dataset to establish the "nodes" in the second nuclear power knowledge graph. Relationship extraction refers to binary relationship extraction, and the purpose is to extract the relationships between entities in the text. For example, extract the relationships between parameters, equipment, screens, and procedures from the alarm card file. The preset deep learning algorithm can be selected according to the actual situation. In some embodiments, as Figure 4 shown, the specific implementation method of step 203 can include steps 401 to 403.

[0061] Step 401: Based on the neural network algorithm, identify nuclear power entities from the text data of the unstructured data in the nuclear power plant system as the second nuclear power entities, and use the second nuclear power entities as the nodes in the second nuclear power knowledge graph.

[0062] Embodiments of the present invention use a neural network-based NER (Named Entity Recognition) algorithm, which uses a bidirectional recurrent neural network (BiRNN), and adds a conditional random field (CRF) on the basis of the recurrent neural network to adjust the order of entity recognition results, improving the recognition accuracy. On this basis, a BERT word vector language model is used to encode text information, enhancing the generalization ability of the model. Compared with NER algorithms based on traditional machine learning, neural networks have stronger feature expression capabilities, can avoid manually customized features, and fully learn the context relationships of entities.

[0063] Step 402: Divide the text of the unstructured data in the nuclear power plant system into multiple segments according to sentences. Among the multiple segments, the relationships between the second nuclear power entities are extracted one by one through a relationship extraction network.

[0064] Embodiments of the present invention use sentence-level relationship extraction according to the characteristics of the data source, and limit the domain according to the extracted entities, restricting the relationship scope to the existing relationship types. For example, the relationship extraction network is based on the text classification model TextCNN, defines the output of the text classification model as different types of relationships, the network accepts the entire text sentence as input, and outputs the classification result of the relationships between entities, which is equivalent to performing end-to-end relationship extraction.

[0065] Step 403: Construct a second nuclear power knowledge graph according to the relationships between the second nuclear power entities.

[0066] It can be understood that after obtaining the second nuclear power entities and the relationships between the second nuclear power entities, a second nuclear power knowledge graph can be constructed.

[0067] Step 204: Perform knowledge fusion on the first nuclear power knowledge graph and the second nuclear power knowledge graph to obtain a nuclear power knowledge graph.

[0068] In some embodiments, the specific implementation method of step 204 may include: performing coreference resolution and entity disambiguation on the knowledge in the first nuclear power knowledge graph and the second nuclear power knowledge graph, screening and merging the knowledge with similar names to obtain a nuclear power knowledge graph.

[0069] Understandably, knowledge with similar names refers to knowledge that has different or similar names but the same actual meaning. In language, or generally in linguistics and our daily language, when using an abbreviation or a substitute name to replace a certain word that has already appeared above, this situation in linguistics is called "anaphora", that is, reference. Anaphora can avoid problems such as bloated and redundant sentences caused by the repeated appearance of the same word; however, it also causes the problem of "ambiguous reference" due to this kind of omission. Formally, the process of dividing different referents representing the same entity into an equivalent set is called anaphora resolution. The ambiguity of an entity refers to the fact that an entity referent can correspond to multiple real-world entities. For example, "apple" can refer to a fruit, a computer brand, or other entities. Determining the real-world entity that an entity referent points to is entity disambiguation. The methods of anaphora resolution and entity disambiguation can fuse knowledge with similar names so that knowledge with the same actual meaning shares the same name.

[0070] In this way, the first nuclear power knowledge graph derived from structured data and the second nuclear power knowledge graph derived from unstructured data can be fused into the same knowledge graph, that is, the nuclear power knowledge graph.

[0071] Step 102: Establish a vector space based on the nuclear power knowledge graph, and obtain the first similarity matrix between nuclear power entities according to the vector space.

[0072] In order to present the required information to the operator, it is necessary to quantify the information, quantify the semantic information into an index that can measure the degree of correlation tightness, and establish a vector space model. First, quantify the entities and relationships in the knowledge graph into vectors, and then represent the information that needs to be inferred or used as information containing entity keywords. Based on the vector space model, use the nuclear power knowledge graph to deeply mine the semantic information therein, so as to accurately quantify the required information into an information vector.

[0073] In some embodiments, as Figure 5 shown, the specific implementation method of step 102 may include step 501 to step 502.

[0074] Step 501: Obtain nuclear power entity vectors and relationship vectors according to the relationship between nuclear power entities and nuclear power entities, and establish a vector space according to the nuclear power entity vectors and relationship vectors.

[0075] Exemplarily, the entities and relationships in the first nuclear power knowledge graph can be quantified based on the TransE algorithm. While preserving semantics, the entities and relationships in the first nuclear power knowledge graph are embedded into a continuous dense low-dimensional vector space. The structured knowledge in the first nuclear power knowledge graph is represented as an undirected graph G. The set of entities, i.e., nodes, is represented as V, and the set of relationships, i.e., edges, is represented as E. In the vector space of the TransE algorithm, if two entity vectors can be connected by a relationship vector, that is, the result of adding the source entity vector and the relationship vector is closer to the target entity vector, it indicates that the vectorized representation of the entity and the relationship is more accurate. By generating an objective function based on this constraint and optimizing the objective function, an entity vectorization network is trained. This network can map entities with similar semantics in the knowledge graph to corresponding positions in the vector space. The above method uses the constructed knowledge graph in the nuclear power field to perform vectorized representation on each knowledge node and map it into the vector space, thereby providing quantitative data reference for the intelligent information presentation model.

[0076] Step 502: Obtain the first similarity matrix between nuclear power entities according to the distances between nuclear power entity vectors in the vector space.

[0077] It can be understood that after information is vectorized, the similarity between every two nuclear power entities can be calculated through the distances in the vector space. According to the similarity between every two nuclear power entities, the first similarity matrix S1 can be obtained.

[0078] Step 103: Obtain the historical operation information of the operator and obtain a set of historical operation sequences according to the historical operation information of the operator.

[0079] It can be understood that the set of historical operation sequences includes multiple historical operation sequences. A historical operation sequence is a sequence of nuclear power entities formed according to the click order of the operator on nuclear power entities. Exemplarily, according to the operation purpose of the operator, the historical operation information of the operator is divided into m historical operation sequences. Taking the number of nuclear power entities as n as an example, the set U of historical operation sequences is U = {U1, U2,..., Ui,..., Um}, where i is an integer and the value range of i is from 1 to m. The set I of n nuclear power entities is I = {I1, I2,..., Ij,..., In}, where j is an integer and the value range of j is from 1 to n.

[0080] Step 104: Obtain the second similarity matrix between nuclear power entities according to the set of historical operation sequences.

[0081] In some embodiments, as Figure 6 shown, the specific implementation method of step 104 may include step 601 to step 602.

[0082] Step 601: Obtain the behavior matrix between multiple historical operation sequences and nuclear power entities according to the set of historical operation sequences.

[0083] Let Rij denote the number of occurrences of the nuclear power entity Ij in the historical operation sequence Ui. Rij can represent the degree of attention of the historical operation sequence Ui to the nuclear power entity Ij. Therefore, the behavior matrix R between the set of historical operation sequences U and the nuclear power entity I can be generated, as shown in Equation (1).

[0084]

[0085] Step 602: Obtain the second similarity matrix between nuclear power entities through the cosine distance according to the behavior matrix.

[0086] According to Equation (1), each nuclear power entity I j can be represented as an m-dimensional vector composed of the degrees of attention of the historical operation sequence U i to the nuclear power entity I j , as shown in Equation (2).

[0087] I j =(R 1j , R 2j ,..., R mj ) T (2)

[0088] Then, the second similarity s2(I k , I j ) between the nuclear power entity I k and the nuclear power entity I j can be obtained according to the cosine distance, as shown in Equation (3). k is an integer, and the value range of k is from 1 to n.

[0089]

[0090] According to the second similarity s2(I k , I j ) between the first nuclear power entity I i and the first nuclear power entity I j , the second similarity matrix S2 can be obtained.

[0091] Step 105: Generate a fused similarity matrix according to the first similarity matrix and the second similarity matrix.

[0092] It can be understood that by performing weighted fusion on the first similarity matrix S1 and the second similarity matrix S2, a fused similarity matrix S can be generated. Then, when the operator is performing real-time operations, recommended information can be displayed to the operator through the fused similarity matrix S.

[0093] Step 106: In response to the real-time operation of the operator, obtain recommendation information according to the fusion similarity matrix, and display the recommendation information.

[0094] Understandably, the recommendation information includes a list of nuclear power entities recommended to the operator.

[0095] In some embodiments, as Figure 7 shown, the implementation method of step 106 may include steps 701 to 704.

[0096] Step 701: For any nuclear power entity, obtain a set of reference entities.

[0097] In step 701, the set of reference entities is the intersection of the set of similar entities and the set of real-time operation entities of the operator; the set of similar entities is the set of nuclear power entities similar to any nuclear power entity in the nuclear power knowledge graph; the set of real-time operation entities of the operator includes the set of nuclear power entities arranged according to the real-time click order of the operator.

[0098] Understandably, as Figure 8 shown, for any nuclear power entity, the method for obtaining the set of reference entities may include steps 7011 to 7013.

[0099] Step 7011: Obtain the set of nuclear power entities similar to any nuclear power entity in the nuclear power knowledge graph as the set of similar entities.

[0100] Step 7012: Obtain the set of real-time operation entities of the operator according to the real-time click order of the operator.

[0101] Step 7013: Take the intersection of the set of similar entities and the set of real-time operation entities of the operator, and obtain the set of reference entities.

[0102] Step 702: According to the fusion similarity matrix, obtain the weight of each nuclear power entity in the set of reference entities.

[0103] In some embodiments, as Figure 9 shown, the implementation method of step 702 may include steps 7021 to 7023.

[0104] Step 7021: Obtain the similarity between each nuclear power entity in the set of reference entities and any nuclear power entity.

[0105] Step 7022: Obtain the sum of the similarities between any nuclear power entity and all other entities in the nuclear power knowledge graph.

[0106] Understandably, in step 7022, all other entities are nuclear power entities in the nuclear power knowledge graph other than any nuclear power entity.

[0107] Step 7023: Obtain the quotient of the similarity between each nuclear power entity in the reference entity set and any nuclear power entity and the sum of the similarities between all other entities in the nuclear power knowledge graph and any nuclear power entity, which is the weight of each nuclear power entity in the reference entity set.

[0108] Step 703: For any nuclear power entity, according to the weights of each nuclear power entity in the reference entity set and the fusion similarity matrix, obtain the predicted correlation degree between any nuclear power entity and the set of real-time operation entities of the operator.

[0109] In some embodiments, as Figure 10 shown, the implementation method of step 703 may include steps 7031 to 7033.

[0110] Step 7031: According to the fusion similarity matrix, obtain the degree of attention of each nuclear power entity in the reference entity set.

[0111] Step 7032: Obtain the product of the weight of each nuclear power entity in the reference entity set and the degree of attention of each nuclear power entity in the reference entity set, which is the weighting of each nuclear power entity in the reference entity set.

[0112] Step 7033: Obtain the sum of the weightings of each nuclear power entity in the reference entity set, which is the predicted correlation degree between the set of real-time operation entities of the operator and any nuclear power entity.

[0113] Step 704: For all nuclear power entities, sort them in descending order according to the values of the predicted correlation degree with the set of real-time operation entities of the operator to generate recommendation information.

[0114] Exemplarily, as Figure 11 shown, the recommendation information includes nuclear power entity 1 with a predicted correlation degree of 0.95, nuclear power entity 2 with a predicted correlation degree of 0.82, and nuclear power entity 3 with a predicted correlation degree of 0.56. To clearly present the recommendation information to the operator, the embodiments of the present invention design an interaction logic and an interactive graphical interface, and display the recommendation information in a graphical and text combination form in a specific area (for example, area Z1) on the control system screen. When the system is in use, the recommendation list is calculated and updated in real time after each operation of the operator, and the graphical interface is refreshed with the recommendation information list as the original data to complete the intelligent information presentation.

[0115] Embodiments of the present invention perform recommendation ranking on the content required by the operator based on the collaborative filtering algorithm of the knowledge graph, and design an intelligent information presentation system to present the recommended content, so as to facilitate the operator to obtain information quickly and conveniently. In the embodiments of the present invention, the recommendation ranking algorithm combines the nuclear power knowledge graph and the collaborative filtering algorithm, and introduces the relationship information from the knowledge graph into the traditional collaborative filtering algorithm to make up for the defects of the collaborative filtering algorithm. Among them, the traditional collaborative filtering algorithm mainly collects the historical operation habit information of the operator, while the knowledge graph integrates more interpretable nuclear power professional knowledge in the information recommendation process. The integration of these two data sources makes the information recommendation result more accurate.

[0116] Embodiments of the present invention study the intelligent information presentation model, and based on the knowledge association in the knowledge graph and the large amount of historical interaction big data of the operator collected by the nuclear power plant, learn the control rules of the nuclear power plant system and the operation habits of the operator, and adaptively adjust the currently displayed information, so as to realize the intelligent push of the information most needed by the operator.

[0117] As Figure 12 shown, embodiments of the present invention provide a nuclear power information recommendation system 1200 based on a knowledge graph, including a data acquisition unit 1201 and a data processing unit 1202. The data acquisition unit 1201 is used to acquire data in the nuclear power plant system, and construct a nuclear power knowledge graph according to the data in the nuclear power plant system. The nuclear power knowledge graph includes multiple nuclear power entities and the relationships between nuclear power entities. The data processing unit 1202 is connected to the data acquisition unit 1201 and is used to: establish a vector space according to the nuclear power knowledge graph, and obtain a first similarity matrix between nuclear power entities according to the vector space. Acquire the historical operation information of the operator, and obtain a set of historical operation sequences according to the historical operation information of the operator. Obtain a second similarity matrix between nuclear power entities according to the set of historical operation sequences. Generate a fusion similarity matrix according to the first similarity matrix and the second similarity matrix. And, in response to the real-time operation of the operator, obtain recommended information according to the fusion similarity matrix, and display the recommended information. The recommended information includes a list of nuclear power entities recommended to the operator. The set of historical operation sequences includes multiple historical operation sequences, and the historical operation sequence is a sequence of nuclear power entities formed according to the click order of the operator on the nuclear power entities.

[0118] For the specific solution and beneficial effects of the nuclear power information recommendation system 1200 based on the knowledge graph provided by the embodiments of the present invention, reference can be made to the relevant descriptions of the nuclear power information recommendation method based on the knowledge graph provided in the above embodiments, and details are not described herein again.

[0119] As Figure 13As shown in the figure, an embodiment of the present invention provides a computer device 1300, which includes a memory 1301 and a processor 1302. The memory 1301 is connected to the processor 1302. The memory 1301 stores a computer program, and when the processor 1302 executes the computer program, the steps of the above-mentioned nuclear power information recommendation method based on the knowledge graph are implemented.

[0120] For the specific solution and beneficial effects of the computer device provided by the embodiment of the present invention, reference may be made to the relevant descriptions of the nuclear power information recommendation method based on the knowledge graph provided in the above embodiments, which will not be elaborated here.

[0121] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the processor executes the steps of the above-mentioned nuclear power information recommendation method based on the knowledge graph.

[0122] For the specific solution and beneficial effects of the computer-readable storage medium provided by the embodiment of the present invention, reference may be made to the relevant descriptions of the nuclear power information recommendation method based on the knowledge graph provided in the above embodiments, which will not be elaborated here.

[0123] It can be understood that the above embodiments are merely exemplary embodiments adopted to illustrate the principles of the present invention. However, the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also regarded as the protection scope of the present invention.

Claims

1. A nuclear power information recommendation method based on a knowledge graph, characterized in that Including: Obtain data in the nuclear power plant system, and construct a nuclear power knowledge graph based on the data in the nuclear power plant system, where the nuclear power knowledge graph includes multiple nuclear power entities and the relationships between the nuclear power entities; Establish a vector space according to the nuclear power knowledge graph, and obtain a first similarity matrix between the nuclear power entities according to the vector space; Obtain the historical operation information of the operator, and obtain a set of historical operation sequences according to the historical operation information of the operator; Obtain a second similarity matrix between the nuclear power entities according to the set of historical operation sequences; the set of historical operation sequences includes multiple historical operation sequences, and the historical operation sequence is a sequence of the nuclear power entities formed according to the click order of the operator on the nuclear power entities; Generate a fusion similarity matrix according to the first similarity matrix and the second similarity matrix; And In response to the real-time operation of the operator, obtain recommended information according to the fusion similarity matrix, and display the recommended information; The recommended information includes a list of the nuclear power entities recommended to the operator.

2. The method for recommending nuclear power information based on a knowledge graph according to claim 1, wherein The obtaining data in the nuclear power plant system and constructing a nuclear power knowledge graph according to the data in the nuclear power plant system includes: Obtain data in the nuclear power plant system, and according to the data in the nuclear power plant system, obtain the structural data and non-structural data in the nuclear power plant system; Construct a first nuclear power knowledge graph according to the structural data in the nuclear power plant system; Extract nuclear power knowledge from the non-structural data in the nuclear power plant system by using a preset deep learning algorithm to obtain a second nuclear power knowledge graph; and Perform knowledge fusion on the first nuclear power knowledge graph and the second nuclear power knowledge graph to obtain the nuclear power knowledge graph.

3. The method for recommending nuclear power information based on a knowledge graph according to claim 2, wherein The constructing a first nuclear power knowledge graph according to the structural data in the nuclear power plant system includes: According to the structural data in the nuclear power plant system, obtain the definition of the data model of the first nuclear power knowledge graph, where the definition of the data model of the first nuclear power knowledge graph includes the definition of the domain, ontology, definition of the first nuclear power entity, and definition of the relationship in the first nuclear power knowledge graph; and Construct the first nuclear power knowledge graph based on the database according to the definition of the data model of the first nuclear power knowledge graph.

4. The method for recommending nuclear power information based on a knowledge graph according to claim 2, wherein The extracting nuclear power knowledge from the non-structural data in the nuclear power plant system by using a preset deep learning algorithm to obtain a second nuclear power knowledge graph includes: Based on the neural network algorithm, identify nuclear power entities from the text data of the non-structural data in the nuclear power plant system as the second nuclear power entities, and use the second nuclear power entities as the nodes in the second nuclear power knowledge graph; Divide the text of the non-structural data in the nuclear power plant system into multiple text segments according to sentences, and in the multiple text segments, extract the relationships between the second nuclear power entities one by one through a relationship extraction network; and Construct the second nuclear power knowledge graph according to the second nuclear power entities and the relationships between the second nuclear power entities.

5. The method for recommending nuclear power information based on a knowledge graph according to claim 2, wherein The performing knowledge fusion on the first nuclear power knowledge graph and the second nuclear power knowledge graph to obtain the nuclear power knowledge graph includes: Perform coreference resolution and entity disambiguation on the knowledge in the first nuclear power knowledge graph and the second nuclear power knowledge graph, screen and merge knowledge with similar names, and obtain the nuclear power knowledge graph.

6. The method for recommending nuclear power information based on a knowledge graph according to any one of claims 1 to 5, characterized in that The establishment of a vector space based on the nuclear power knowledge graph and the acquisition of a first similarity matrix between the nuclear power entities according to the vector space include: Obtain nuclear power entity vectors and relationship vectors according to the relationship between the nuclear power entities and the nuclear power entities, and establish a vector space according to the nuclear power entity vectors and the relationship vectors; and Obtain a first similarity matrix between the nuclear power entities according to the distances of the nuclear power entity vectors in the vector space.

7. The method for recommending nuclear power information based on a knowledge graph according to any one of claims 1 to 5, characterized in that, The acquisition of a second similarity matrix between the nuclear power entities according to the set of historical operation sequences includes: Obtain a behavior matrix between multiple historical operation sequences and the nuclear power entities according to the set of historical operation sequences; and Obtain the second similarity matrix between the nuclear power entities through cosine distance according to the behavior matrix.

8. The method for recommending nuclear power information based on a knowledge graph according to any one of claims 1 to 5, characterized in that The obtaining of recommended information according to the fusion similarity matrix in response to the real-time operation of the operator includes: For any one of the nuclear power entities, obtain a reference entity set; the reference entity set is the intersection of the similar entity set and the real-time operation entity set of the operator; the similar entity set is the set of nuclear power entities similar to any one of the nuclear power entities in the nuclear power knowledge graph; the real-time operation entity set of the operator includes the set of nuclear power entities arranged according to the real-time click order of the operator; Obtain the weight of each nuclear power entity in the reference entity set according to the fusion similarity matrix; For any one of the nuclear power entities, obtain the predicted association degree between any one of the nuclear power entities and the real-time operation entity set of the operator according to the weight of each nuclear power entity in the reference entity set and the fusion similarity matrix; and For all the nuclear power entities, sort them in descending order according to the values of the predicted association degrees with the real-time operation entity set of the operator to generate the recommended information.

9. The method for recommending nuclear power information based on a knowledge graph according to claim 8, wherein The obtaining of the weight of each nuclear power entity in the reference entity set according to the fusion similarity matrix includes: Obtain the similarity between each nuclear power entity in the reference entity set and any one of the nuclear power entities, and obtain the sum of the similarities between any one of the nuclear power entities and all other entities in the nuclear power knowledge graph; all other entities are nuclear power entities in the nuclear power knowledge graph other than any one of the nuclear power entities; Calculate the quotient of the similarity between each nuclear power entity in the reference entity set and any one of the nuclear power entities and the sum of the similarities between any one of the nuclear power entities and all other entities in the nuclear power knowledge graph as the weight of each nuclear power entity in the reference entity set.

10. The method for recommending nuclear power information based on a knowledge graph according to claim 8, characterized in that, For any one of the nuclear power entities, obtaining the predicted association degree between the any one of the nuclear power entities and the real-time operation entity set of the operator according to the weight of each nuclear power entity in the reference entity set and the fusion similarity matrix includes: Obtaining the attention degree of each nuclear power entity in the reference entity set according to the fusion similarity matrix; and Obtaining the weighting of each nuclear power entity in the reference entity set, where the weighting of each nuclear power entity in the reference entity set is the product of the weight of each nuclear power entity in the reference entity set and the attention degree of each nuclear power entity in the reference entity set; Calculating the sum of the weightings of each nuclear power entity in the reference entity set as the predicted association degree between the real-time operation entity set of the operator and the any one of the nuclear power entities.

11. A nuclear power information recommendation system based on a knowledge graph, characterized in that, Including: A data acquisition unit for acquiring data in the nuclear power plant system and constructing a nuclear power knowledge graph according to the data in the nuclear power plant system, where the nuclear power knowledge graph includes a plurality of nuclear power entities and the relationships between the nuclear power entities; A data processing unit connected to the data acquisition unit and configured to: establish a vector space according to the nuclear power knowledge graph and obtain a first similarity matrix between the nuclear power entities according to the vector space; obtain the historical operation information of the operator and obtain a set of historical operation sequences according to the historical operation information of the operator; obtain a second similarity matrix between the nuclear power entities according to the set of historical operation sequences; the set of historical operation sequences includes a plurality of historical operation sequences, and the historical operation sequence is a sequence of the nuclear power entities formed according to the click order of the operator on the nuclear power entities; generate a fusion similarity matrix according to the first similarity matrix and the second similarity matrix; And, in response to the real-time operation of the operator, obtaining recommended information according to the fusion similarity matrix and displaying the recommended information; The recommended information includes a list of the nuclear power entities recommended to the operator.

12. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 10 are implemented.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the processor executes the steps of the method according to any one of claims 1 to 10.

Images