Method and system for realizing power big data automatic label based on multi-source heterogeneity

Abstract

The application provides a kind of based on multi-source heterogeneous power big data automation label implementation method, first each power participant is regarded as node and constitutes distributed computer model, obtains the storage data and information data of each node, and according to theme category, the classified storage data and information data are stored in meta information database, the data in meta information database is preprocessed to form standardized data, then the preset automation label generation rule or automation label generation model is called to create computing container locally, and the corresponding label is generated for standardized data by executing label rule or algorithm based on computing container through MapReduce parallel framework, so that the label technology of multi-source heterogeneous power big data is realized, facing equipment label and user label, providing business and intelligent label service and application display for power monitoring and intelligent analysis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of information management, and more particularly to a multi-source heterogeneous power big data automatic label implementation method and system. BACKGROUND

[0002] With the rapid growth of power demand, the scale of power grid is expanding, and the structure and management mode of power business system are also becoming increasingly complex. Under the background of rapid growth of power equipment, the traditional centralized power data management and analysis mode cannot meet the application requirements of high parallelism, high reliability and high fault tolerance in multi-source data management. In the collaborative scenario of multi-system and cross-department, there are still problems such as non-uniform data caliber, inconsistent governance standards, and incompatible analysis methods. The equipment and corresponding user data of different institutions and different stations cannot be uniformly gathered on the platform for management and analysis, and the interconnection between systems cannot be realized. The management and analysis of multi-source heterogeneous data still have problems such as low efficiency and data privacy security.

[0003] Therefore, there is an urgent need for a method or system to manage multi-source cross-institutional power equipment and user data. SUMMARY

[0004] In view of the above problems, the purpose of the present application is to provide a multi-source heterogeneous power big data automatic label implementation method and system to solve the problem that the traditional centralized power data management and analysis mode cannot meet the application requirements of high parallelism, high reliability and high fault tolerance in multi-source data management. In the collaborative scenario of multi-system and cross-department, there are still problems such as non-uniform data caliber, inconsistent governance standards, and incompatible analysis methods.

[0005] The present application provides a multi-source heterogeneous power big data automatic label implementation method, which comprises:

[0006] Each power participant is taken as a node to form a distributed computer model;

[0007] The storage data and information data of each node in the distributed computer model are obtained, classified according to the theme category, and then the storage data and information data classified according to the theme category are stored in a meta-information database;

[0008] The data in the meta-information database is preprocessed based on an algorithm library to form standardized data;

[0009] A preset automatic label generation rule or automatic label generation model is called to create a computing container locally, and the label rule or algorithm is executed based on the computing container through a MapReduce parallel framework to generate corresponding labels for the standardized data.

[0010] Preferably, the power participant at least includes a station system, a business system, and a production system.

[0011] Preferably, the process of obtaining storage data and information data of each node and storing the classified storage data and information data in the meta-information database according to the subject category comprises:

[0012] Collecting local data about each node and storing the local data into a pre-constructed local database to form storage data;

[0013] Registering each node to a service scheduling module of the distributed computer model, obtaining physical location and node website of each node to obtain information data;

[0014] Establishing a subject for the storage data and the information data;

[0015] Dividing the storage data and the information data according to the data specification to form specification data, and storing the specification data in the meta-information database corresponding to the subject.

[0016] Preferably, the data specification at least includes data standard, data structure, data format, data type, data precision, and data belonging equipment naming.

[0017] Preferably, the process of dividing the storage data and the information data according to the data specification comprises:

[0018] Each node obtains a preset subject list through the service scheduling module;

[0019] Selecting a subject in the subject list for subscription and obtaining data specification related to the selected subject;

[0020] Dividing the storage data and the information data according to the data specification.

[0021] Preferably, before the data in the meta-information database is preprocessed based on the algorithm library to form standardized data, it further comprises:

[0022] Formulating an automatic tag generation rule through a rule engine and establishing an algorithm model; wherein the algorithm model comprises a preprocessing model based on a preprocessing algorithm and an automatic tag generation model based on machine learning;

[0023] Storing the automatic tag generation rule and the algorithm model into the algorithm library.

[0024] Preferably, the process of preprocessing data in the meta-information database based on the algorithm library to form normalized data comprises:

[0025] Each node invokes the preprocessing model in the algorithm library according to data in the meta-information database.

[0026] It is judged whether there is a preprocessing algorithm corresponding to the data in the meta-information database in the preprocessing model; wherein,

[0027] If there is a preprocessing algorithm corresponding to the data in the meta-information database, the data in the meta-information database is preprocessed based on the preprocessing algorithm to form normalized data; if there is no preprocessing algorithm corresponding to the data in the meta-information database, a local preprocessing algorithm corresponding to the data in the meta-information database is built locally at the node, and the data in the meta-information database is preprocessed by the local preprocessing algorithm to form normalized data.

[0028] Preferably, the preprocessing algorithm comprises data cleaning, data integration and data reduction.

[0029] Preferably, after the common execution of the label rule or algorithm based on the computing container through the MapReduce parallel framework to generate corresponding labels for the normalized data, it further comprises:

[0030] The labels and models formed after running or training according to the labels are stored in a label library.

[0031] The label library and the meta-information database are connected, and an external search query link of the label library is established.

[0032] The label query is performed through the search query link to invoke data corresponding to the labels.

[0033] The application also provides a multi-source heterogeneous power big data automatic label implementation system, which realizes the above-mentioned system, comprising:

[0034] A computer service unit is used to construct a distributed computer model by taking each power participant as a node.

[0035] A data service unit is used to acquire storage data and information data of each node in the distributed computer model, classify them according to a theme category, store the classified storage data and information data in a meta-information database, and preprocess data in the meta-information database based on an algorithm library to form normalized data.

[0036] A label center unit is configured to invoke preset automatic label generation rules or automatic label generation models to create a computing container locally, and to jointly execute the label rules or algorithms based on the computing container through a MapReduce parallel framework to generate corresponding labels for the normalized data.

[0037] Preferably, the computer service unit comprises a service scheduling module and a storage module, wherein,

[0038] The service scheduling module is configured to provide a service interface for each node, the service interface comprising at least a registration interface for node registration of each node, and a topic interface for each node to obtain a preset topic list; the topic interface comprises a topic subscription interface, a topic meta-information interface, a subject standard interface and a topic modeling interface.

[0039] The storage module is configured to carry an algorithm library and store a rule engine to formulate automatic label generation rules; wherein the algorithm library stores the automatic label generation rules and algorithm models; the algorithm models comprise a preprocessing model based on a preprocessing algorithm and an automatic label generation model based on machine learning.

[0040] From the above technical solution, the method and system for implementing automatic labels of power big data based on multi-source heterogeneity provided by the present application first construct a distributed computer model by taking each power participant as a node, obtain storage data and information data of each node, and store the classified storage data and information data in a meta-information database according to topic categories, then preprocess the data in the meta-information database based on an algorithm library to form normalized data, and then invoke preset automatic label generation rules or automatic label generation models to create a computing container locally, and jointly execute the label rules or algorithms based on the computing container through a MapReduce parallel framework to generate corresponding labels for the normalized data. In this way, the problem of inability to implement unified standards in multi-department cross-systems is solved, information silos in data management systems are broken, and interconnection and intercommunication of power information management are realized, rather than directly providing raw data. Through rule processing and machine learning processing, distributed storage and computing frameworks are utilized, combined with containerization technology, to implement label technology for multi-source heterogeneous power big data, face equipment labels and user labels, and provide business-oriented and intelligent label services and application displays for power monitoring and intelligent analysis. BRIEF DESCRIPTION OF DRAWINGS

[0041] Other objects and results of the present application will become more apparent and easily understood with reference to the following description of the application taken in conjunction with the accompanying drawings, and with the fuller understanding of the present application, it will be apparent that further objects and advantages of the application can be realized and attained by the structure particularly pointed out in the specification. In the drawings:

[0042] Figure 1A flow chart of a multi-source heterogeneous-based power big data automatic label implementation method according to an embodiment of the present application;

[0043] Figure 2 A schematic diagram of a multi-source heterogeneous-based power big data automatic label implementation system according to an embodiment of the present application. DETAILED DESCRIPTION

[0044] The traditional centralized power data management and analysis mode cannot meet the application requirements of high parallelism, high reliability and high fault tolerance in multi-source data management. In the collaborative scene of multi-system cross-department, there are still problems such as non-uniform data caliber, inconsistent governance standards, and incompatible analysis methods. The equipment and corresponding user data between different institutions and different stations cannot be uniformly gathered on the platform for management and analysis, the interconnection between systems cannot be realized, and there are still problems such as low efficiency and data privacy security in the management and analysis of multi-source heterogeneous data.

[0045] In view of the above problems, the present application provides a multi-source heterogeneous-based power big data automatic label implementation method and system, which will be described in detail below with reference to the accompanying drawings.

[0046] In order to illustrate the multi-source heterogeneous-based power big data automatic label implementation method and system provided by the present application, Figure 1 The multi-source heterogeneous-based power big data automatic label implementation method of the embodiment of the present application is exemplarily indicated; Figure 2 The multi-source heterogeneous-based power big data automatic label implementation system of the embodiment of the present application is exemplarily indicated.

[0047] The description of the following exemplary embodiments is actually only illustrative, but by no means as any limitation on the present application and its application or use. The technology and equipment known to those skilled in the related art can not be discussed in detail, but should be considered as part of the specification when appropriate.

[0048] As Figure 1 indicated, the multi-source heterogeneous-based power big data automatic label implementation method of the embodiment of the present application provided by the present application comprises:

[0049] S1: Each power participant is taken as a node to constitute a distributed computer model;

[0050] S2: The storage data and information data of each node in the distributed computer model are acquired, classified according to the theme category, and then the storage data and information data classified according to the theme category are stored in the meta-information database;

[0051] S3: Preprocessing data in the meta-information database based on an algorithm library to form normalized data;

[0052] S4: Calling preset automatic label generation rules or an automatic label generation model to create a computing container locally, and executing label rules or algorithms based on the computing container through a MapReduce parallel framework to generate corresponding labels for the normalized data.

[0053] As shown in Figure 1 Step S1 is a process of building a distributed computer model, i.e., a process of establishing a whole computer model of multi-source heterogeneity, wherein each participant (each node) in the multi-source heterogeneity at least includes a station system, a business system, and a production system, i.e., data of the physical resources or entity nodes of the station system, the business system, and the production system are respectively established and stored to construct a local distributed database, and the data of each node is stored into the distributed database of the respective node.

[0054] In the embodiment shown in Figure 1 Step S2 is to obtain the stored data and information data of each node, and store the classified stored data and information data in the meta-information database according to the subject categories, wherein:

[0055] S21: Obtaining local data about each node, and storing the local data into a pre-constructed local database to form stored data;

[0056] S22: Making each node register with a service scheduling module of the distributed computer model, obtaining the physical location and node website of each node to obtain information data;

[0057] S23: Establishing a subject for the stored data and the information data;

[0058] S24: Dividing the stored data and the information data according to the subject to which the data belongs according to a data specification to form specification data, and storing the specification data in the meta-information database corresponding to the subject; wherein the data specification at least includes: data standard, data structure, data format, data type, data precision, and data belonging to equipment naming;

[0059] The process of dividing the stored data and the information data according to the subject to which the data belongs according to the data specification includes:

[0060] S241: Each node obtains a preset subject list through the service scheduling module;

[0061] S242: Selecting a subject in the subject list for subscription, and obtaining the data specification related to the selected subject;

[0062] S243: Subject classification of the stored data and information data according to data specifications.

[0063] Specifically, in steps S21 and S22, first, the physical resources or entity nodes in the multi-source heterogeneous system register with the service scheduling module, provide relevant information of the entity nodes including physical locations, node URLs to obtain information data, and establish stored data about each node to build a local distributed database, and then store the stored data and information data in the system into the distributed database.

[0064] In steps S23 and S24, subjects are established for the stored data and information data, i.e., a system administrator or project leader establishes corresponding subjects according to application requirements, defines relevant information of the subjects including subject purposes, subject states, and label meta-information, and then performs subject classification of the stored data and information data according to data specifications including data standards, data structures, data formats, data types, data precisions, and data device naming, and stores them into a meta-information database of the subjects.

[0065] In Figure 1 In the embodiment shown, step S3 is a process of preprocessing data in the meta-information database based on an algorithm library to form standardized data, wherein before preprocessing data in the meta-information database based on the algorithm library to form standardized data, step S0 is further included, which comprises:

[0066] S01: Formulate automatic label generation rules through a rule engine and establish an algorithm model; wherein the algorithm model comprises a preprocessing model based on a preprocessing algorithm and an automatic label generation model based on machine learning;

[0067] S02: Store the automatic label generation rules and the algorithm model into the algorithm library.

[0068] Specifically, before step S3, rules need to be pre-established, i.e., according to different data structures and business requirements, the rule engine in the computing service module is used to formulate automatic label generation rules, which are formulated by experts' experience and judgment in this embodiment, and an algorithm model is established and stored into the algorithm library, i.e., a data preprocessing algorithm is constructed, a big data analysis model and a machine learning model are constructed, a preprocessing model is formed based on the preprocessing algorithm, and the big data analysis model and the machine learning model are coded and converted into code images and stored in a mirror warehouse to generate an automatic label generation model, so that labels can be generated directly through the automatic label generation rules in the future, or labels can be generated through the automatic label generation model.

[0069] The process of preprocessing data in the meta-information database based on the algorithm library to form standardized data comprises:

[0070] S31: causing each node to invoke the pre-established preprocessing model in the algorithm library according to the data in the meta-information database;

[0071] S32: determining whether there is a preprocessing algorithm corresponding to the data in the meta-information database in the preprocessing model; wherein,

[0072] S33: if there is a preprocessing algorithm corresponding to the data in the meta-information database, preprocessing the data in the meta-information database based on the preprocessing algorithm to form standardized data; if there is no preprocessing algorithm corresponding to the data in the meta-information database, building a local preprocessing algorithm corresponding to the data in the meta-information database locally, and preprocessing the data in the meta-information database through the local preprocessing algorithm to form standardized data;

[0073] Preprocessing algorithms include data cleaning, data integration, and data reduction.

[0074] Specifically, for multi-source heterogeneous data preprocessing, the entity node (each node) selects the corresponding data preprocessing algorithm from the preprocessing model in the algorithm library. If the algorithm does not exist, the entity node can build the preprocessing algorithm and then create a mirror and store it in the mirror library; wherein the subject modeling in the service scheduling module is used to determine the data preprocessing method. The data preprocessing method includes data cleaning, data integration, data transformation, and data reduction, mainly for normalizing the abnormal values, different sources, formats, characteristics and properties in the original data set. In this embodiment, a series of standardized algorithm preprocessing methods are formulated according to the device type and data properties for participants to use, which helps to improve the efficiency of data analysis.

[0075] Moreover, the entity node uses the mirror (preset) of the data preprocessing algorithm called from the algorithm library for the corresponding data in the subject, creates a preprocessing computing container using the data and the preprocessing algorithm mirror, the preprocessing computing container will execute the specified data preprocessing algorithm, and then obtains the preprocessed data from the computing container and stores it in the local distributed database.

[0076] In Figure 1In the illustrated embodiment, step S4 is to call preset automatic label generation rules or automatic label generation models to create a computing container locally, and to execute label rules or algorithms based on the computing container through a MapReduce parallel framework to generate corresponding labels for the normalized data. The process is a label generation process. In step S0, the automatic label generation rules and the automatic label generation models have been preset in the algorithm library. Therefore, in the label generation process, each entity node executes the generation of labels based on the MapReduce parallel framework. The label generation method can be based on automatic label generation rules or based on machine learning processing, i.e., automatic label generation models. Based on the automatic label generation rules, a specified label generation image is obtained from a preset image repository. Similarly, the method based on machine learning includes a feature extraction method, a supervised learning model, a clustering method, a deep learning model, and the like to obtain a specified algorithm from the preset image repository. Then, each node uses the obtained rules or algorithms to create a label computing container locally, and then generates labels based on the label computing container.

[0077] In addition, after the labels are generated based on the computing container through the MapReduce parallel framework, step S5 is further included, which comprises:

[0078] S51: storing the labels and models formed after running or training according to the labels into a label library;

[0079] S52: connecting the label library with a meta-information database, and establishing an external search query link of the label library;

[0080] S53: performing label query through the search query link to call data corresponding to the labels.

[0081] Specifically, after the label computing container is executed, the generated labels are stored into a label library of the label center. The storage includes the generated labels, models or rules after running or training, and the like. Subsequently, a label search and application service can be applied. The labels can be searched and queried through the label center, intelligent sorting of label query and recommendation of labels are provided, and the like. In the application service of the labels, the label center provides functions of label combination and label management for different business requirements. Meanwhile, the trained label models can be used to provide functions of prediction and display of labels for newly generated data, and the like.

[0082] As described above, the method for realizing automatic labeling of electric power big data based on multi-source heterogeneity provided by the present invention first forms a distributed computer model with each electric power participant as a node, obtains the storage data and information data of each node, and stores the classified storage data and information data in the metadata database according to the subject category, and then pre-processes the data in the metadata database to form normalized data, and then calls the preset automatic label generation rules or automatic label generation model to create a computing container locally, and executes the labeling rules or algorithms based on the computing container through the MapReduce parallel framework to generate corresponding labels for the normalized data. In this way, the problem of the inability to achieve a unified caliber across multiple departments and systems is solved, the information islands in the data management system are broken, and the interconnection and interoperability of electric power information management is realized, rather than directly providing raw data. Through rule-based processing and machine learning processing, distributed storage and computing frameworks are utilized, and containerization technology is combined to realize labeling technology of multi-source heterogeneous electric power big data, oriented to device labels and user labels, to provide business-oriented and intelligent labeling services and application displays for electric power monitoring and intelligent analysis.

[0083] like Figure 2 As shown, the present invention also provides a system 100 for realizing automatic labeling of electric power big data based on multi-source heterogeneity, which realizes the aforementioned method for realizing automatic labeling of electric power big data based on multi-source heterogeneity, including:

[0084] Computer service unit 101, used to construct a distributed computer model by taking each power participant as a node;

[0085] The data service unit 102 is used to obtain the storage data and information data of each node of the distributed computer model, classify them according to the subject category, store the storage data and information data classified according to the subject category in the metadata database, and pre-process the data in the metadata database based on the algorithm library to form normalized data;

[0086] The label center unit 103 is used to call the preset automatic label generation rules or automatic label generation model to create a computing container locally, and jointly execute the label rules or algorithms based on the computing container through the MapReduce parallel framework to generate corresponding labels for normalized data.

[0087] like Figure 2 In the embodiment shown, the computer service unit 101 includes a service scheduling module 101-1 and a storage module 101-2, wherein:

[0088] The service scheduling module 101-1 is configured to provide a service interface for each node, and the service interface at least includes a registration interface for node registration of each node, and a topic interface for each node to obtain a preset topic list; the topic interface includes a topic subscription interface, a topic meta-information interface, a subject standard interface and a topic modeling interface;

[0089] The storage module 101-2 is configured to carry an algorithm library and store a rule engine to formulate an automatic label generation rule; wherein the algorithm library stores the automatic label generation rule and an algorithm model; the algorithm model includes a preprocessing model based on a preprocessing algorithm and an automatic label generation model based on machine learning.

[0090] As can be seen from the above embodiments, the multi-source heterogeneous power big data automatic label implementation system provided by the application first constructs a distributed computer model by taking each power participant as a node through the computer service unit 101, then acquires storage data and information data of each node through the data service unit 102, and stores the classified storage data and information data in the meta-information database according to the topic category, and then preprocesses the data in the meta-information database to form standardized data, and then calls the preset automatic label generation rule or automatic label generation model through the label center unit 103 to create a computing container locally, and executes the label rule or algorithm based on the computing container through the MapReduce parallel framework to generate corresponding labels for the standardized data, thus solving the problem that unified standards cannot be realized in multi-department cross-systems, breaking the information silos in the data management system, realizing the interconnection of power information management, and not directly providing raw data, but through the rule processing and machine learning processing mode, using the distributed storage and computing framework, combining the containerization technology, realizing the label technology of multi-source heterogeneous power big data, facing the device label and user label, and providing business and intelligent label services and application display for power monitoring and intelligent analysis.

[0091] The multi-source heterogeneous power big data automatic label implementation method and system according to the application are described above with reference to the accompanying drawings by way of example. However, those skilled in the art should understand that various improvements can be made to the multi-source heterogeneous power big data automatic label implementation method and system according to the application described above without departing from the content of the application. Therefore, the protection scope of the application should be determined by the content of the appended claims.

Claims

1. A method for realizing automatic labeling of electric power big data based on multi-source heterogeneity, characterized in that: include: Each power participant is regarded as a node to form a distributed computer model; Acquiring storage data and information data of each node in the distributed computer model, classifying them according to subject categories, and then storing the storage data and information data classified according to the subject categories in a metadata database; The method includes: collecting local data about each node and storing the local data in a pre-built local database to form storage data; registering each node with the service scheduling module of the distributed computer model to obtain the physical location and node URL of each node to obtain information data; establishing topics for the storage data and the information data; dividing the storage data and the information data into topics according to data specifications to form specification data, and storing the specification data in a meta-information database corresponding to the topic; the meta-information database is a database that stores relevant information about the topic; Preprocessing the data in the metadata database based on an algorithm library to form normalized data; A preset automatic label generation rule or automatic label generation model is called to create a computing container locally, and the label rule or algorithm is jointly executed based on the computing container through the MapReduce parallel framework to generate corresponding labels for the normalized data.

2. The method for realizing automatic labeling of electric power big data based on multi-source heterogeneity according to claim 1, characterized in that: The power participants include at least the site system, business system, and production system.

3. The method for realizing automatic labeling of electric power big data based on multi-source heterogeneity according to claim 1, characterized in that: The data specifications include at least: data standards, data structure, data format, data type, data precision, and the name of the device to which the data belongs.

4. The method for realizing automatic labeling of electric power big data based on multi-source heterogeneity according to claim 3, characterized in that: The process of classifying the stored data and the information data by subject according to data specifications includes: Each node obtains a preset topic list through the service scheduling module; Select a topic from the topic list to subscribe to, and obtain data specifications related to the selected topic; The stored data and the information data are divided into subject categories according to the data specifications.

5. The method for realizing automatic labeling of electric power big data based on multi-source heterogeneity according to claim 1, characterized in that: Before preprocessing the data in the metadata database based on the algorithm library to form normalized data, the method further includes: Formulate automated label generation rules through a rule engine and establish an algorithm model; wherein the algorithm model includes a preprocessing model based on a preprocessing algorithm and an automated label generation model based on machine learning; The automatic tag generation rules and algorithm models are stored in the algorithm library.

6. The method for realizing automatic labeling of electric power big data based on multi-source heterogeneity according to claim 5, characterized in that: The process of preprocessing the data in the metadata database based on the algorithm library to form normalized data includes: enabling each node to call the preprocessing model in the algorithm library according to the data in the metadata database; Determine whether there is a preprocessing algorithm corresponding to the data in the metadata database in the preprocessing model; wherein, If there is a preprocessing algorithm corresponding to the data in the metadata database, the data in the metadata database is preprocessed based on the preprocessing algorithm to form normalized data; if there is no preprocessing algorithm corresponding to the data in the metadata database, a local preprocessing algorithm corresponding to the data in the metadata database is built locally on the node, and the data in the metadata database is preprocessed using the local preprocessing algorithm to form normalized data; The preprocessing algorithm includes: data cleaning, data integration, and data reduction.

7. The method for realizing automatic labeling of electric power big data based on multi-source heterogeneity according to claim 5, characterized in that: After executing labeling rules or algorithms based on the computing containers through the MapReduce parallel framework to generate corresponding labels for the normalized data, the method further includes: Storing the labels and the models formed after running or training according to the labels in a label library; Connecting the tag library with the meta information database and establishing an external search query link for the tag library; A tag query is performed through the search query link to retrieve data corresponding to the tag.

8. A system for realizing automatic labeling of electric power big data based on multi-source heterogeneity, implementing the method for realizing automatic labeling of electric power big data based on multi-source heterogeneity according to any one of claims 1 to 7, comprising: A computer service unit, used to construct a distributed computer model using various power participants as nodes; a data service unit, configured to obtain storage data and information data of each node in the distributed computer model, classify the storage data and information data according to the categories of themes, store the storage data and information data classified according to the categories of themes in a metadata database, and preprocess the data in the metadata database based on an algorithm library to form normalized data; specifically configured to: collect local data about each node, and store the local data in a pre-built local database to form storage data; enable each node to register with the service scheduling module of the distributed computer model, obtain the physical location and node website of each node to obtain information data; establish topics for the storage data and information data; classify the storage data and information data by topics according to data specifications to form specification data, and store the specification data in the metadata database corresponding to the topics to which they belong; The meta information database is a database that stores relevant information of the subject; The label center unit is used to call the preset automatic label generation rules or automatic label generation model to create a computing container locally, and jointly execute the label rules or algorithms based on the computing container through the MapReduce parallel framework to generate corresponding labels for the normalized data.

9. The multi-source heterogeneous power big data automation labeling system according to claim 8, characterized in that: The computer service unit includes a service scheduling module and a storage module, wherein: The service scheduling module is used to provide a service interface for each node, and the service interface includes at least a registration interface for each node to register the node, and a topic interface for each node to obtain a preset topic list; the topic interface includes a topic subscription interface, a topic meta-information interface, a subject standard interface, and a topic modeling interface; The storage module is used to carry the algorithm library and store the rule engine to formulate automatic label generation rules; wherein, the algorithm library stores the automatic label generation rules and algorithm models; the algorithm model includes a preprocessing model based on the preprocessing algorithm and an automatic label generation model based on machine learning.

Images