Construction method of standard digital reasoning model in power field based on knowledge graph

Abstract

The invention discloses a construction method of a power field standard digital reasoning model based on a knowledge graph, and relates to the technical field of electric digital data processing. The construction method of the power field standard digital reasoning model based on the knowledge graph comprises the following steps: collecting data; constructing a knowledge graph; performing model construction verification; and optimizing the model. The electric power standard knowledge graph is constructed through the electric power standard data and the electric power knowledge data, then the digital inference model is constructed based on the preset inference rule in the electric power field and in combination with the electric power performance parameters of the electric power standard knowledge graph, and the model verification result is obtained through model verification. And then the power field in the preset time interval is monitored in real time, and the power change degree is obtained and optimized, so that the effect of improving the data utilization rate of the knowledge graph in the model construction process is achieved, and the problem of insufficient data utilization of the knowledge graph in the model construction process in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical digital data processing, and in particular to a method for constructing a standard digital reasoning model in the electric power field based on a knowledge graph. Background Art

[0002] With the advancement of national standardization, standard documents are becoming more and more numerous and richer, and these documents are stored in the database in the form of pictures, texts, etc. As an important branch of artificial intelligence technology, knowledge graph (KG) has been successfully applied in the big data environment, providing people with a more effective cross-media big data organization, management and cognitive ability.

[0003] At the same time, combined with the application of big data in intelligent search and intelligent questioning, by expressing data information in a specific field into a knowledge representation form that is closer to the human cognitive world, domain-specific knowledge graph (DKG) has gradually been applied in many fields or industries such as medical care and finance due to its huge data scale, rich semantic relationships, excellent quality, and friendly structure.

[0004] The existing knowledge graph in the power field makes full use of the data information carried by the power Internet of Things to describe the concepts, entities, events and their relationships in the power system in a structured way, providing a more effective cross-media big data organization, management and cognitive capability for the power industry chain. Combining big data and artificial intelligence technology, domain knowledge graphs are gradually becoming one of the core driving forces for the development of power artificial intelligence.

[0005] For example, the invention patent with announcement number CN115577095B discloses a method for recommending electricity standard information based on graph theory, including: obtaining user information and electricity standard information; generating a user vector based on the user information; determining an item offset vector based on the electricity standard information; obtaining corresponding structured knowledge from a pre-constructed electricity standard knowledge graph based on the electricity standard information; determining corresponding text knowledge based on the electricity standard information; vectorizing the structured knowledge and text knowledge to obtain a structured vector and a textual vector respectively; determining an item vector based on the item offset vector, the structured vector and the textual vector; and calculating a recommendation index for the electricity standard information based on the user vector and the item vector.

[0006] For example, the patent application with publication number: CN115934955A discloses a method for constructing an electric power standard knowledge graph, a knowledge question-answering system and a device, which include: constructing an ontology structure of an electric power standard knowledge graph through collected electric power standard data, wherein the ontology structure includes entities, attributes and relationships between entities; obtaining basic data containing electric power standard knowledge, and extracting knowledge from the basic data to extract entities, attributes and relationships between entities; performing knowledge fusion based on the extracted knowledge, and storing the fused knowledge to construct an electric power standard knowledge graph.

[0007] However, in the process of implementing the technical solution of the invention in the embodiments of the present application, the present application found that the above technology has at least the following technical problems:

[0008] In the existing technology, relevant knowledge in the field of electricity usually relies on search engines, and in the process of constructing the electricity standard knowledge graph, it is difficult to extract relevant knowledge, which makes it difficult to construct the knowledge graph. There is also the problem of insufficient data utilization of the knowledge graph during the model construction process. Summary of the invention

[0009] The embodiment of the present application solves the problem of insufficient utilization of knowledge graphs by standard digital reasoning models in the power field during model construction in the prior art by providing a method for constructing a standard digital reasoning model in the power field based on a knowledge graph, thereby improving the data utilization rate of the knowledge graph during model construction.

[0010] The embodiment of the present application provides a method for constructing a digital reasoning model of power field standards based on a knowledge graph, comprising the following steps: obtaining power standard data, and obtaining power knowledge data in combination with preset actual requirements of the knowledge graph to be constructed; extracting the entity structure of the knowledge graph to be constructed from the power standard data, and constructing the power standard knowledge graph in combination with the acquired power knowledge data, and at the same time performing association mapping between the power knowledge data and the circuit network in the power standard knowledge graph; constructing a digital reasoning model in combination with the power performance parameters of the power standard knowledge graph, and performing model verification to obtain a model verification result, wherein the digital reasoning model is used to output the reasoning result by inputting the entity structure and association mapping in the power standard knowledge graph.

[0011] Furthermore, the specific extraction method of the entity structure is as follows: the standard text is divided into different text domains according to the meaning and attributes of the standard text in the power standard data, the text domain is segmented and the content of the text domain is segmented by a string matching method to obtain valid words and invalid words, and the invalid words are filtered by an invalid vocabulary list in a preset knowledge base; it is determined whether the obtained valid words exist in the preset knowledge base, and if not, the valid words are added to the knowledge base vocabulary list, otherwise no operation is performed; entity relationships are extracted from the standard text according to the preset entity relationship type and entity features of the entity structure, and the entity relationships are converted into feature vectors by syntactic analysis and lexical analysis; the feature vector similarity is obtained, and the entity relationships are classified according to the feature vector similarity to obtain the entity structure.

[0012] Furthermore, the specific steps of performing word segmentation processing on the text domain are: performing text cleaning and text standardization on the content in the text domain; selecting a word segmentation method to perform word segmentation processing on the text domain; and performing part-of-speech tagging and named entity recognition on the text domain after word segmentation processing.

[0013] Furthermore, the specific construction method of the power standard knowledge graph is as follows: establish a linkage model; convert the acquired power knowledge data into graph nodes and graph edges in the power standard knowledge graph, and use ontology design software to draw the power standard knowledge graph.

[0014] Furthermore, the specific process of the association mapping is: extracting entity-directly related data items from the power knowledge data, and obtaining mapping indicators by analyzing the circuit network structure in the power standard knowledge graph; locating the entity structure of the entity-directly related data items that meet the mapping indicators to obtain the corresponding entity structure of the entity-directly related data items in the power standard knowledge graph; filling the text information in the power knowledge data into the corresponding position of the power standard knowledge graph according to the created mapping rules; verifying and optimizing the power standard knowledge graph after the association mapping, and regularly updating the power knowledge data and the power standard knowledge graph.

[0015] Furthermore, the steps for constructing the digital reasoning model are as follows: step one, construct a feature data set, and divide the feature data set into a training feature data set and a verification feature data set according to a preset ratio; step two, obtain a training machine model, wherein the training machine model is a converged model obtained by training a preset reasoning model using the training feature data set; step three, use the verification feature data set to perform model verification on the training machine model to obtain a model verification result, and screen the training machine model in combination with a preset threshold condition to obtain a digital reasoning model, wherein the model verification result includes a model performance verification index and a model simulation test index, wherein the model simulation test index is used to describe the model simulation condition of the training machine model, and the model performance verification index is used to describe the performance verification condition of the training machine model.

[0016] Furthermore, the method for obtaining the model simulation test index is as follows: performing model simulation on the training machine model to obtain a corresponding simulation test score, wherein the simulation test score includes fit, accuracy and recall; obtaining the model simulation test index of the training machine model according to the obtained simulation test score, wherein the model simulation test index is calculated using the following formula:

[0017]

[0018] Where e is a natural constant, m represents the number of the training machine model, m = 1, 2, ..., M, M is the total number of training machine models, MN m is the model simulation test index of the mth training machine model, is the fitness of the mth training machine model, Z' m is the accuracy of the mth training machine model, H' m is the recall rate of the mth training machine model.

[0019] Furthermore, the method for obtaining the model performance verification index is as follows: obtaining a performance verification score from a power performance parameter, the performance verification score including computing efficiency and memory consumption, and obtaining a memory consumption coefficient based on memory consumption; and obtaining the model performance verification index of the training machine model based on the obtained computing efficiency and memory consumption coefficient.

[0020] Furthermore, the model performance verification index is calculated using the following formula:

[0021]

[0022] Where m represents the number of the training machine model, m = 1, 2, ..., M, M is the total number of training machine models, JY m is the model performance verification index of the mth training machine model, δ' mis the computational efficiency of the mth training machine model, σ' m is the memory consumption coefficient of the mth training machine model.

[0023] Furthermore, after obtaining the model verification result, it also includes: real-time monitoring of the power field performance data within a preset time interval and obtaining the power change degree, and optimizing the constructed digital reasoning model accordingly; the process of optimizing the constructed digital reasoning model is as follows: extracting the power change score from the power field performance data of the physical structure, the power change score includes the voltage fluctuation score, the voltage flicker frequency, the frequency deviation score and the harmonic content score, obtaining the power change degree of the physical structure according to the obtained power change score, and updating the digital reasoning model in real time according to the power change degree; performing secondary processing on the power standard data and using data enhancement method to expand the power standard data; evaluating the digital reasoning model according to the preset time and updating the digital reasoning model using real-time data stream.

[0024] The embodiment of the present application provides a system for constructing a digital reasoning model of electric power field standards, including: an acquisition module, used to acquire electric power standard data, and acquire electric power knowledge data in combination with preset actual requirements of the knowledge graph to be constructed; a mapping module, used to extract the entity structure of the knowledge graph to be constructed from the electric power standard data, and construct the electric power standard knowledge graph in combination with the acquired electric power knowledge data, and at the same time, perform correlation mapping between the electric power knowledge data and the circuit network in the electric power standard knowledge graph; an execution module, used to construct a digital reasoning model in combination with the electric power performance parameters of the electric power standard knowledge graph, and perform model verification to obtain model verification results, wherein the digital reasoning model is used to output reasoning results by inputting entity structures and correlation mappings in the electric power standard knowledge graph. 12. The system for constructing a digital reasoning model of electric power field standards based on a knowledge graph as claimed in claim 11, characterized in that the mapping module is specifically used to:

[0025] According to the meaning and attributes of the standard text in the power standard data, the standard text is divided into different text domains, the text domain is segmented and the string matching method is used to segment the content of the text domain to obtain valid words and invalid words, and the invalid words are filtered using the invalid word list in the preset knowledge base;

[0026] Determine whether the acquired valid vocabulary exists in the preset knowledge base. If not, add the valid vocabulary to the knowledge base vocabulary list, otherwise, do not perform any operation;

[0027] Entity relations are extracted from standard texts according to the preset entity relation types and entity features of entity structures, and the entity relations are converted into feature vectors by using syntactic analysis and lexical analysis.

[0028] Obtain feature vector similarity, and classify entity relationships according to feature vector similarity to obtain entity structure;

[0029] The mapping module is specifically used to: perform text cleaning and text standardization on the content in the text domain; select a word segmentation method to perform word segmentation processing on the text domain; and perform part-of-speech tagging and named entity recognition on the text domain after the word segmentation processing.

[0030] The mapping module is specifically used to: establish a linkage model; convert the acquired power knowledge data into graph nodes and graph edges in the power standard knowledge graph, and use the ontology design software to draw the power standard knowledge graph.

[0031] The mapping module is specifically used to: extract entity-directly related data items from the power knowledge data, and obtain mapping indicators by analyzing the circuit network structure in the power standard knowledge graph; locate the entity structure of the entity-directly related data items that meet the mapping indicators to obtain the entity structure corresponding to the entity-directly related data items in the power standard knowledge graph; fill the text information in the power knowledge data into the corresponding position of the power standard knowledge graph according to the created mapping rules; verify and optimize the power standard knowledge graph after the association mapping, and regularly update the power knowledge data and the power standard knowledge graph.

[0032] The execution module is specifically used for: step one, constructing a feature data set, and dividing the feature data set into a training feature data set and a verification feature data set according to a preset ratio; step two, obtaining a training machine model, wherein the training machine model is a converged model obtained by training a preset reasoning model using the training feature data set; step three, performing model verification on the training machine model using the verification feature data set to obtain a model verification result, and screening the training machine model in combination with a preset threshold condition to obtain a digital reasoning model, wherein the model verification result includes a model performance verification index and a model simulation test index, wherein the model simulation test index is used to describe the model simulation condition of the training machine model, and the model performance verification index is used to describe the performance verification condition of the training machine model;

[0033] The method for obtaining the model simulation test index is as follows: perform model simulation on the training machine model to obtain the corresponding simulation test score, and the simulation test score includes fit, accuracy and recall; obtain the model simulation test index of the training machine model according to the obtained simulation test score, and the model simulation test index is calculated using the following formula:

[0034]

[0035] Where e is a natural constant, m represents the number of the training machine model, m = 1, 2, ..., M, M is the total number of training machine models, MN m is the model simulation test index of the mth training machine model, is the fitness of the mth training machine model, Z' m is the accuracy of the mth training machine model, H' m is the recall rate of the mth training machine model;

[0036] The method for obtaining the model performance verification index is as follows: obtaining a performance verification score from the power performance parameter, the performance verification score includes computing efficiency and memory consumption, and obtaining a memory consumption coefficient according to the memory consumption; obtaining a model performance verification index of the training machine model according to the obtained computing efficiency and memory consumption coefficient; the model performance verification index is calculated using the following formula:

[0037]

[0038] Where m represents the number of the training machine model, m = 1, 2, ..., M, M is the total number of training machine models, JY m is the model performance verification index of the mth training machine model, δ' m is the computational efficiency of the mth training machine model, σ' m is the memory consumption coefficient of the mth training machine model.

[0039] The mapping module is specifically used to: monitor the power field performance data in the power field within a preset time interval in real time and obtain the power change degree, and optimize the constructed digital reasoning model based on this; the process of optimizing the constructed digital reasoning model is as follows: extract the power change score from the power field performance data of the physical structure, the power change score includes the voltage fluctuation score, the voltage flicker frequency, the frequency deviation score and the harmonic content score, obtain the power change degree of the physical structure according to the obtained power change score, and update the digital reasoning model in real time according to the power change degree; perform secondary processing on the power standard data and use data enhancement method to expand the power standard data; evaluate the digital reasoning model according to the preset time and update the digital reasoning model using real-time data stream.

[0040] In another aspect, an electronic device includes:

[0041] A processor; and a memory arranged to store computer executable instructions which, when executed, cause the processor to perform any of the methods described.

[0042] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0043] 1. Build an electric power standard knowledge graph through electric power standard data and electric power knowledge data, and then build a digital reasoning model based on the preset reasoning rules in the electric power field and combined with the electric power performance parameters of the electric power standard knowledge graph, and perform model verification to obtain the model verification results. Then, monitor the electric power field performance data within the preset time interval in real time, obtain the electric power change degree and optimize it, thereby realizing the efficient construction of the standard digital reasoning model in the electric power field, and then realizing the improvement of the data utilization rate of the knowledge graph in the model construction process, and effectively solving the problem of insufficient data utilization of the knowledge graph in the model construction process in the prior art.

[0044] 2. The simulation test score of the training machine model is obtained by performing a simulation test on the training machine model, and then the model simulation test index of the training machine model is obtained according to the obtained simulation test score. Finally, the model simulation situation of the training machine model is evaluated according to the obtained model simulation test index, thereby realizing the digitization of the model simulation situation of the training machine model, and then realizing a more accurate evaluation of the model simulation situation of the training machine model.

[0045] 3. By obtaining the performance verification score from the power performance parameters, obtaining the memory consumption coefficient based on the memory consumption, and then obtaining the model performance verification index of the training machine model based on the obtained computing efficiency and memory consumption coefficient, and finally evaluating the performance verification of the training machine model based on the obtained model performance verification index, the performance verification of the training machine model is digitized, thereby achieving a more accurate evaluation of the performance verification of the training machine model. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A flow chart of a method for constructing a standard digital reasoning model in the electric power field based on a knowledge graph provided in an embodiment of the present application;

[0047] Figure 2 A flowchart for constructing a digital reasoning model provided in an embodiment of the present application;

[0048] Figure 3 A statistical chart of changes in the model performance verification index provided in an embodiment of the present application. DETAILED DESCRIPTION

[0049] The embodiment of the present application solves the problem of insufficient data utilization of knowledge graphs in the model construction process in the prior art by providing a method for constructing a standard digital reasoning model of the power field based on a knowledge graph. The power standard data is obtained by preprocessing the initial power standard data collected within a preset time interval, and the power knowledge data is obtained in combination with the preset actual needs of the knowledge graph to be constructed. Then, the entity structure of the knowledge graph to be constructed is extracted from the power standard data, and the power standard knowledge graph is constructed in combination with the obtained power knowledge data. At the same time, the power knowledge data and the circuit network in the power standard knowledge graph are mapped by correlation. Then, a digital reasoning model is constructed based on the preset reasoning rules of the power field and in combination with the power performance parameters of the power standard knowledge graph, and the model is verified to obtain the model verification results. Finally, the power field performance data within the preset time interval is monitored in real time and the power change degree is obtained. The constructed digital reasoning model is optimized accordingly, thereby improving the data utilization rate of the knowledge graph in the model construction process.

[0050] The technical solution in the embodiment of the present application is to solve the problem of insufficient data utilization of the knowledge graph in the above model construction process. The overall idea is as follows:

[0051] An electric power standard knowledge graph is constructed through electric power standard data and electric power knowledge data, and then a digital reasoning model is constructed based on the preset reasoning rules in the electric power field and combined with the electric power performance parameters of the electric power standard knowledge graph. The model is verified to obtain the model verification results, and then the electric power performance data in the preset time interval is monitored in real time, and the degree of electric power change is obtained and optimized, thereby achieving the effect of improving the data utilization rate of the knowledge graph during the model construction process.

[0052] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0053] like Figure 1 As shown, it is a flow chart of a method for constructing a digital reasoning model of power field standards based on a knowledge graph provided in an embodiment of the present application, the method comprising the following steps: obtaining power standard data, and obtaining power knowledge data in combination with preset actual requirements of the knowledge graph to be constructed; extracting the entity structure of the knowledge graph to be constructed from the power standard data, and constructing a power standard knowledge graph in combination with the acquired power knowledge data, and at the same time performing association mapping between the power knowledge data and the circuit network in the power standard knowledge graph; constructing a digital reasoning model in combination with power performance parameters of the power standard knowledge graph, and performing model verification to obtain model verification results, wherein the power performance parameters represent parameters in the power standard knowledge graph related to power field performance, and the digital reasoning model is used to output reasoning results by inputting the entity structure and association mapping in the power standard knowledge graph.

[0054] In this embodiment, the initial power standard data represents standard data related to the power field, including power grid operation data, equipment monitoring data and energy market data. The preset actual demand is set by the preset personnel. The power knowledge data represents the data in the initial power standard data related to the preset actual demand of the knowledge graph to be constructed; the preset reasoning rules are obtained from the database and are used to obtain the physical parameter quantity in the power field according to the power performance parameters. The preset reasoning rules represent preset logic or formulas, which are used to convert one type of power performance parameters into another type of physical parameter quantity. The above logic or formula rules are preset before the system is run, usually based on domain knowledge or experience. The rules follow a specific logical structure and are used to derive conclusions under known conditions. The reasoning process can be automatically executed. In a specific embodiment, for example, the input is current and voltage, and the output is the corresponding power according to the preset reasoning rules. For example, the input is power and time, and the output is the corresponding watt-second according to the preset reasoning rules; the entity structure is a component of the power system in the power field, and the power standard knowledge graph is used to store the entity structure and association mapping, which is used to establish the connection between power knowledge data and circuit network. The circuit network includes busbar, generator, synchronous motor, load point, transmission line and circuit breaker entities and the connection relationship between them; power performance parameters are parameters extracted from the power system in the power field for model performance verification, generally including efficiency parameters, reliability parameters, quality parameters, environmental parameters and economic parameters. Efficiency parameters generally include energy efficiency, power factor and load rate, reliability parameters generally include failure rate, reliability and maintenance time, quality parameters generally include fluctuation noise and voltage stability, environmental parameters generally include power consumption, electromagnetic radiation and thermal effect, and economic parameters generally include cost-effectiveness and return on investment; efficiency power system includes sensors and measuring equipment; power field performance data includes stability, efficiency, reliability and quality indicators of power system in the power field; preprocessing includes data cleaning, data deduplication and data formatting operations on the collected initial power standard data to ensure the quality and consistency of power standard data, thereby effectively improving the data utilization rate of knowledge graph in the model construction process.

[0055] Furthermore, the specific extraction method of the entity structure is as follows: the standard text is divided into different text domains according to the meaning and attributes of the standard text in the power standard data, the text domain is segmented and the content of the text domain is segmented by a string matching method to obtain valid words and invalid words, and the invalid words are filtered by an invalid vocabulary list in a preset knowledge base; it is determined whether the obtained valid words exist in the preset knowledge base, if not, the valid words are added to the knowledge base vocabulary list, otherwise no operation is performed, and the knowledge base vocabulary list is used to store valid words; entity relationships are extracted from the standard text according to the preset entity relationship type and the entity features of the entity structure, and the entity relationships are converted into feature vectors by syntactic analysis and lexical analysis; the feature vector similarity is obtained, and the entity relationships are classified according to the feature vector similarity to obtain the entity structure, and the feature vector similarity indicates the similarity between the entity structure and the entity relationship.

[0056] In this embodiment, word segmentation is used to classify the content of the text domain. Standard text represents text information in power standard data. Standard text includes national standards, industry standards, local standards and group standards. String matching methods include maximum forward matching, maximum reverse matching and minimum segmentation. Valid vocabulary represents vocabulary with practical meaning, and invalid vocabulary represents vocabulary without practical meaning. In order to convert the identified entities and relationships into feature vectors, we use the naive Bayes algorithm to train the standard text and learn the representation of entities and relationships. This algorithm has strong generalization ability. The feature vector similarity is obtained through convolutional neural networks, which are used to process complex nonlinear relationships and large amounts of data. Through these methods, we not only improve the accuracy of entity structure extraction, but also use dimensionality reduction technology to reduce the dimension of feature vectors, thereby improving the efficiency of feature vector acquisition. These improvements further improve the data utilization rate of knowledge graphs in the model construction process.

[0057] Furthermore, the specific steps of word segmentation processing on the text domain are: text cleaning and text standardization of the content in the text domain; selecting a word segmentation method to perform word segmentation processing on the text domain; performing part-of-speech tagging and named entity recognition on the text domain after word segmentation processing, where part-of-speech tagging means assigning part-of-speech tags to words after word segmentation processing, and named entity recognition means identifying specific entities in the text.

[0058] In this embodiment, text cleaning means removing HTML tags, URLs and special characters from the content of the text domain, and text specification includes converting the content of the text domain to lowercase and removing punctuation marks; word segmentation methods include space-based word segmentation, rule-based word segmentation and statistical-based word segmentation; space-based word segmentation means segmenting the language of natural separators by spaces, rule-based word segmentation means segmenting by building a dictionary containing all words, and statistical-based word segmentation means learning the probability of vocabulary boundaries using a hidden Markov model HMM and performing word segmentation; text specification depends on the preset actual needs in the power field; the language of natural separators is generally English; part-of-speech tags generally include nouns and verbs; specific entities generally include names of people, places and organizations. These processes improve the efficiency of word segmentation and further improve the data utilization rate of the knowledge graph in the model construction process.

[0059] Furthermore, the specific construction method of the power standard knowledge graph is as follows: establish a linkage model, which is a model for statistical indicators-detailed data-related data-related objects constructed by utilizing the entity structure and connection relationship of power production statistical indicators in the power field; convert the acquired power knowledge data into graph nodes and graph edges in the power standard knowledge graph, and use ontology design software to draw the power standard knowledge graph.

[0060] In this embodiment, the graph nodes represent entity results in the power field, and the graph edges represent the relationship between entity structures; the entity structures of the graph nodes generally include power stations, power grids, equipment, and users; the relationship between the entity structures of the graph edges generally includes power supply relationships and equipment connection relationships. This method improves the efficiency of building the power standard knowledge graph and further improves the data utilization rate of the knowledge graph during the model building process.

[0061] Furthermore, the specific process of association mapping is as follows: extracting entity-directly related data items from the electric power knowledge data, and obtaining mapping indicators by analyzing the circuit network structure in the electric power standard knowledge graph; locating the entity structure of the entity-directly related data items that meet the mapping indicators to obtain the entity structure corresponding to the entity-directly related data items in the electric power standard knowledge graph, and the entity structure positioning is used to determine the entity structure of the entity-directly related data items in the electric power standard knowledge graph; filling the text information in the electric power knowledge data into the corresponding position of the electric power standard knowledge graph according to the created mapping rules, and the mapping rules are used to associate the entity-directly related data items with the entity structure and entity structure relationship in the electric power standard knowledge graph; verifying and optimizing the electric power standard knowledge graph after association mapping, and regularly updating the electric power knowledge data and the electric power standard knowledge graph.

[0062] In this embodiment, the entity directly related data items include power field operating parameters, power field performance indicators and power field maintenance records. The mapping indicators represent the entity structure and entity structure relationship that need to be mapped to the power knowledge data; the circuit network structure analysis requires the analysis of the components in the circuit and their connection methods. The components generally include resistors, capacitors, inductors and power supplies; according to the circuit structure, a mathematical model of the circuit is established, and the relationship between the voltage, current and component parameters of the circuit is listed. The circuit model is the mathematical basis of circuit analysis, so that the circuit analysis can be transformed into the solution of mathematical problems; identify the key nodes and paths in the circuit network, analyze the connection relationship between the nodes, and understand the overall structure and function of the network; based on this, the mapping indicators are determined, which are generally voltage, current and power. These methods improve the connection accuracy between the power knowledge data and the circuit network mapping relationship in the power standard knowledge graph, and further improve the data utilization rate of the knowledge graph during the model construction process.

[0063] Further, such as Figure 2 As shown, it is a flow chart for constructing a digital reasoning model provided in an embodiment of the present application, and the steps for constructing the digital reasoning model are as follows: Step 1, construct a feature data set, the feature data set is a set of power performance parameters extracted from the power standard knowledge graph, and the feature data set is divided into a training feature data set and a verification feature data set according to a preset ratio; Step 2, obtain a training machine model, the training machine model is a converged model obtained by training a preset reasoning model using the training feature data set, and the preset reasoning model is a model with an inference output based on the input content; Step 3, use the verification feature data set to perform model verification on the training machine model to obtain a model verification result, and combine the preset threshold condition to screen the training machine model to obtain a digital reasoning model, the model verification result includes a model performance verification index and a model simulation test index, the model simulation test index is used to describe the model simulation situation of the training machine model, and the model performance verification index is used to describe the performance verification situation of the training machine model.

[0064] In this embodiment, the feature data set is a collection of power performance parameters; the preset threshold conditions include the model performance verification index threshold and the model simulation test index, and the preset threshold conditions are the compliance restriction conditions for the training machine model; screening means putting the training machine model that meets the preset threshold conditions into practical application, and retraining the training machine model that does not meet the preset threshold conditions; using the training feature data set to train the model, adjusting the model parameters through the optimization algorithm to minimize the loss function; using the verification feature data set to evaluate the model performance, and adjusting the model hyperparameters to improve the performance. These steps realize the efficient construction of the digital reasoning model and further improve the data utilization rate of the knowledge graph in the model construction process.

[0065] Furthermore, the method for obtaining the model simulation test index is as follows: perform model simulation on the training machine model to obtain the corresponding simulation test score, which includes fit, accuracy and recall; obtain the model simulation test index of the training machine model according to the obtained simulation test score, and the model simulation test index is calculated using the following formula:

[0066]

[0067] Where e is a natural constant, m represents the number of the training machine model, m = 1, 2, ..., M, M is the total number of training machine models, MN m is the model simulation test index of the mth training machine model, is the fitness of the mth training machine model, Z' m is the accuracy of the mth training machine model, H' m is the recall rate of the mth training machine model.

[0068] In this embodiment, let the simulation rate Then the model simulation test index can be simplified as: like Figure 3 As shown, it is a statistical graph of the changes in the model simulation test index provided in the embodiment of the present application. It can be seen from the graph that the degree of fit and the simulation rate are positively correlated with the model simulation test index; when the degree of fit increases, the larger the model simulation test index is, the better the model simulation situation of the training machine model is; when the simulation rate increases, the larger the model simulation test index is, the better the model simulation situation of the training machine model is; a more accurate evaluation of the model simulation situation of the training machine model is achieved, and the data utilization rate of the knowledge graph in the model construction process is further improved.

[0069] Furthermore, the method for obtaining the model performance verification index is as follows: obtain the performance verification score from the power performance parameters, the performance verification score includes computing efficiency and memory consumption, and obtain the memory consumption coefficient according to the memory consumption; obtain the model performance verification index of the training machine model according to the obtained computing efficiency and memory consumption coefficient.

[0070] In this embodiment, the computational efficiency indicates the computational output speed of the training machine model, and the computational efficiency is obtained by obtaining the computational output of the training machine model per second; the memory consumption coefficient refers to the absolute value ratio of the difference between the memory consumption and the reference value. These indicators realize the numerical verification of the performance of the training machine model, thereby further improving the data utilization rate of the knowledge graph in the model construction process.

[0071] Furthermore, the model performance verification index is calculated using the following formula:

[0072]

[0073] Where m represents the number of the training machine model, m = 1, 2, ..., M, M is the total number of training machine models, JY m is the model performance verification index of the mth training machine model, δ' m is the computational efficiency of the mth training machine model, σ' m is the memory consumption coefficient of the mth training machine model.

[0074] In this embodiment, the expression of the memory consumption coefficient is: Among them, σ m is the memory consumption of the mth training machine model, is the reference memory consumption of the mth training machine model, Δσ is the reference deviation of memory consumption, and the statistical table of changes in the model performance verification index is shown in Table 1:

[0075] Table 1. Statistics of changes in model performance verification index

[0076] <![CDATA[Computational efficiency δ‘ m > <![CDATA[Memory consumption coefficient σ‘ m > <![CDATA[Model performance verification index JY m > 0.424 0.072 0.621 0.646 0.088 0.769 0.438 0.021 0.734 0.963 0.778 0.519 ... ... ...

[0077] It can be seen from the above table that the model performance verification index cannot be directly obtained through a single data, but needs to be obtained through a comprehensive analysis of the computing efficiency and memory consumption coefficient. For example, the computing efficiency of the second group of data is lower than that of the fourth group of data, but the model performance verification index of the second group obtained in the end is higher than that of the fourth group. Because the memory consumption coefficient also has an impact, a comprehensive analysis is required; the performance verification of the training machine model is quantified, and the data utilization rate of the knowledge graph in the model construction process is further improved.

[0078] It should be understood that the reference memory consumption is obtained by analyzing and counting the memory consumption of the training machine model and taking the average value, and the memory consumption reference deviation is obtained by analyzing and counting the memory consumption of the training machine model using the reference deviation calculation formula.

[0079] Furthermore, after obtaining the model verification result, it also includes: real-time monitoring of the power field performance data within a preset time interval and obtaining the power change degree, and optimizing the constructed digital reasoning model based on this; the process of optimizing the constructed digital reasoning model is as follows: extracting the power change score from the power field performance data of the physical structure, the power change score includes the voltage fluctuation score, the voltage flicker frequency, the frequency deviation score and the harmonic content score, obtaining the power change degree of the physical structure according to the obtained power change score, and updating the digital reasoning model in real time according to the power change degree; performing secondary processing on the power standard data and using data enhancement methods to expand the power standard data; evaluating the digital reasoning model according to the preset time and updating the digital reasoning model using real-time data streams.

[0080] In this embodiment, the power variation is calculated using the following formula:

[0081]

[0082] Where d represents the number of the entity structure, d = 1, 2, ..., D, D is the total number of entity structures, BH d is the power variation of the d-th entity structure, V' d is the voltage fluctuation fraction of the d-th entity structure, μ' d is the voltage flicker frequency of the dth entity structure, τ' d is the frequency deviation score of the d-th entity structure, X' d is the harmonic content fraction of the d-th entity structure.

[0083] Among them, a device for constructing a standard digital reasoning model in the power field based on a knowledge graph includes: a memory and power measurement components; the memory is used to store power standard data and power performance parameters; the power measurement components are used to measure entity directly related data items and power change scores.

[0084] Specifically, voltage fluctuation refers to the phenomenon that voltage changes dramatically in a short period of time, which may cause serious damage to the user's electrical equipment. Voltage fluctuation is usually caused by instantaneous load mutations, starting of large motors, etc. The voltage fluctuation score is obtained by measuring the peak and valley values ​​of the voltage curve and subtracting them, and then comparing them with the average voltage fluctuation index. The average voltage fluctuation index is the average value of the voltage fluctuation index, which is obtained by averaging the voltage fluctuation index; voltage flicker refers to the phenomenon that voltage decreases or increases briefly at a certain moment. Voltage flicker may cause the flickering of lighting equipment and the jitter of motors. The voltage flicker frequency measures the duration, amplitude and number of temporary sags and swells; frequency deviation refers to the degree to which the grid frequency deviates from the standard value. Frequency deviation may cause the clock of electrical equipment to run faster or slower, thereby affecting its normal operation. The frequency deviation score is measured by comparing it with the reference frequency and calculating the mean of the deviation; harmonics are the frequency in the power system. The voltage or current component is an integer multiple of the fundamental frequency. Harmonics may cause overload, interference or damage to electrical equipment. The harmonic content score is obtained by measuring the harmonic components of voltage and current and calculating their total distortion rate; the power variability is obtained by; secondary processing is used to improve the data quality of power standard data. The data enhancement method refers to the process of transforming, expanding or synthesizing the original data in deep learning tasks to increase the diversity and quantity of training data, thereby improving the generalization ability and performance of the digital reasoning model. Generally, random noise can be added to the performance data in the power field. This method can improve the model's tolerance to noise, thereby improving its generalization ability; secondary processing includes removing noise, outliers and data balancing. Data balancing means using oversampling, undersampling and data synthesis methods for data with unbalanced data categories to avoid the bias of the digital reasoning model on data categories; the preset time is set by the preset professional; real-time data stream refers to the power field performance data that is continuously generated and flowed, which is characterized by the real-time and continuity of power field performance data; it realizes the improvement of the optimization efficiency of the digital reasoning model and further realizes the improvement of the data utilization rate of the knowledge graph in the model construction process.

[0085] The embodiment of the present application provides a system for constructing a standard digital reasoning model in the electric power field, including: an acquisition module, used to acquire electric power standard data, and acquire electric power knowledge data in combination with preset actual requirements of the knowledge graph to be constructed; a mapping module, used to extract the entity structure of the knowledge graph to be constructed from the electric power standard data, and construct the electric power standard knowledge graph in combination with the acquired electric power knowledge data, and at the same time perform correlation mapping between the electric power knowledge data and the circuit network in the electric power standard knowledge graph; an execution module, used to construct a digital reasoning model in combination with the electric power performance parameters of the electric power standard knowledge graph, and perform model verification to obtain a model verification result, wherein the digital reasoning model is used to output the reasoning result by inputting the entity structure and correlation mapping in the electric power standard knowledge graph.

[0086] The mapping module is specifically used to: divide the standard text into different text domains according to the meaning and attributes of the standard text in the power standard data, perform word segmentation on the text domain and use a string matching method to perform vocabulary segmentation on the content of the text domain to obtain valid words and invalid words, and use an invalid vocabulary list in a preset knowledge base to filter invalid words; determine whether the acquired valid words exist in the preset knowledge base, if not, add the valid words to the knowledge base vocabulary list, otherwise no operation is performed; extract entity relationships from the standard text according to the preset entity relationship type and entity features of the entity structure, and use syntactic analysis and lexical analysis to convert the entity relationships into feature vectors; obtain feature vector similarity, and classify the entity relationships according to the feature vector similarity to obtain the entity structure;

[0087] The mapping module is specifically used for: cleaning and standardizing the content in the text domain; selecting a word segmentation method to perform word segmentation on the text domain; and performing part-of-speech tagging and named entity recognition on the text domain after word segmentation.

[0088] The mapping module is specifically used to: establish a linkage model; convert the acquired power knowledge data into graph nodes and graph edges in the power standard knowledge graph, and use the ontology design software to draw the power standard knowledge graph.

[0089] The mapping module is specifically used to: extract entity-directly related data items from the power knowledge data, and obtain mapping indicators by analyzing the circuit network structure in the power standard knowledge graph; locate the entity structure of the entity-directly related data items that meet the mapping indicators to obtain the corresponding entity structure of the entity-directly related data items in the power standard knowledge graph; fill the text information in the power knowledge data into the corresponding position of the power standard knowledge graph according to the created mapping rules; verify and optimize the power standard knowledge graph after the association mapping, and regularly update the power knowledge data and the power standard knowledge graph.

[0090] The execution module is specifically used for: step 1, constructing a feature data set, and dividing the feature data set into a training feature data set and a verification feature data set according to a preset ratio; step 2, obtaining a training machine model, wherein the training machine model is a converged model obtained by model training a preset reasoning model using the training feature data set; step 3, using the verification feature data set to perform model verification on the training machine model to obtain a model verification result, and screening the training machine model in combination with a preset threshold condition to obtain a digital reasoning model, wherein the model verification result includes a model performance verification index and a model simulation test index, wherein the model simulation test index is used to describe the model simulation situation of the training machine model, and the model performance verification index is used to describe the performance verification situation of the training machine model; a method for obtaining the model simulation test index is as follows: performing model simulation on the training machine model to obtain a corresponding simulation test score, wherein the simulation test score includes fit, accuracy and recall; obtaining the model simulation test index of the training machine model according to the obtained simulation test score, wherein the model simulation test index is calculated using the following formula:

[0091]

[0092] Where e is a natural constant, m represents the number of the training machine model, m = 1, 2, ..., M, M is the total number of training machine models, MN m is the model simulation test index of the mth training machine model, is the fitness of the mth training machine model, Z' m is the accuracy of the mth training machine model, H' m is the recall rate of the mth training machine model.

[0093] The method for obtaining the model performance verification index is as follows: obtain a performance verification score from the power performance parameters, the performance verification score including computing efficiency and memory consumption, and obtain a memory consumption coefficient based on the memory consumption; obtain the model performance verification index of the training machine model based on the obtained computing efficiency and memory consumption coefficient.

[0094] The model performance verification index is calculated using the following formula:

[0095]

[0096] Where m represents the number of the training machine model, m = 1, 2, ..., M, M is the total number of training machine models, JY m is the model performance verification index of the mth training machine model, δ' m is the computational efficiency of the mth training machine model, σ' m is the memory consumption coefficient of the mth training machine model.

[0097] The mapping module is specifically used to: monitor the power field performance data in the preset time interval in real time and obtain the power change degree, and optimize the constructed digital reasoning model based on this; the process of optimizing the constructed digital reasoning model is as follows: extract the power change score from the power field performance data of the physical structure, the power change score includes the voltage fluctuation score, the voltage flicker frequency, the frequency deviation score and the harmonic content score, obtain the power change degree of the physical structure according to the obtained power change score, and update the digital reasoning model in real time according to the power change degree; perform secondary processing on the power standard data and use data enhancement method to expand the power standard data; evaluate the digital reasoning model according to the preset time and update the digital reasoning model using real-time data stream.

[0098] To summarize, the embodiment of the present application constructs an electric power standard knowledge graph through electric power standard data and electric power knowledge data, and then constructs a digital reasoning model based on the preset reasoning rules in the electric power field and in combination with the electric power performance parameters of the electric power standard knowledge graph, and performs model verification to obtain the model verification results, and then monitors the electric power field performance data within the preset time interval in real time and obtains the electric power change degree and optimizes it, thereby realizing the efficient construction of the standard digital reasoning model of the electric power field, and then realizing the improvement of the data utilization rate of the knowledge graph in the model construction process, and effectively solving the problem of insufficient data utilization of the knowledge graph in the model construction process in the prior art.

[0099] At the hardware level, the electronic device includes a processor, and optionally an internal bus, a network interface, and a memory. The memory may include a memory, such as a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage. Of course, the electronic device may also include hardware required for other services.

[0100] The processor, network interface and memory can be interconnected through an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 2 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0101] The memory is used to store the program. Specifically, the program may include a program code, and the program code includes a computer operation instruction. The memory may include a memory and a non-volatile memory, and provides instructions and data to the processor.

[0102] The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming a high-resolution weather forecast device based on multi-source detection and machine learning at the logical level. The processor executes the program stored in the memory and is specifically used to execute any of the aforementioned methods for constructing a standard digital reasoning model in the power field based on the knowledge graph.

[0103] The above application Figure 1 The method for constructing a standard digital reasoning model in the power field based on a knowledge graph disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be directly embodied as being executed by a hardware decoding processor, or executed by a combination of hardware and software modules in a decoding processor. The software module can be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0104] The electronic device may also perform Figure 1 A method for constructing a standard digital reasoning model in the power field based on knowledge graph is proposed, and the Figure 1 The functions of the illustrated embodiment will not be described in detail in the embodiments of the present application.

[0105] An embodiment of the present application also proposes a computer-readable storage medium, which stores one or more programs, and the one or more programs include instructions. When the instructions are executed by an electronic device including multiple applications, any one of the aforementioned methods for constructing a standard digital reasoning model in the power field based on the knowledge graph is executed.

[0106] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0107] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0108] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0109] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0110] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0111] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0112] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0113] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0114] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0115] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for constructing a standard digital reasoning model in the power field, characterized in that: The following steps are involved: Obtain power standard data, and obtain power knowledge data based on the preset actual needs of the knowledge graph to be constructed; Extracting the entity structure of the knowledge graph to be constructed from the power standard data, and constructing the power standard knowledge graph in combination with the acquired power knowledge data, and mapping the power knowledge data with the circuit network in the power standard knowledge graph in terms of association; A digital reasoning model is constructed in combination with the power performance parameters of the power standard knowledge graph, and the model verification is performed to obtain the model verification result. The digital reasoning model is used to output the reasoning result by inputting the entity structure and association mapping in the power standard knowledge graph.

2. The method for constructing a standard digital reasoning model in the electric power field based on a knowledge graph as claimed in claim 1, characterized in that: The specific extraction method of the entity structure is: According to the meaning and attributes of the standard text in the power standard data, the standard text is divided into different text domains, the text domain is segmented and the string matching method is used to segment the content of the text domain to obtain valid words and invalid words, and the invalid words are filtered using the invalid word list in the preset knowledge base; Determine whether the acquired valid vocabulary exists in the preset knowledge base. If not, add the valid vocabulary to the knowledge base vocabulary list, otherwise, do not perform any operation; Entity relations are extracted from standard texts according to the preset entity relation types and entity features of entity structures, and the entity relations are converted into feature vectors by using syntactic analysis and lexical analysis. The feature vector similarity is obtained, and the entity relationship is classified according to the feature vector similarity to obtain the entity structure.

3. The method for constructing a standard digital reasoning model in the electric power field based on a knowledge graph as claimed in claim 2, characterized in that: The specific steps of performing word segmentation processing on the text domain are: Perform text cleaning and text standardization on the content in the text domain; Select the word segmentation method to segment the text field; Perform part-of-speech tagging and named entity recognition on the text domain after word segmentation.

4. The method for constructing a standard digital reasoning model in the electric power field based on a knowledge graph as claimed in claim 1, characterized in that: The specific construction method of the power standard knowledge graph is as follows: Establish linkage model; The acquired power knowledge data is converted into graph nodes and graph edges in the power standard knowledge graph, and the power standard knowledge graph is drawn using ontology design software.

5. The method for constructing a standard digital reasoning model in the electric power field based on a knowledge graph as claimed in claim 1, characterized in that: The specific process of the association mapping is as follows: Extracting directly relevant data items of entities from the power knowledge data, and obtaining mapping indicators by analyzing the circuit network structure in the power standard knowledge graph; Perform entity structure positioning on entity directly related data items that meet mapping indicators to obtain the entity structure corresponding to the entity directly related data items in the electric power standard knowledge graph; Fill the text information in the power knowledge data into the corresponding position of the power standard knowledge graph according to the created mapping rules; The power standard knowledge graph after association mapping is verified and optimized, and the power knowledge data and power standard knowledge graph are updated regularly.

6. The method for constructing a standard digital reasoning model in the electric power field based on a knowledge graph as claimed in claim 1, characterized in that: The steps for constructing the digital reasoning model are as follows: Step 1: construct a feature data set and divide the feature data set into a training feature data set and a verification feature data set according to a preset ratio; Step 2: Obtain a training machine model, where the training machine model is a converged model obtained by training a preset inference model using a training feature data set; Step three, use the verification feature data set to perform model verification on the training machine model to obtain a model verification result, and screen the training machine model in combination with a preset threshold condition to obtain a digital reasoning model. The model verification result includes a model performance verification index and a model simulation test index. The model simulation test index is used to describe the model simulation situation of the training machine model, and the model performance verification index is used to describe the performance verification situation of the training machine model.

7. The method for constructing a standard digital reasoning model in the electric power field based on a knowledge graph as claimed in claim 6, characterized in that: The method for obtaining the model simulation test index is as follows: Performing model simulation on the training machine model to obtain corresponding simulation test scores, wherein the simulation test scores include fit, accuracy and recall; The model simulation test index of the training machine model is obtained according to the obtained simulation test score, and the model simulation test index is calculated using the following formula: Where e is a natural constant, m represents the number of the training machine model, m = 1, 2, ..., M, M is the total number of training machine models, MN m is the model simulation test index of the mth training machine model, is the fitness of the mth training machine model, Z ‘ m is the accuracy of the mth training machine model, H ‘ m is the recall rate of the mth training machine model.

8. The method for constructing a standard digital reasoning model in the electric power field based on a knowledge graph as claimed in claim 6, characterized in that: The method for obtaining the model performance verification index is as follows: Obtaining a performance verification score from the power performance parameter, the performance verification score including computing efficiency and memory consumption, and obtaining a memory consumption coefficient according to the memory consumption; The model performance verification index of the training machine model is obtained according to the obtained computing efficiency and memory consumption coefficient.

9. The method for constructing a standard digital reasoning model in the electric power field based on a knowledge graph as claimed in claim 8, characterized in that: The model performance verification index is calculated using the following formula: Where m represents the number of the training machine model, m = 1, 2, ..., M, M is the total number of training machine models, JY m is the model performance verification index of the mth training machine model, δ ‘ m is the computational efficiency of the mth training machine model, σ ‘ m is the memory consumption coefficient of the mth training machine model.

10. The method for constructing a standard digital reasoning model in the electric power field based on a knowledge graph as claimed in claim 1, characterized in that: After obtaining the model verification result, the method further includes: Real-time monitoring of power performance data in the power sector within a preset time interval and obtaining power variation, based on which the constructed digital reasoning model is optimized; The process of optimizing the constructed digital reasoning model is as follows: Extracting a power change score from the power field performance data of the physical structure, the power change score includes a voltage fluctuation score, a voltage flicker frequency, a frequency deviation score, and a harmonic content score, obtaining a power change degree of the physical structure according to the obtained power change score, and updating the digital reasoning model in real time according to the power change degree; Conduct secondary processing on power standard data and expand power standard data using data enhancement methods; The digital reasoning model is evaluated at a preset time and updated using real-time data streams.

11. A system for constructing a standard digital reasoning model in the electric power field, characterized in that: include: An acquisition module is used to acquire power standard data and acquire power knowledge data in combination with the preset actual requirements of the knowledge graph to be constructed; A mapping module, used to extract the entity structure of the knowledge graph to be constructed from the power standard data, and to construct the power standard knowledge graph in combination with the acquired power knowledge data, and to perform association mapping between the power knowledge data and the circuit network in the power standard knowledge graph; An execution module is used to construct a digital reasoning model in combination with the power performance parameters of the power standard knowledge graph, and perform model verification to obtain a model verification result. The digital reasoning model is used to output the reasoning result by inputting the entity structure and association mapping in the power standard knowledge graph.

12. The system for constructing a standard digital reasoning model in the electric power field based on a knowledge graph as claimed in claim 11, characterized in that: The mapping module is specifically used for: According to the meaning and attributes of the standard text in the power standard data, the standard text is divided into different text domains, the text domain is segmented and the string matching method is used to segment the content of the text domain to obtain valid words and invalid words, and the invalid words are filtered using the invalid word list in the preset knowledge base; Determine whether the acquired valid vocabulary exists in the preset knowledge base. If not, add the valid vocabulary to the knowledge base vocabulary list, otherwise, do not perform any operation; Entity relations are extracted from standard texts according to the preset entity relation types and entity features of entity structures, and the entity relations are converted into feature vectors by using syntactic analysis and lexical analysis. The feature vector similarity is obtained, and the entity relationship is classified according to the feature vector similarity to obtain the entity structure.

13. The system for constructing a standard digital reasoning model in the electric power field based on a knowledge graph as claimed in claim 11, characterized in that: The mapping module is specifically used for: Perform text cleaning and text standardization on the content in the text domain; Select the word segmentation method to segment the text field; Perform part-of-speech tagging and named entity recognition on the text domain after word segmentation.

14. The system for constructing a standard digital reasoning model in the electric power field based on a knowledge graph as claimed in claim 11, characterized in that: The mapping module is specifically used for: Establish linkage model; The acquired power knowledge data is converted into graph nodes and graph edges in the power standard knowledge graph, and the power standard knowledge graph is drawn using ontology design software.

15. The system for constructing a standard digital reasoning model in the electric power field based on a knowledge graph as claimed in claim 11, characterized in that: The mapping module is specifically used for: Extracting directly relevant data items of entities from the power knowledge data, and obtaining mapping indicators by analyzing the circuit network structure in the power standard knowledge graph; Perform entity structure positioning on entity directly related data items that meet mapping indicators to obtain the entity structure corresponding to the entity directly related data items in the electric power standard knowledge graph; Fill the text information in the power knowledge data into the corresponding position of the power standard knowledge graph according to the created mapping rules; The power standard knowledge graph after association mapping is verified and optimized, and the power knowledge data and power standard knowledge graph are updated regularly.

16. The system for constructing a standard digital reasoning model in the electric power field based on a knowledge graph as claimed in claim 11, characterized in that: The execution module is specifically used for: Step 1: construct a feature data set and divide the feature data set into a training feature data set and a verification feature data set according to a preset ratio; Step 2: Obtain a training machine model, where the training machine model is a converged model obtained by training a preset inference model using a training feature data set; Step three, using the verification feature data set to perform model verification on the training machine model to obtain a model verification result, and combining the preset threshold condition to screen the training machine model to obtain a digital reasoning model, the model verification result includes a model performance verification index and a model simulation test index, the model simulation test index is used to describe the model simulation situation of the training machine model, and the model performance verification index is used to describe the performance verification situation of the training machine model; The method for obtaining the model simulation test index is as follows: Performing model simulation on the training machine model to obtain corresponding simulation test scores, wherein the simulation test scores include fit, accuracy and recall; The model simulation test index of the training machine model is obtained according to the obtained simulation test score, and the model simulation test index is calculated using the following formula: Where e is a natural constant, m represents the number of the training machine model, m = 1, 2, ..., M, M is the total number of training machine models, MN m is the model simulation test index of the mth training machine model, is the fitness of the mth training machine model, Z ‘ m is the accuracy of the mth training machine model, H ‘ m is the recall rate of the mth training machine model; The method for obtaining the model performance verification index is as follows: Obtaining a performance verification score from the power performance parameter, the performance verification score including computing efficiency and memory consumption, and obtaining a memory consumption coefficient according to the memory consumption; The model performance verification index of the training machine model is obtained according to the obtained computing efficiency and memory consumption coefficient; The model performance verification index is calculated using the following formula: Where m represents the number of the training machine model, m = 1, 2, ..., M, M is the total number of training machine models, JY m is the model performance verification index of the mth training machine model, δ ‘ m is the computational efficiency of the mth training machine model, σ ‘ m is the memory consumption coefficient of the mth training machine model.

17. The system for constructing a standard digital reasoning model in the electric power field based on a knowledge graph as claimed in claim 11, characterized in that: The mapping module is specifically used for: Real-time monitoring of power performance data in the power sector within a preset time interval and obtaining power variation, based on which the constructed digital reasoning model is optimized; The process of optimizing the constructed digital reasoning model is as follows: Extracting a power change score from the power field performance data of the physical structure, the power change score includes a voltage fluctuation score, a voltage flicker frequency, a frequency deviation score, and a harmonic content score, obtaining a power change degree of the physical structure according to the obtained power change score, and updating the digital reasoning model in real time according to the power change degree; Conduct secondary processing on power standard data and expand power standard data using data enhancement methods; The digital reasoning model is evaluated at a preset time and updated using real-time data streams.

18. An electronic device, comprising: processor; and a memory arranged to store computer executable instructions, which, when executed, cause the processor to perform the method of any one of claims 1 to 10.

Images