A knowledge graph determination method, apparatus, device, and storage medium

Abstract

The application discloses a kind of knowledge graph determination method, device, equipment and storage medium.The method comprises: determining preset knowledge graph, and according to the data information in power grid subsystem, construct subsystem knowledge graph;First splicing quantity and second splicing quantity are input into first preset model, to obtain the first similarity of subsystem entity and preset entity;Third splicing quantity is input into second preset model, to obtain the second similarity of subsystem attribute and preset attribute, and according to the second similarity, the third similarity of subsystem entity and preset entity is determined;According to the first similarity and the third similarity, the association between subsystem entity and preset entity is determined, and according to the association, the target knowledge graph composed of preset knowledge graph and subsystem knowledge graph is determined.The technical scheme of the embodiment of the application solves the problem of low efficiency and accuracy caused by relying on manual processing of data scattered in different subsystems during power grid evaluation data collection.

Description

Technical Field

[0001] This invention relates to the field of power technology, and in particular to a method, apparatus, device, and storage medium for determining a knowledge graph. Background Technology

[0002] With the increasing demand for new characteristics of power systems and system evaluation and diagnostic methods, especially in the context of dual carbon, decision-making for new power systems is inseparable from diagnostic evaluation. The entire power grid diagnostic evaluation chain mainly includes power grid fault diagnosis, power grid evaluation, pre-project evaluation, and post-project evaluation.

[0003] Data acquisition is fundamental to all types of power grid evaluations. For example, post-evaluation of power grid projects mainly includes data collection, the construction of an evaluation indicator system, and the design of evaluation methods. Evaluation indicator systems are primarily divided into those for traditional energy sources and those considering the characteristics of new energy sources. Evaluation indicator systems for traditional energy sources mainly include individual and comprehensive evaluations of the project implementation process, power grid technical performance, economic impact, and social and environmental impact. Evaluation systems for new energy sources require additional consideration of the sustainability of technical, financial, managerial, environmental, and social benefits.

[0004] However, most of the data is stored in different subsystems or various scattered tables within the central platform. With the development of subsystems and the continuous expansion of the central platform, this leads to the dispersed distribution of various data. Currently, the data collection phase for power grid evaluation mainly relies on manual retrieval of various database tables and attributes across different subsystems and the central platform to find the required evaluation data. This method is time-consuming, labor-intensive, and its accuracy cannot be guaranteed. Summary of the Invention

[0005] This invention provides a method, apparatus, device, and storage medium for determining knowledge graphs, in order to solve the problem of low efficiency caused by manually processing data scattered across different subsystems.

[0006] In a first aspect, embodiments of the present invention provide a method for determining a knowledge graph, comprising:

[0007] A preset knowledge graph is determined, and a subsystem knowledge graph is constructed based on the data information in the power grid subsystem. The preset knowledge graph contains multiple preset entities, and each preset entity is associated with a preset entity description, a preset attribute, and a preset attribute description. The subsystem knowledge graph contains multiple subsystem entities, and each subsystem entity is associated with a subsystem entity description, a subsystem attribute, and a subsystem attribute description.

[0008] The first concatenation amount and the second concatenation amount are input into the first preset model to obtain the first similarity between the subsystem entity and the preset entity. The first concatenation amount is the concatenation amount of the name of the subsystem entity and the embedding vector of the subsystem entity description. The second concatenation amount is the concatenation amount of the name of the preset entity and the embedding vector of the preset entity description. The first preset model includes a first preset semantic extraction model, a preset graph neural network model and a first classifier.

[0009] The third concatenation quantity is input into the second preset model to obtain the second similarity between the subsystem attribute and the preset attribute, and the third similarity between the subsystem entity and the preset entity is determined based on the second similarity. The third concatenation quantity is the concatenation quantity of the subsystem attribute, the subsystem attribute description, the preset attribute, and the preset attribute description. The second preset model includes a second preset semantic extraction model and a second classifier.

[0010] Based on the first similarity and the third similarity, the association between the subsystem entity and the preset entity is determined, and based on the association, a target knowledge graph composed of the preset knowledge graph and the subsystem knowledge graph is determined.

[0011] Secondly, embodiments of the present invention provide a knowledge graph determination apparatus, comprising:

[0012] The knowledge graph determination module is used to determine a preset knowledge graph and construct a subsystem knowledge graph based on data information in the power grid subsystem. The preset knowledge graph contains multiple preset entities, and each preset entity is associated with a preset entity description, a preset attribute, and a preset attribute description. The subsystem knowledge graph contains multiple subsystem entities, and each subsystem entity is associated with a subsystem entity description, a subsystem attribute, and a subsystem attribute description.

[0013] The first similarity determination module is used to input the first concatenation amount and the second concatenation amount into the first preset model to obtain the first similarity between the subsystem entity and the preset entity. The first concatenation amount is the concatenation amount of the name of the subsystem entity and the embedding vector of the subsystem entity description, and the second concatenation amount is the concatenation amount of the name of the preset entity and the embedding vector of the preset entity description. The first preset model includes a first preset semantic extraction model, a preset graph neural network model, and a first classifier.

[0014] The second similarity determination module is used to input the third concatenation quantity into the second preset model to obtain the second similarity between the subsystem attribute and the preset attribute, and to determine the third similarity between the subsystem entity and the preset entity based on the second similarity. The third concatenation quantity is the concatenation quantity of the subsystem attribute, the subsystem attribute description, the preset attribute, and the preset attribute description. The second preset model includes a second preset semantic extraction model and a second classifier.

[0015] The association determination module is used to determine the association between the subsystem entity and the preset entity based on the first similarity and the third similarity, and to determine the target knowledge graph composed of the preset knowledge graph and the subsystem knowledge graph based on the association.

[0016] Thirdly, embodiments of the present invention provide an electronic device, the electronic device comprising:

[0017] At least one processor;

[0018] and memory that is communicatively connected to at least one processor;

[0019] The memory stores a computer program that can be executed by at least one processor, which is executed by at least one processor to enable the at least one processor to perform the method for determining the knowledge graph described in the first aspect above.

[0020] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing computer instructions for causing a processor to execute the knowledge graph determination method described in the first aspect.

[0021] The knowledge graph determination scheme provided in this embodiment of the invention determines a preset knowledge graph and constructs a subsystem knowledge graph based on data information from a power grid subsystem. The preset knowledge graph contains multiple preset entities, each associated with a preset entity description, preset attributes, and preset attribute descriptions. The subsystem knowledge graph contains multiple subsystem entities, each associated with a subsystem entity description, subsystem attributes, and subsystem attribute descriptions. A first concatenation value and a second concatenation value are input into a first preset model to obtain a first similarity between the subsystem entity and the preset entity. The first concatenation value is the concatenation value of the name of the subsystem entity and the embedding vector of the subsystem entity description, and the second concatenation value is the concatenation value of the name of the preset entity and the embedding vector of the preset entity description. The first preset model includes a first preset semantic extraction model, a preset graph neural network model, and a first classifier. The third concatenation quantity is input into the second preset model to obtain the second similarity between the subsystem attribute and the preset attribute. Based on the second similarity, the third similarity between the subsystem entity and the preset entity is determined. The third concatenation quantity is the concatenation quantity of the subsystem attribute, the subsystem attribute description, the preset attribute, and the preset attribute description. The second preset model includes a second preset semantic extraction model and a second classifier. Based on the first similarity and the third similarity, the association relationship between the subsystem entity and the preset entity is determined. Based on the association relationship, a target knowledge graph composed of the preset knowledge graph and the subsystem knowledge graph is determined. By adopting the above technical solution, the first similarity between the subsystem entity and the preset entity can be obtained using the first preset model. Then, the second similarity between the subsystem entity and the preset entity can be obtained using the second preset model. Based on the second similarity, the third similarity between the subsystem attribute and the preset attribute can be obtained. Finally, based on the first and third similarities, a relationship can be established between the subsystem entity with high similarity and the preset entity. Based on this relationship, the preset knowledge graph and the subsystem knowledge graph can be linked to obtain the target knowledge graph. Using the preset knowledge graph of the target knowledge graph, the original data (subsystem entity data) in each subsystem of the power grid can be collected across systems. This solves the problem of low efficiency and inaccuracy caused by relying on manual processing of data scattered in different subsystems during the power grid evaluation data collection stage, and improves the automation level of the power grid evaluation process.

[0022] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart of a method for determining a knowledge graph according to Embodiment 1 of the present invention;

[0025] Figure 2 This is an architecture diagram of a target knowledge graph provided in Embodiment 1 of the present invention;

[0026] Figure 3 This is a flowchart of a method for determining a knowledge graph according to Embodiment 2 of the present invention;

[0027] Figure 4 This is a schematic diagram of the structure of a knowledge graph determination device according to Embodiment 3 of the present invention;

[0028] Figure 5 This is a schematic diagram of the structure of an electronic device provided according to Embodiment 4 of the present invention. Detailed Implementation

[0029] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0030] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. In the description of this invention, unless otherwise stated, "a plurality of" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist; for example, A and / or B can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0031] Example 1

[0032] Figure 1 The flowchart of a method for determining a knowledge graph is provided in Embodiment 1 of the present invention. This embodiment is applicable to the construction of knowledge graphs. The method can be executed by a knowledge graph determining device, which can be implemented in hardware and / or software. The knowledge graph determining device can be configured in an electronic device, which can be composed of two or more physical entities or a single physical entity.

[0033] like Figure 1 As shown, the method for determining a knowledge graph provided in Embodiment 1 of the present invention specifically includes the following steps:

[0034] S101. Determine the preset knowledge graph and construct the subsystem knowledge graph based on the data information in the power grid subsystem.

[0035] The preset knowledge graph contains multiple preset entities, each preset entity being associated with a preset entity description, preset attributes, and preset attribute descriptions. The subsystem knowledge graph contains multiple subsystem entities, each subsystem entity being associated with a subsystem entity description, subsystem attributes, and subsystem attribute descriptions.

[0036] In this embodiment, a knowledge graph covering business operations such as power grid evaluation can be pre-constructed, i.e., a preset knowledge graph. This preset knowledge graph can contain multiple preset entities, and each preset entity can contain multiple preset attributes. Each preset attribute and preset entity has its own description, i.e., a preset attribute description and a preset entity description. Here, a preset entity can be understood as an empty data table, where the columns of the empty data table serve as entity attributes, the description of the empty data table serves as the entity description, and the descriptions of the columns of the empty data table serve as entity attribute descriptions.

[0037] For example, if the preset knowledge graph is an evaluation knowledge graph for the power grid system, it can include preset entities such as green environmental protection, renewable energy access ratio, total power generation, hydropower generation ratio, hydropower generation, details of hydropower generation, photovoltaic power generation, details of photovoltaic power generation, wind power generation, details of wind power generation, carbon dioxide emission reduction, and details of carbon dioxide emission reduction.

[0038] Accordingly, data tables can be obtained from the database of each power grid subsystem, thereby identifying multiple subsystem entities, subsystem entity descriptions, subsystem attributes, and subsystem attribute descriptions. The obtained data tables are the subsystem entities, the table descriptions serve as subsystem entity descriptions, the table columns serve as subsystem attributes, and the column descriptions serve as subsystem attribute descriptions.

[0039] For example, if a photovoltaic (PV) power generation customer management system includes a PV power generation customer table, a customer power outage event table, and a PV customer power generation table, then three entities can be identified. These three entities are named PV power generation customer, customer power outage event, and PV customer power generation, respectively. If the PV power generation customer table includes columns such as customer account number, customer name, power generation customer type, installed capacity, and grid connection date, then the entity name is PV power generation customer, and the corresponding attribute names are customer number, customer name, power generation customer type, installed capacity, and grid connection date, etc. The subsystem entity description for PV power generation customer can be "records information such as PV power generation-related customer number, power generation, and grid connection time." The subsystem attribute description for customer account number can be "each customer's account number."

[0040] In this embodiment, since the primary key of a table uniquely determines the values ​​of other columns in the table, and the foreign key of a table may be the primary key of another table, the foreign key relationship between tables can effectively reflect the relationship between tables. Therefore, based on the primary and foreign key relationships between the data tables corresponding to entities, the association relationship between entities within each subsystem can be established. This association relationship is also the association relationship between subsystem entities in the subsystem knowledge graph. There can be multiple subsystems. The primary key of a table can be preset by the table creator. Once the primary key is determined, the remaining columns are the foreign keys. The association relationship between multiple preset entities in the preset knowledge graph can be preset according to the actual situation. For example, if the preset entities are annual power generation, monthly power generation, and daily power generation, then the association relationship between these three can be preset, that is, the foreign key in the annual power generation table is monthly power generation, and the foreign key in the monthly power generation table is daily power generation.

[0041] For example, if the customer account number in the customer power outage event table is a foreign key, and that foreign key is the primary key in the photovoltaic power generation customer table, then a relationship can be established between the actual customer power outage event and the actual photovoltaic power generation customer. Similarly, if the customer account number in the photovoltaic customer power generation table is a foreign key, and that customer account number is the primary key in the photovoltaic power generation customer table, then a relationship can also be established between the actual photovoltaic customer power generation and the actual photovoltaic power generation customer.

[0042] S102. Input the first splicing amount and the second splicing amount into the first preset model to obtain the first similarity between the subsystem entity and the preset entity.

[0043] Wherein, the first concatenation amount is the concatenation amount of the name of the subsystem entity and the embedding vector of the subsystem entity description, the second concatenation amount is the concatenation amount of the name of the preset entity and the embedding vector of the preset entity description, and the first preset model includes a first preset semantic extraction model, a preset graph neural network model, and a first classifier.

[0044] In this embodiment, the name of each subsystem entity and its corresponding description are converted into embedding vectors, and these embedding vectors are concatenated to obtain a first concatenation quantity equal to the number of subsystem entities. The name of each preset entity and its corresponding description are also converted into embedding vectors, and these embedding vectors are concatenated to obtain a second concatenation quantity equal to the number of preset entities. The first and second concatenation quantities are input into a first preset model to obtain a first similarity. A preset semantic extraction model in the first preset model can be used to extract semantic features from the first and second concatenation quantities, obtaining semantic feature vectors. These semantic feature vectors are then input into a preset graph neural network model to obtain structural feature vectors for the first and second concatenation quantities. After inputting these structural feature vectors into a first classifier, the similarity (first similarity) between the subsystem entity and the preset entity is obtained. The embedding vectors include word vectors, position vectors, and segment vectors. The preset semantic extraction model can be BERT (Bidirectional Enoceder Representations from Transformers) or similar models. The preset graph neural network model can be a graph neural network, a graph convolutional neural network, or an attention-based graph neural network. The classifier can be a softmax classifier or similar models.

[0045] S103. Input the third splicing amount into the second preset model to obtain the second similarity between the subsystem attributes and the preset attributes, and determine the third similarity between the subsystem entity and the preset entity based on the second similarity.

[0046] The third splicing quantity is the splicing quantity of the subsystem attribute, the subsystem attribute description, the preset attribute, and the preset attribute description. The second preset model includes a second preset semantic extraction model and a second classifier.

[0047] In this embodiment, the subsystem attributes, subsystem attribute descriptions, preset attributes, and preset attribute descriptions can be concatenated to obtain a third concatenated quantity. Inputting the third concatenated quantity into a second preset model yields a second similarity. The second preset semantic extraction model within the second preset model can be used to extract semantic features from the third concatenated quantity. After processing by a second classifier, these semantic features yield the similarity between the subsystem attributes and the preset attributes (second similarity). Performing preset processing on the obtained second similarity yields the similarity between the subsystem entities and the preset entities (third similarity). For example, if a subsystem entity contains subsystem attributes a and b, and the preset entity contains preset attributes A and B, then the average or maximum similarity of a and A, a and B, b and A, and b and B can be used as the third similarity. The subsystem attributes and preset attributes in the third concatenated quantity can be concatenated using attribute names.

[0048] S104. Based on the first similarity and the third similarity, determine the association relationship between the subsystem entity and the preset entity, and based on the association relationship, determine the target knowledge graph composed of the preset knowledge graph and the subsystem knowledge graph.

[0049] In this embodiment, the first and third similarities can be processed using a preset method to obtain the target similarity between the subsystem entity and the preset entity. If set p contains entity pairs of the subsystem entity and the preset entity corresponding to the first similarity, and set q contains entity pairs of the subsystem entity and the preset entity corresponding to the third similarity, then sets a and b can be integrated. The larger similarity among the overlapping subsystem entity and preset entity pairs in sets a and b is determined as the target similarity. If the target similarity is greater than a preset similarity threshold, such as 0.6, then a relationship between the subsystem entity and the preset entity can be determined. Figure 2 As an architectural diagram of a target knowledge graph, such as Figure 2 As shown, based on this relationship, the preset knowledge graph and the subsystem knowledge graph can be linked together, thereby obtaining a target knowledge graph that includes the preset knowledge graph and the subsystem knowledge graph. Figure 2 The dashed lines in the diagram represent the relationships between the subsystem knowledge graph and the preset knowledge graph, as well as the relationships between the subsystem knowledge graph and the subsystem database. Figure 2 The solid lines in the graph represent the relationships between entities within the knowledge graph. Figure 2 The knowledge graph for each subsystem contains three subsystem knowledge graphs, with each circle representing the knowledge graph of a subsystem. Power grid personnel can locate and retrieve raw data from the database of a power grid subsystem by using the pre-defined entities in the pre-defined knowledge graphs within the target knowledge graph and leveraging the associated entities in the subsystem knowledge graphs.

[0050] The knowledge graph determination method provided in this invention involves determining a preset knowledge graph and constructing a subsystem knowledge graph based on data information from a power grid subsystem. The preset knowledge graph contains multiple preset entities, each associated with a preset entity description, preset attributes, and preset attribute descriptions. The subsystem knowledge graph contains multiple subsystem entities, each associated with a subsystem entity description, subsystem attributes, and subsystem attribute descriptions. A first concatenation value and a second concatenation value are input into a first preset model to obtain a first similarity between the subsystem entity and the preset entity. The first concatenation value is the concatenation value between the name of the subsystem entity and the embedding vector of the subsystem entity description, and the second concatenation value is the concatenation value between the name of the preset entity and the embedding vector of the preset entity description. The first preset model includes a first preset semantic extraction model, a preset graph neural network model, and a first classifier. The third concatenation quantity is input into the second preset model to obtain the second similarity between the subsystem attribute and the preset attribute. Based on the second similarity, the third similarity between the subsystem entity and the preset entity is determined. The third concatenation quantity is the concatenation quantity of the subsystem attribute, the subsystem attribute description, the preset attribute, and the preset attribute description. The second preset model includes a second preset semantic extraction model and a second classifier. Based on the first similarity and the third similarity, the association relationship between the subsystem entity and the preset entity is determined. Based on the association relationship, a target knowledge graph composed of the preset knowledge graph and the subsystem knowledge graph is determined. The technical solution of this invention first uses a first preset model to obtain a first similarity between a subsystem entity and a preset entity. Then, a second preset model is used to obtain a second similarity between the subsystem entity and the preset entity. Based on the second similarity, a third similarity between a subsystem attribute and a preset attribute is obtained. Finally, based on the first and third similarities, an association relationship can be established between subsystem entities with high similarity and preset entities. Based on this association relationship, a preset knowledge graph and a subsystem knowledge graph can be linked to obtain a target knowledge graph. Using the preset knowledge graph of this target knowledge graph, it is possible to collect raw data (subsystem entity data) from various subsystems of the power grid across systems. This solves the problem of low efficiency and inaccuracy caused by relying on manual processing of data scattered across different subsystems during the power grid evaluation data collection stage, and improves the automation level of the power grid evaluation process.

[0051] Example 2

[0052] Figure 3 This is a flowchart of a method for determining a knowledge graph according to Embodiment 2 of the present invention. The technical solution of the present invention is further optimized based on the above optional technical solutions, and a specific method for determining a knowledge graph is given.

[0053] Optionally, determining the third similarity between the subsystem entity and the preset entity based on the second similarity includes: determining candidate subsystem attributes and candidate preset attributes corresponding to the second similarity greater than a first preset similarity threshold; determining candidate subsystem entities corresponding to the candidate attributes from multiple subsystem entities, and determining candidate preset entities corresponding to the candidate preset attributes from multiple preset entities; determining a first ratio of the number of attributes of the candidate subsystem attributes to the total number of attributes of the candidate subsystem entities; determining a second ratio of the number of preset attributes of the candidate preset attributes to the total number of attributes of the candidate preset entities; and determining the third similarity between the subsystem entity and the preset entity based on the first ratio and the second ratio. The advantage of this configuration is that the first and second ratios can characterize the similarity information between the subsystem entity and the preset entity, and the third similarity can be accurately determined using the first and second ratios.

[0054] Optionally, determining the association between the subsystem entity and the preset entity based on the first similarity and the third similarity includes: determining, from multiple subsystem entities and multiple preset entities, a first target entity and a first target preset entity corresponding to the first similarity, and a second target entity and a second target preset entity corresponding to the third similarity; determining a first set based on the first target entity, the first target preset entity, and the first association between the first target entity and the first target preset entity; determining a second set based on the second target entity, the second target preset entity, and the second association between the second target entity and the second target preset entity; performing a union operation on the first set and the second set to obtain a target set; and determining the association between the subsystem entity and the preset entity based on the target associations contained in the target set. The advantage of this configuration is that by calculating the union of the first set and the second set, the association between the subsystem entity and the preset entity can be comprehensively determined.

[0055] like Figure 3 As shown in Embodiment 2 of the present invention, a method for determining a knowledge graph specifically includes the following steps:

[0056] S201. Determine the preset knowledge graph and construct the subsystem knowledge graph based on the data information in the power grid subsystem.

[0057] S202. Input the first splicing amount and the second splicing amount into the first preset model to obtain the first similarity between the subsystem entity and the preset entity.

[0058] Optionally, the method for determining the first preset model includes:

[0059] 1) Determine multiple sample vector pairs based on the first sample knowledge graph and the second sample knowledge graph, and input the multiple sample vector pairs into the first initial preset semantic extraction model to obtain multiple sample semantic embedding vector pairs composed of the first sample semantic embedding vector and the second sample semantic embedding vector. The sample vector pairs include the first sample concatenation amount and the second sample concatenation amount. The first sample concatenation amount is the concatenation amount of the name of the first sample entity and the embedding vector of the description of the first sample entity. The second sample concatenation amount is the concatenation amount of the name of the second sample entity and the embedding vector of the description of the second sample entity. The first sample entity belongs to the first sample knowledge graph, and the second sample entity belongs to the second sample knowledge graph. The first sample knowledge graph is a preset knowledge graph based on the second sample knowledge graph.

[0060] For example, if a second sample entity in the second sample knowledge graph is c, and its associated second sample entity is described as c', and the first sample entity in the preset first sample knowledge graph is C, and its associated first sample entity is described as C', then the first sample concatenation amount is the concatenation amount of the embedding vectors of c and c', and the second sample concatenation amount is the concatenation amount of the embedding vectors of C and C'. Inputting the sample vector pair composed of the first sample concatenation amount and the second sample concatenation amount into the first initial preset semantic extraction model yields a sample semantic embedding vector pair. This sample semantic embedding vector pair includes the first sample semantic embedding vector corresponding to the first sample concatenation amount and the second sample semantic embedding vector corresponding to the second sample concatenation amount.

[0061] 2) Input the first sample semantic embedding vector and the second sample semantic embedding vector into two different preset initial graph neural network models respectively to obtain sample structure feature vector pairs, and input the sample structure feature vector pairs into the first initial classifier to obtain the first sample similarity between the first sample entity and the second sample entity. Each sample structure feature vector pair contains the output results of two different preset initial graph neural network models.

[0062] For example, if the initial graph neural network model is preset to be an attention mechanism graph neural network, then two different graph neural networks can be used to convert the adjacency relationships between all first sample entities and the adjacency relationships between all second sample entities into structural feature vectors containing structural information.

[0063] The specific process by which attention-based graph neural networks process sample semantic embedding vector pairs can include:

[0064] ① If the initial features of the first sample entity i are h i The initial features of the first sample entity i and its neighboring entity j are h.j Then the attention weight e of entity j for the first sample entity i ij for

[0065] e ij =LeakyReLU(θ*(W*h) i W*h j ))

[0066] Where W and θ are trainable parameters, LeakyReLU() is the weight function, i, j = 1, 2, ..., M, and M is the total number of sample vector pairs.

[0067] ②If N i Let w be the set of neighboring nodes of the first sample entity i. To differentiate the attention weights between different sample entity pairs, a softmax function can be introduced for normalization. Then the weights w between sample entity pairs are... ij for

[0068]

[0069] Where exp() is an exponential function with the natural constant e as the base, ∑ represents the summation operation, n represents the number of nodes in the neighbor set, and h is the structural feature vector of node i in the current layer. i ′ for

[0070]

[0071] Where δ is a non-linear activation function, and the initial graph neural network model can be a multi-layer attention mechanism graph neural network.

[0072] For each sample vector pair, the sample structure feature vectors output by two different attention mechanism graph neural networks are input into the first initial classifier to obtain the first sample similarity of the sample vector pair (first sample entity and second sample entity).

[0073] 3) Receive verification information for the first set sample pair, determine the first aligned sample pair from the first set sample pair according to the verification information, and train the first initial preset semantic extraction model, the preset initial graph neural network model and the first initial classifier according to the first alignment similarity of the first aligned sample pair to obtain the first preset model. The first set sample pair is a first sample entity and a second sample entity whose first sample similarity is greater than a first preset threshold. The first sample entity and the second sample entity in the first aligned sample pair meet the first preset matching requirements.

[0074] Specifically, manual verification can be used to check whether the first sample entity and the second sample entity meet the first preset matching requirements, such as checking whether the first sample entity and the second sample entity have the same meaning. If they do, the verification result is a match; if they do not, the verification result is a mismatch. The manual verification result (verification information) is received, allowing the selection of sample entity pairs (first aligned sample pairs) whose verification information matches from the first set of sample pairs. If the similarity of the first aligned sample pair is less than a first threshold, such as 0.8, the first initial preset semantic extraction model, the preset initial graph neural network model, and the first initial classifier can be trained until the similarity of the first aligned sample pair is greater than the first threshold. The first set sample pair is a sample entity pair with a similarity greater than a first preset threshold, such as 0.5. It can update the sample vector pair input each time. That is, the sample vector pair input each time can be composed of the sample vector pairs input in the previous training rounds and the new sample vector pair. Correspondingly, the first aligned sample pair determined each time will also be updated. In the sample vector pair of the current round, the first aligned sample pair determined in the previous round does not need to participate in the manual verification of the current round. The verification information of the previous round can be used directly in the current round.

[0075] The advantage of this setup is that by using multiple sample vector pairs to determine the similarity of sample entity pairs (first sample similarity), and then using the verification information to filter the similarity of sample pairs, the initial model of the first preset model can be trained using the magnitude of the filtered similarity (first alignment similarity), thus training a first preset model with more accurate output results.

[0076] S203. Input the third splicing amount into the second preset model to obtain the second similarity between the subsystem attributes and the preset attributes.

[0077] Optionally, the method for determining the second preset model includes:

[0078] 1) Input multiple sample vectors into the second initial preset semantic extraction model to obtain sample semantic embedding vectors, wherein the sample vectors are the concatenation of the first sample attribute, the first sample attribute description, the second sample attribute and the second sample attribute description, and the first sample attribute and its associated first sample attribute description are preset attributes and attribute descriptions based on the second sample attribute and its associated second sample attribute description.

[0079] Specifically, if the second sample attribute is d and its corresponding second sample attribute description is d', then the first sample attribute preset based on d is D, and the first sample attribute preset based on d' is D', then the sample vector is the concatenation of d, d', D and D'. By inputting this sample vector into the second initial preset semantic extraction model, the corresponding sample semantic embedding vector can be obtained.

[0080] For example, the sample vector can be represented as (CLS||am||dm||SEP||an||dn||SEP), such as (CLS||PV Customer Power Generation.Power Generation||PV Customer Total Power Generation||SEP||PV Power Generation Details.Annual Power Generation||PV Customer Annual Total Power Generation||SEP), where PV Customer Power Generation and PV Power Generation Details are the names of the entities used to represent the association between sample attributes and sample entities. Here, || indicates vector concatenation, CLS indicates the start of a sequence, SEP indicates the end of a sequence, am is the first sample attribute, dm is the description of the first sample attribute, an is the second sample attribute, and dn is the description of the second sample attribute.

[0081] 2) Input the sample semantic embedding vector into the second initial classifier to obtain the second sample similarity between the first sample attribute and the second sample attribute.

[0082] Specifically, by inputting the sample semantic embedding vector into the second initial classifier, the second sample similarity between the first sample attribute and the second sample attribute can be obtained. As in the example above, the similarity between power generation and annual power generation can be obtained, i.e., the second sample similarity.

[0083] 3) Receive verification information for the second set sample pair, determine the second aligned sample pair from the second set sample pair according to the verification information, train the second initial preset semantic extraction model and the second initial classifier according to the second alignment similarity of the second aligned sample pair to obtain the second preset model, wherein the second set sample pair is the first sample attribute and the second sample attribute corresponding to the second sample similarity greater than the second preset threshold, and the first sample attribute and the second sample attribute in the second aligned sample pair satisfy the second preset matching requirement.

[0084] Specifically, manual verification can be used to check whether the first and second sample attributes meet the second preset matching requirements. For example, the meaning of the first and second sample attributes can be checked. If they are consistent, the verification result is a match; if they are inconsistent, the verification result is a mismatch. The manual verification results (verification information) are received, allowing the selection of sample attribute pairs (first and second sample attributes) with matching verification information from the second set of sample pairs. This sample attribute pair is the second aligned sample pair. If the similarity of this second aligned sample pair is less than a second threshold, such as 0.8, multiple rounds of training can be performed on the second initial preset semantic extraction model and the second initial classifier until the similarity of the second aligned sample pair is greater than the second threshold. The second set sample pair is a sample attribute pair with a similarity greater than a second preset threshold, such as 0.5. It can update the sample vector for each input. That is, the sample vector for each input can be composed of the sample vector input in the previous training round and the new sample vector. Correspondingly, the second aligned sample pair determined each time will also be updated. In the sample vector pair of the current round, the second aligned sample pair determined in the previous round does not need to participate in the manual verification of the current round. The verification information of the previous round can be used directly in the current round. The first threshold and the second threshold can be the same or different.

[0085] The advantage of this setup is that by using multiple sample vectors to determine the similarity of the second sample, and by using the verification information to filter the similarity of the sample pairs, and then using the magnitude of the filtered similarity to train the initial model of the second preset model, a second preset model with more accurate output results can be trained.

[0086] S204. Determine the candidate subsystem attribute and candidate preset attribute corresponding to the second similarity greater than the first preset similarity threshold, and determine the candidate subsystem entity corresponding to the candidate attribute from multiple subsystem entities, and determine the candidate preset entity corresponding to the candidate preset attribute from multiple preset entities.

[0087] For example, if the first preset similarity threshold is 0.8, then similarities greater than 0.8 can be filtered from multiple second similarities. The subsystem attribute corresponding to this similarity is a candidate subsystem attribute, and the preset attribute corresponding to this similarity is a candidate preset attribute. Candidate subsystem entities corresponding to the candidate subsystem attributes can be determined from multiple subsystem entities, and candidate preset entities corresponding to the candidate preset attributes can be determined from multiple preset entities.

[0088] S205. Determine the first proportion of the number of attributes of the candidate subsystem attributes to the total number of attributes of the candidate subsystem entities.

[0089] For example, if the number of attributes of a candidate subsystem is 15 and the total number of attributes of a candidate subsystem entity is 20, then the first ratio is 75%.

[0090] S206. Determine the second ratio of the number of preset attributes of the candidate preset attributes to the total number of attributes of the candidate preset entities.

[0091] For example, if the number of preset attributes of the candidate preset attributes is 17 and the total number of attributes of the candidate preset entities is 20, then the second ratio is 85%.

[0092] S207. Determine the third similarity between the subsystem entity and the preset entity based on the first ratio and the second ratio.

[0093] Specifically, the first and second ratios can be processed to obtain the third similarity, such as using the mean of the first and second ratios as the third similarity, or using the weighted sum of the first and second ratios as the third similarity, etc.

[0094] Optionally, determining the third similarity between the subsystem entity and the preset entity based on the first ratio and the second ratio includes: determining the maximum value of the first ratio and the second ratio as a candidate ratio; if the candidate ratio is greater than a second preset similarity threshold, then determining the candidate ratio as the third similarity between the subsystem entity and the preset entity. The advantage of this setting is that by using the second preset similarity threshold, larger values ​​among the candidate ratios are filtered out, that is, preset entities and subsystem entities with higher correlation are filtered out, ensuring the accuracy of the target correlation.

[0095] For example, if the first ratio is 75% and the second ratio is 85%, then the candidate ratio is 85%. If the second preset similarity threshold is 0.6 and the candidate ratio is 85%, which is greater than the second preset similarity threshold, then the third similarity is 85%. Wherein, if the candidate ratio is less than or equal to the second preset similarity threshold, it is considered that the preset entity and the subsystem entity corresponding to the candidate ratio are not related, and the candidate ratio can be discarded.

[0096] S208. From the plurality of subsystem entities and the plurality of preset entities, determine the first target entity and the first target preset entity corresponding to the first similarity, and the second target entity and the second target preset entity corresponding to the third similarity.

[0097] Optionally, before determining the first target entity and the first target preset entity corresponding to the first similarity, the method further includes: deleting the first similarity values ​​that are less than a third preset similarity threshold to obtain the deleted first similarity values. The advantage of this setting is that it can delete the smaller similarities in the first similarity values, retaining the higher similarities, thus ensuring the accuracy of the target association.

[0098] S209. Determine a first set based on the first target entity, the first target preset entity, and the first association relationship between the first target entity and the first target preset entity.

[0099] For example, if the first target entities are e and f, the first target preset entities are E and F, e and E have a first association relationship, and f and F have a first association relationship, then the first set can be represented as {(e, E), (f, F)}.

[0100] S210. Determine the second set based on the second target entity, the second target preset entity, and the second association relationship between the second target entity and the second target preset entity.

[0101] For example, if the second target entities are g and h, and the second target preset entities are G and H, and g and G have a second association relationship, and h and H have a second association relationship, then the second set can be represented as {(g, G), (h, H)}.

[0102] S211. Perform a union operation on the first set and the second set to obtain the target set.

[0103] For example, as described in the example above, the target set can be represented as {(e, E), (f, F), (g, G), (h, H)}.

[0104] S212. Determine the association relationship between the subsystem entity and the preset entity based on the target association relationships contained in the target set.

[0105] For example, if the target set can be represented as {(e, E), (f, F), (g, G), (h, H)}, then it can be determined that there are associations between e and E, f and F, g and G, and h and H.

[0106] S213. Determine the target knowledge graph composed of the preset knowledge graph and the subsystem knowledge graph based on the association relationship.

[0107] The knowledge graph determination method provided in this invention first uses a first preset model to obtain a first similarity between subsystem entities and preset entities. Then, a second preset model is used to obtain a second similarity between subsystem entities and preset entities. A third similarity is determined based on the second similarity and a first preset similarity threshold. Finally, the preset entities and subsystem entities corresponding to the first and third similarities are integrated to determine the association relationships between all subsystem entities and preset entities. Based on these association relationships, the preset knowledge graph and the subsystem knowledge graph can be linked to obtain a target knowledge graph. Using the preset knowledge graph of this target knowledge graph, cross-system data collection (subsystem entity data) from various subsystems of the power grid can be achieved. The target knowledge graph can fully explore the asset value of data, improve the intelligent utilization level of data assets, and solve the problems of low efficiency and inaccurate accuracy caused by relying on manual processing of data scattered across different subsystems during the power grid evaluation data collection stage. This improves the automation level of the power grid evaluation process.

[0108] Based on the above embodiments, the method may further include:

[0109] 1) Determine the first calculation relationship of the evaluation indicators. Based on the first calculation relationship and the second calculation relationship of the target knowledge graph, determine the target preset entities from the target knowledge graph. The first calculation relationship includes calculation relationships between preset entity data, and the second calculation relationship is the calculation relationship between all preset entity data in the preset knowledge graph. The second calculation relationship may include calculation relationships between preset entities, calculation relationships between preset entities and preset attributes, and calculation relationships between preset attributes. The second calculation relationship can be displayed in the preset knowledge graph.

[0110] For example, if the first calculation relationship of the evaluation index N is: N = E + F + L, and the second calculation relationship includes the calculation relationship of L, L = (EF) * (G / H), then the target preset entities can be determined as E, F and L.

[0111] 2) Determine intermediate calculation relationships based on the target preset entity and the third calculation relationship of the target knowledge graph, wherein the third calculation relationship is the calculation relationship between all subsystem entity data in the subsystem knowledge graph, and the intermediate calculation relationship is the calculation relationship between subsystem entity data.

[0112] For example, as described above, there are relationships between e and E, f and F, g and G, and h and H. If the third computational relationship of the target knowledge graph includes the computational relationship of e, e = a + b, then the intermediate computational relationship can be obtained as N = a + b + f + (a + bf) * (g / h). The third computational relationship can also include computational relationships between subsystem entity data and subsystem attribute data, as well as computational relationships between subsystem attribute data. This third computational relationship can be displayed in the subsystem knowledge graph.

[0113] 3) Determine the index value of the evaluation index based on the intermediate calculation relationship and the subsystem entity data.

[0114] For example, if in the subsystem entity data a=1, b=2, f=3, g=4, h=5, then the index value of evaluation index L is 6.

[0115] The advantage of this setup is that it decouples the raw data from the power grid evaluation process, enhancing the robustness of the power grid evaluation to changes in the raw system data structure.

[0116] Example 3

[0117] Figure 4 This is a schematic diagram of a knowledge graph determination device provided in Embodiment 3 of the present invention. Figure 4 As shown, the device includes: a knowledge graph determination module 301, a first similarity determination module 302, a second similarity determination module 303, and an association determination module 304, wherein:

[0118] The knowledge graph determination module is used to determine a preset knowledge graph and construct a subsystem knowledge graph based on data information in the power grid subsystem. The preset knowledge graph contains multiple preset entities, and each preset entity is associated with a preset entity description, a preset attribute, and a preset attribute description. The subsystem knowledge graph contains multiple subsystem entities, and each subsystem entity is associated with a subsystem entity description, a subsystem attribute, and a subsystem attribute description.

[0119] The first similarity determination module is used to input the first concatenation amount and the second concatenation amount into the first preset model to obtain the first similarity between the subsystem entity and the preset entity. The first concatenation amount is the concatenation amount of the name of the subsystem entity and the embedding vector of the subsystem entity description, and the second concatenation amount is the concatenation amount of the name of the preset entity and the embedding vector of the preset entity description. The first preset model includes a first preset semantic extraction model, a preset graph neural network model, and a first classifier.

[0120] The second similarity determination module is used to input the third concatenation quantity into the second preset model to obtain the second similarity between the subsystem attribute and the preset attribute, and to determine the third similarity between the subsystem entity and the preset entity based on the second similarity. The third concatenation quantity is the concatenation quantity of the subsystem attribute, the subsystem attribute description, the preset attribute, and the preset attribute description. The second preset model includes a second preset semantic extraction model and a second classifier.

[0121] The association determination module is used to determine the association between the subsystem entity and the preset entity based on the first similarity and the third similarity, and to determine the target knowledge graph composed of the preset knowledge graph and the subsystem knowledge graph based on the association.

[0122] The knowledge graph determination device provided in this embodiment of the invention first uses a first preset model to obtain a first similarity between a subsystem entity and a preset entity, then uses a second preset model to obtain a second similarity between the subsystem entity and the preset entity, and based on the second similarity, obtains a third similarity between a subsystem attribute and a preset attribute. Finally, based on the first and third similarities, an association relationship can be established between subsystem entities with high similarity and preset entities. Based on this association relationship, the preset knowledge graph and the subsystem knowledge graph can be linked to obtain a target knowledge graph. Using the preset knowledge graph of the target knowledge graph, it is possible to collect raw data (subsystem entity data) from various subsystems of the power grid across systems. This solves the problems of low efficiency and inability to guarantee accuracy caused by relying on manual processing of data scattered in different subsystems during the power grid evaluation data collection stage, and improves the automation level of the power grid evaluation process.

[0123] Optionally, the method for determining the first preset model includes: determining multiple sample vector pairs based on the first sample knowledge graph and inputting the multiple sample vector pairs into the first initial preset semantic extraction model to obtain multiple sample semantic embedding vector pairs composed of the first sample semantic embedding vector and the second sample semantic embedding vector, wherein the sample vector pairs include the first sample concatenation amount and the second sample concatenation amount, the first sample concatenation amount is the concatenation amount of the name of the first sample entity and the embedding vector of the description of the first sample entity, the second sample concatenation amount is the concatenation amount of the name of the second sample entity and the embedding vector of the description of the second sample entity, the first sample entity belongs to the first sample knowledge graph, the second sample entity belongs to the second sample knowledge graph, and the first sample knowledge graph is a knowledge graph preset based on the second sample knowledge graph; inputting the first sample semantic embedding vector and the second sample semantic embedding vector into two... In different preset initial graph neural network models, sample structure feature vector pairs are obtained, and the sample structure feature vector pairs are input into a first initial classifier to obtain the first sample similarity between the first sample entity and the second sample entity. Each sample structure feature vector pair contains the output results of two different preset initial graph neural network models. Verification information for a first set sample pair is received. Based on the verification information, a first aligned sample pair is determined from the first set sample pair. Based on the magnitude of the first alignment similarity of the first aligned sample pair, the first initial preset semantic extraction model, the preset initial graph neural network model, and the first initial classifier are trained to obtain a first preset model. The first set sample pair consists of a first sample entity and a second sample entity whose first sample similarity is greater than a first preset threshold. The first sample entity and the second sample entity in the first aligned sample pair satisfy a first preset matching requirement.

[0124] Optionally, the determination method of the second preset model includes: inputting multiple sample vectors into a second initial preset semantic extraction model to obtain a sample semantic embedding vector, wherein the sample vector is a concatenation of a first sample attribute, a first sample attribute description, a second sample attribute, and a second sample attribute description, and the first sample attribute and its associated first sample attribute description are preset attributes and attribute descriptions based on the second sample attribute and its associated second sample attribute description; inputting the sample semantic embedding vector into a second initial classifier to obtain a second sample similarity between the first sample attribute and the second sample attribute; receiving verification information for a second set sample pair, determining a second aligned sample pair from the second set sample pair based on the verification information, and training the second initial preset semantic extraction model and the second initial classifier based on the magnitude of the second alignment similarity of the second aligned sample pair to obtain the second preset model, wherein the second set sample pair is the first sample attribute and the second sample attribute corresponding to a second sample similarity greater than a second preset threshold, and the first sample attribute and the second sample attribute in the second aligned sample pair satisfy the second preset matching requirements.

[0125] Optionally, the second similarity determination module includes:

[0126] The entity and attribute determination unit is used to determine the candidate subsystem attribute and the candidate preset attribute corresponding to the second similarity greater than the first preset similarity threshold, and to determine the candidate subsystem entity corresponding to the candidate attribute from multiple subsystem entities, and to determine the candidate preset entity corresponding to the candidate preset attribute from multiple preset entities.

[0127] The first proportion determining unit is used to determine a first proportion of the number of attributes of the candidate subsystem attributes to the total number of attributes of the candidate subsystem entities;

[0128] The second ratio determination unit is used to determine a second ratio of the number of preset attributes of the candidate preset attributes to the total number of attributes of the candidate preset entities.

[0129] A similarity determination unit is used to determine a third similarity between the subsystem entity and the preset entity based on the first ratio and the second ratio.

[0130] Optionally, determining the third similarity between the subsystem entity and the preset entity based on the first ratio and the second ratio includes: determining the maximum value of the first ratio and the second ratio as a candidate ratio; if the candidate ratio is greater than a second preset similarity threshold, then determining the candidate ratio as the third similarity between the subsystem entity and the preset entity.

[0131] Optionally, the association determination module includes:

[0132] An entity determination unit is used to determine, from multiple subsystem entities and multiple preset entities, a first target entity and a first target preset entity corresponding to the first similarity, and a second target entity and a second target preset entity corresponding to the third similarity;

[0133] The first set determination unit is used to determine a first set based on the first target entity, the first target preset entity, and the first association relationship between the first target entity and the first target preset entity;

[0134] The second set determination unit is used to determine a second set based on the second target entity, the second target preset entity, and the second association relationship between the second target entity and the second target preset entity;

[0135] The target set determination unit is used to perform a union operation on the first set and the second set to obtain the target set;

[0136] The association relationship determination unit is used to determine the association relationship between the subsystem entity and the preset entity based on the target association relationships contained in the target set.

[0137] Optionally, the device may also include:

[0138] The target preset entity determination module is used to determine a first calculation relationship of the evaluation index, and determine the target preset entity from the target knowledge graph based on the first calculation relationship and the second calculation relationship of the target knowledge graph. The first calculation relationship includes the calculation relationship between preset entity data, and the second calculation relationship is the calculation relationship between all preset entity data in the preset knowledge graph.

[0139] The intermediate calculation relationship determination module is used to determine intermediate calculation relationships based on the target preset entity and the target knowledge graph's third calculation relationship, wherein the third calculation relationship is the calculation relationship between all subsystem entity data in the subsystem knowledge graph, and the intermediate calculation relationship is the calculation relationship between subsystem entity data;

[0140] The indicator value determination module is used to determine the indicator value of the evaluation indicator based on the intermediate calculation relationship and the subsystem entity data.

[0141] The knowledge graph determination device provided in the embodiments of the present invention can execute the knowledge graph determination method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0142] Example 4

[0143] Figure 5A schematic diagram of an electronic device 40 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0144] like Figure 5 As shown, the electronic device 40 includes at least one processor 41 and a memory, such as a read-only memory (ROM) 42 or a random access memory (RAM) 43, communicatively connected to the at least one processor 41. The memory stores computer programs executable by the at least one processor. The processor 41 can perform various appropriate actions and processes based on the computer program stored in the ROM 42 or loaded into the RAM 43 from storage unit 48. The RAM 43 may also store various programs and data required for the operation of the electronic device 40. The processor 41, ROM 42, and RAM 43 are interconnected via a bus 44. An input / output (I / O) interface 45 is also connected to the bus 44.

[0145] Multiple components in electronic device 40 are connected to I / O interface 45, including: input unit 46, such as keyboard, mouse, etc.; output unit 47, such as various types of monitors, speakers, etc.; storage unit 48, such as disk, optical disk, etc.; and communication unit 49, such as network card, modem, wireless transceiver, etc. Communication unit 49 allows electronic device 40 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0146] Processor 41 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 41 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 41 performs the various methods and processes described above, such as methods for determining knowledge graphs.

[0147] In some embodiments, the knowledge graph determination method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 48. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 40 via ROM 42 and / or communication unit 49. When the computer program is loaded into RAM 43 and executed by processor 41, one or more steps of the knowledge graph determination method described above may be performed. Alternatively, in other embodiments, processor 41 may be configured to perform the knowledge graph determination method by any other suitable means (e.g., by means of firmware).

[0148] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0149] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0150] The computer equipment provided above can be used to execute the knowledge graph determination method provided in any of the above embodiments, and has corresponding functions and beneficial effects.

[0151] Example 5

[0152] In the context of this invention, the computer-readable storage medium may be a tangible medium, and the computer-executable instructions, when executed by a computer processor, are used to perform a method for determining a knowledge graph, the method comprising:

[0153] A preset knowledge graph is determined, and a subsystem knowledge graph is constructed based on the data information in the power grid subsystem. The preset knowledge graph contains multiple preset entities, and each preset entity is associated with a preset entity description, a preset attribute, and a preset attribute description. The subsystem knowledge graph contains multiple subsystem entities, and each subsystem entity is associated with a subsystem entity description, a subsystem attribute, and a subsystem attribute description.

[0154] The first concatenation amount and the second concatenation amount are input into the first preset model to obtain the first similarity between the subsystem entity and the preset entity. The first concatenation amount is the concatenation amount of the name of the subsystem entity and the embedding vector of the subsystem entity description. The second concatenation amount is the concatenation amount of the name of the preset entity and the embedding vector of the preset entity description. The first preset model includes a first preset semantic extraction model, a preset graph neural network model and a first classifier.

[0155] The third concatenation quantity is input into the second preset model to obtain the second similarity between the subsystem attribute and the preset attribute, and the third similarity between the subsystem entity and the preset entity is determined based on the second similarity. The third concatenation quantity is the concatenation quantity of the subsystem attribute, the subsystem attribute description, the preset attribute, and the preset attribute description. The second preset model includes a second preset semantic extraction model and a second classifier.

[0156] Based on the first similarity and the third similarity, the association between the subsystem entity and the preset entity is determined, and based on the association, a target knowledge graph composed of the preset knowledge graph and the subsystem knowledge graph is determined.

[0157] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by, or in conjunction with, an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0158] The computer equipment provided above can be used to execute the knowledge graph determination method provided in any of the above embodiments, and has corresponding functions and beneficial effects.

[0159] It is worth noting that in the embodiments of the knowledge graph determination device described above, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0160] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A method for determining a knowledge graph, characterized in that, include: A preset knowledge graph is determined, and a subsystem knowledge graph is constructed based on the data information in the power grid subsystem. The preset knowledge graph contains multiple preset entities, and each preset entity is associated with a preset entity description, a preset attribute, and a preset attribute description. The subsystem knowledge graph contains multiple subsystem entities, and each subsystem entity is associated with a subsystem entity description, a subsystem attribute, and a subsystem attribute description. The first concatenation amount and the second concatenation amount are input into the first preset model to obtain the first similarity between the subsystem entity and the preset entity. The first concatenation amount is the concatenation amount of the name of the subsystem entity and the embedding vector of the subsystem entity description. The second concatenation amount is the concatenation amount of the name of the preset entity and the embedding vector of the preset entity description. The first preset model includes a first preset semantic extraction model, a preset graph neural network model and a first classifier. The third concatenation quantity is input into the second preset model to obtain the second similarity between the subsystem attribute and the preset attribute, and the third similarity between the subsystem entity and the preset entity is determined based on the second similarity. The third concatenation quantity is the concatenation quantity of the subsystem attribute, the subsystem attribute description, the preset attribute, and the preset attribute description. The second preset model includes a second preset semantic extraction model and a second classifier. Based on the first similarity and the third similarity, the association between the subsystem entity and the preset entity is determined, and based on the association, a target knowledge graph composed of the preset knowledge graph and the subsystem knowledge graph is determined; The step of determining the third similarity between the subsystem entity and the preset entity based on the second similarity includes: Determine the candidate subsystem attribute and candidate preset attribute corresponding to the second similarity greater than the first preset similarity threshold, and determine the candidate subsystem entity corresponding to the candidate subsystem attribute from multiple subsystem entities, and determine the candidate preset entity corresponding to the candidate preset attribute from multiple preset entities; Determine a first proportion of the number of attributes of the candidate subsystem to the total number of attributes of the candidate subsystem entities; Determine a second ratio of the number of preset attributes of the candidate preset attributes to the total number of attributes of the candidate preset entities; Based on the first ratio and the second ratio, a third similarity between the subsystem entity and the preset entity is determined; The step of determining the association between the subsystem entity and the preset entity based on the first similarity and the third similarity includes: From the plurality of subsystem entities and the plurality of preset entities, determine the first target entity and the first target preset entity corresponding to the first similarity, and the second target entity and the second target preset entity corresponding to the third similarity; A first set is determined based on the first target entity, the first target preset entity, and the first association relationship between the first target entity and the first target preset entity; A second set is determined based on the second target entity, the second target preset entity, and the second association relationship between the second target entity and the second target preset entity; Perform a union operation on the first set and the second set to obtain the target set; Based on the target associations contained in the target set, the association between the subsystem entity and the preset entity is determined.

2. The method according to claim 1, characterized in that, The methods for determining the first preset model include: Multiple sample vector pairs are determined based on the first sample knowledge graph and the second sample knowledge graph, and these multiple sample vector pairs are input into the first initial preset semantic extraction model to obtain multiple sample semantic embedding vector pairs composed of the first sample semantic embedding vector and the second sample semantic embedding vector. Each sample vector pair includes a first sample concatenation amount and a second sample concatenation amount. The first sample concatenation amount is the concatenation amount of the name of the first sample entity and the embedding vector of the first sample entity description. The second sample concatenation amount is the concatenation amount of the name of the second sample entity and the embedding vector of the second sample entity description. The first sample entity belongs to the first sample knowledge graph, and the second sample entity belongs to the second sample knowledge graph. The first sample knowledge graph is a preset knowledge graph based on the second sample knowledge graph. The first sample semantic embedding vector and the second sample semantic embedding vector are respectively input into two different preset initial graph neural network models to obtain sample structure feature vector pairs. The sample structure feature vector pairs are then input into a first initial classifier to obtain the first sample similarity between the first sample entity and the second sample entity. Each sample structure feature vector pair contains the output results of two different preset initial graph neural network models. The system receives verification information for a first set sample pair, determines a first aligned sample pair from the first set sample pair based on the verification information, and trains a first initial preset semantic extraction model, a preset initial graph neural network model, and a first initial classifier based on the magnitude of the first alignment similarity of the first aligned sample pair to obtain a first preset model. The first set sample pair consists of a first sample entity and a second sample entity whose first sample similarity is greater than a first preset threshold, and the first sample entity and the second sample entity in the first aligned sample pair satisfy a first preset matching requirement.

3. The method according to claim 1, characterized in that, The methods for determining the second preset model include: Multiple sample vectors are input into the second initial preset semantic extraction model to obtain sample semantic embedding vectors. The sample vectors are concatenations of the first sample attribute, the first sample attribute description, the second sample attribute, and the second sample attribute description. The first sample attribute and its associated first sample attribute description are preset attributes and attribute descriptions based on the second sample attribute and its associated second sample attribute description. The sample semantic embedding vector is input into the second initial classifier to obtain the second sample similarity between the first sample attribute and the second sample attribute; The system receives verification information for a second set sample pair, determines a second aligned sample pair from the second set sample pair based on the verification information, and trains a second initial preset semantic extraction model and a second initial classifier based on the second alignment similarity of the second aligned sample pair to obtain a second preset model. The second set sample pair consists of a first sample attribute and a second sample attribute corresponding to a second sample similarity greater than a second preset threshold, and the first sample attribute and the second sample attribute in the second aligned sample pair satisfy a second preset matching requirement.

4. The method according to claim 1, characterized in that, The step of determining the third similarity between the subsystem entity and the preset entity based on the first ratio and the second ratio includes: The maximum value between the first ratio and the second ratio is determined as the candidate ratio; If the candidate ratio is greater than the second preset similarity threshold, then the candidate ratio is determined as the third similarity between the subsystem entity and the preset entity.

5. The method according to claim 1, characterized in that, Also includes: A first calculation relationship for evaluation indicators is determined. Based on the first calculation relationship and the second calculation relationship of the target knowledge graph, a target preset entity is determined from the target knowledge graph. The first calculation relationship includes the calculation relationship between preset entity data, and the second calculation relationship is the calculation relationship between all preset entity data in the preset knowledge graph. Based on the target preset entity and the third calculation relationship of the target knowledge graph, an intermediate calculation relationship is determined, wherein the third calculation relationship is the calculation relationship between all subsystem entity data in the subsystem knowledge graph, and the intermediate calculation relationship is the calculation relationship between subsystem entity data; The evaluation index value is determined based on the intermediate calculation relationship and the subsystem entity data.

6. A device for determining a knowledge graph, characterized in that, include: The knowledge graph determination module is used to determine a preset knowledge graph and construct a subsystem knowledge graph based on data information in the power grid subsystem. The preset knowledge graph contains multiple preset entities, and each preset entity is associated with a preset entity description, a preset attribute, and a preset attribute description. The subsystem knowledge graph contains multiple subsystem entities, and each subsystem entity is associated with a subsystem entity description, a subsystem attribute, and a subsystem attribute description. The first similarity determination module is used to input the first concatenation amount and the second concatenation amount into the first preset model to obtain the first similarity between the subsystem entity and the preset entity. The first concatenation amount is the concatenation amount of the name of the subsystem entity and the embedding vector of the subsystem entity description, and the second concatenation amount is the concatenation amount of the name of the preset entity and the embedding vector of the preset entity description. The first preset model includes a first preset semantic extraction model, a preset graph neural network model, and a first classifier. The second similarity determination module is used to input the third concatenation quantity into the second preset model to obtain the second similarity between the subsystem attribute and the preset attribute, and to determine the third similarity between the subsystem entity and the preset entity based on the second similarity. The third concatenation quantity is the concatenation quantity of the subsystem attribute, the subsystem attribute description, the preset attribute, and the preset attribute description. The second preset model includes a second preset semantic extraction model and a second classifier. The association relationship determination module is used to determine the association relationship between the subsystem entity and the preset entity based on the first similarity and the third similarity, and to determine the target knowledge graph composed of the preset knowledge graph and the subsystem knowledge graph based on the association relationship; The second similarity determination module includes: The entity and attribute determination unit is used to determine the candidate subsystem attribute and the candidate preset attribute corresponding to the second similarity greater than the first preset similarity threshold, and to determine the candidate subsystem entity corresponding to the candidate subsystem attribute from multiple subsystem entities, and to determine the candidate preset entity corresponding to the candidate preset attribute from multiple preset entities. The first proportion determining unit is used to determine a first proportion of the number of attributes of the candidate subsystem attributes to the total number of attributes of the candidate subsystem entities; The second ratio determination unit is used to determine a second ratio of the number of preset attributes of the candidate preset attributes to the total number of attributes of the candidate preset entities. A similarity determination unit is used to determine a third similarity between the subsystem entity and the preset entity based on the first ratio and the second ratio; The association determination module includes: An entity determination unit is used to determine, from multiple subsystem entities and multiple preset entities, a first target entity and a first target preset entity corresponding to the first similarity, and a second target entity and a second target preset entity corresponding to the third similarity; The first set determination unit is used to determine a first set based on the first target entity, the first target preset entity, and the first association relationship between the first target entity and the first target preset entity; The second set determination unit is used to determine a second set based on the second target entity, the second target preset entity, and the second association relationship between the second target entity and the second target preset entity; The target set determination unit is used to perform a union operation on the first set and the second set to obtain the target set; The association relationship determination unit is used to determine the association relationship between the subsystem entity and the preset entity based on the target association relationships contained in the target set.

7. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the method for determining the knowledge graph according to any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the method for determining the knowledge graph as described in any one of claims 1-5.

Images