Multi-domain data fusion method, system and corresponding equipment based on multi-source data

Abstract

The embodiment of the present application provides a multi-domain data fusion method, system and corresponding equipment based on multi-source data, including: S10, acquiring business data in multiple power fields; S20, data identification of the acquired business data so that each business data has a unique identification; S30, preprocessing the identified business data; S40, building a knowledge graph based on the preprocessed business data; S50, optimizing the knowledge graph based on a deep reinforcement learning model and fusing multi-source data; S60, performing data query based on the optimized knowledge graph to obtain query results; it can effectively improve the interoperability of multi-source data and improve the operating efficiency and service level of power business. Applicable to the technical field of power data fusion.

Description

Technical Field

[0001] The present application relates to the technical field of power data fusion, and in particular to a multi-domain data fusion method, system and corresponding equipment based on multi-source data. Background Art

[0002] Data fusion technology refers to the process of comprehensively processing data from multiple data sources or sensors to obtain more accurate, complete and useful information. Data fusion technology originated from the sonar signal processing system of the US Department of Defense, and was later gradually applied to many fields, such as geology, seismology, remote sensing, ships, gas concentration, news, scientific evaluation, copyright, metadata integration framework, etc. The research path and methods of data fusion technology include four levels: data processing model, stage-based strategy, feature representation-based method and semantic-based application. At present, the application fields of data fusion technology include multi-sensor systems, biomedicine, environmental research and urban construction. Each field has common or special application challenges.

[0003] Foreign countries started research on data fusion technology earlier. In 1973, the U.S. Department of Defense developed a sonar signal processing system and promoted data fusion technology as an emerging discipline. Subsequently, the United States established a data fusion expert group (DataFusion Group) to formulate standards and models related to data fusion. At present, the United States is in a leading position in the world in data fusion technology, which is mainly used in military, security, medical, transportation and other fields. For example, the United States uses multi-source data fusion technology for monitoring and analysis. In addition to the United States, Europe, Japan and other places are also actively carrying out research and application of data fusion technology. In 2004, Europe launched a project called "Information Society Technology" to promote the development and application of multi-source data fusion technology. In 2007, Japan established the "Large-Scale Sensor Network and Intelligent Information Processing" Research Center, which is committed to exploring the theory and methods of multi-sensor information fusion technology. The data fusion technology of these countries and regions is mainly used in environmental monitoring, disaster warning, intelligent transportation and other fields.

[0004] my country's research on data fusion technology began in the early 1980s, initially focusing on multi-target tracking and multi-sensor information fusion; in the early 1990s, with the development of computer technology, artificial intelligence and other technologies, data fusion technology gradually involved military, aviation, transportation, medical, environmental and other fields.

[0005] At present, my country's data fusion technology has many application cases in the fields of Internet, government, finance, telecommunications, etc., such as Alibaba's City Brain Project, Huawei's Smart City Solutions, and JD's Intelligent Supply Chain Platform.

[0006] However, in terms of multi-domain data fusion, there are widespread problems such as poor data coordination and interoperability, which restrict the application and development of multi-source data. Summary of the invention

[0007] In order to solve one of the above-mentioned technical defects, the present application provides an embodiment.

[0008] According to a first aspect of an embodiment of the present application, a multi-domain data fusion method based on multi-source data is provided, comprising the following steps:

[0009] S10, obtaining business data in multiple power fields;

[0010] S20, performing data identification on the acquired business data so that each business data has a unique identification;

[0011] S30, pre-processing the marked business data;

[0012] S40, building a knowledge graph based on the preprocessed business data;

[0013] S50, optimizes the knowledge graph based on the deep reinforcement learning model and integrates multi-source data;

[0014] S60, performing data query according to the optimized knowledge graph to obtain query results.

[0015] According to a second aspect of an embodiment of the present application, a multi-domain data fusion system based on multi-source data is provided, including:

[0016] A data acquisition module is used to acquire business data in multiple power fields; and to identify the acquired business data so that each business data has a unique identification; and to pre-process the identified business data;

[0017] The data fusion module is used to build a knowledge graph based on preprocessed business data, optimize the knowledge graph based on the deep reinforcement learning model, and fuse multi-source data;

[0018] A data storage module, used for storing the fused data;

[0019] The data query module is used to receive input query operations and perform data queries based on the optimized knowledge graph to obtain query results.

[0020] According to a third aspect of an embodiment of the present application, there is provided an electronic device, including:

[0021] memory; processor; and computer program;

[0022] The computer program is stored in the memory and is configured to be executed by the processor to implement the method as described above.

[0023] According to a fourth aspect of an embodiment of the present application, a computer-readable storage medium is provided, on which a computer program is stored; the computer program is executed by a processor to implement the method as described above.

[0024] The embodiment of the present application adopts the above technical solution, which has the following technical effects:

[0025] This application adopts a multi-domain data fusion application method based on multi-source data, and uses multi-source data preprocessing, fusion and query technologies to solve problems such as semantic differences, structural differences, and quality differences among multi-source data, achieve unified management and interoperability of multi-source data, and improve the operating efficiency and service level of the power business. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0027] Figure 1 A flowchart of a multi-domain data fusion method based on multi-source data provided in an embodiment of the present application;

[0028] Figure 2 This is a flow chart of step S40 in the embodiment of the present application;

[0029] Figure 3 This is a flow chart of step S50 in the embodiment of the present application;

[0030] Figure 4 This is a schematic diagram of the process of multi-source data fusion in an embodiment of the present application;

[0031] Figure 5 A schematic diagram of the structure of a multi-domain data fusion system based on multi-source data provided in an embodiment of the present application;

[0032] Figure 6 A specific application diagram of a multi-domain data fusion system based on multi-source data provided in an embodiment of the present application;

[0033] In the figure:

[0034] 10 is a data acquisition module, 20 is a data fusion module, 30 is a data storage module, 40 is a data query module; 50 is a data display module;

[0035] 201 is a knowledge graph module, and 202 is a multi-source data fusion module;

[0036] 2021 is the deep reinforcement learning model and 2022 is the training module. DETAILED DESCRIPTION

[0037] In order to make the technical solutions and advantages in the embodiments of the present application more clearly understood, the exemplary embodiments of the present application are further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than an exhaustive list of all the embodiments. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0038] In this application, driven by typical business scenarios such as customer profiling, precision marketing, risk control, etc., we explore how to extract valuable information from massive data of different sources, types, and fields, and conduct effective integration and analysis, aiming to solve the current problems of data silos, data quality, data utilization, etc., and enhance the data value and competitiveness of the power business.

[0039] This application is based on business scenarios and uses multi-source data fusion technology to integrate and process data from different fields and sources. On the basis of data traceability, knowledge graphs, reinforcement learning and other technologies, it can realize the application of multiple scenarios of power system data (such as: intelligent analysis of power theft and leakage, public transformer loss analysis, and transformer line loss analysis).

[0040] Embodiment 1

[0041] like Figure 1 As shown, the multi-domain data fusion method based on multi-source data includes the following steps:

[0042] S10, obtaining business data in multiple power fields;

[0043] S20, performing data identification on the acquired business data so that each business data has a unique identification;

[0044] S30, pre-processing the marked business data;

[0045] S40, building a knowledge graph based on the preprocessed business data;

[0046] S50, optimizes the knowledge graph based on the deep reinforcement learning model and integrates multi-source data;

[0047] S60, performing data query according to the optimized knowledge graph to obtain query results.

[0048] In this embodiment, it also includes:

[0049] Establish a query interface to receive data query operations.

[0050] In this embodiment, power business data includes data from different sources, types, and fields, such as power grid operation and equipment detection or monitoring data, power enterprise marketing data, and power enterprise management data. By using multi-source data fusion technology, power business data from different sources, types, and fields can be effectively integrated and analyzed, thereby improving the operating efficiency and service quality of the power industry and realizing intelligent and digital transformation.

[0051] Specifically, the business data in multiple power fields are various data generated during the operation of the power system, ensuring the diversification of data sources and covering different types and fields; specifically, they may include: power load data, generator operation data, transmission line data, substation equipment data, etc.

[0052] For example: real-time data collected by monitoring equipment and sensors, such as current, voltage, power, etc.; historical data of power system operation, such as status records, fault information, maintenance records, etc.

[0053] like Figure 2 As shown, the S40 constructs a knowledge graph according to the preprocessed operation data; including:

[0054] S401, modeling the entity of the preprocessed operation data;

[0055] S402, performing relationship modeling according to the relationship between entities;

[0056] S403, mapping entities and relationships to entities and relationships in the knowledge graph to achieve semantic representation of business data.

[0057] Specifically, the entity is a noun with a specific meaning or strong referentiality in the operation data, including: a person's name, a place name, an event name, a product name or a professional name; each entity has attributes that describe its characteristics;

[0058] The S403, mapping entities and relationships to entities and relationships in the knowledge graph to achieve semantic representation of business data, includes:

[0059] S4031, mapping the identified entity to a node in the knowledge graph, and storing the attribute of the node into the attribute value of the node;

[0060] S4032, mapping the relationship between entities into an edge connecting two nodes in the knowledge graph, and storing the attributes describing the relationship characteristics into the attribute values ​​of the edge;

[0061] S4033, realize semantic representation of data based on the nodes and relationships mapped to the knowledge graph.

[0062] In this embodiment, the properties of a node (entity) may include:

[0063] Data type: describes the type or format of data, such as text, numbers, dates, etc. In the subsequent process, appropriate processing methods and analysis tools can be selected according to the data type to ensure effective processing and utilization of data;

[0064] Content description: Content description provides a detailed description of the data content, including the meaning, characteristics, and structure of the data; these descriptions can help users better understand the data and conduct data analysis and application more accurately.

[0065] Generation time: The generation time records the time when the data is generated or obtained. In the subsequent process, the data can be analyzed for time series, predicted, and other operations based on the generation time to better understand the changing trends and patterns of the data.

[0066] The role of the above-mentioned attributes is to help users better understand and use data, and provide important references and bases in data processing, analysis and application. By considering and utilizing these attributes, data management, analysis and mining can be achieved more effectively.

[0067] In this embodiment, the entity may include:

[0068] Power load data: load size, load type, load period, etc.;

[0069] Generator data: generator type, power generation, power generation efficiency, etc.;

[0070] Transmission line data: line length, line load, line impedance, etc.;

[0071] Substation equipment data: equipment type, equipment status, equipment parameters, etc.

[0072] Interrelationships may include:

[0073] The connection relationship between the generator and the transmission line means that the electric energy output by the generator is transmitted to the load end through the transmission line;

[0074] The dependency relationship between substation equipment represents the interaction relationship between substation equipment, such as the connection relationship between transformers and circuit breakers.

[0075] Mapping entities and relations to entities and relations in the knowledge graph may include:

[0076] A knowledge graph model is constructed based on the data elements (entities) of the power system and the relationships between them. The nodes in the model represent the components of the power system, and the edges represent the relationships between them. Therefore, generators, transmission lines, substations and other equipment can be represented as nodes in the graph, and the connections between them can be represented as edges in the knowledge graph.

[0077] In a specific embodiment, there is a set of raw data (business data), which describes the operation status of a generator in the power system; the raw data may include the following fields:

[0078] Generator number: used to uniquely identify the generator;

[0079] Generator output power: describes the current output power of the generator;

[0080] Generator operating status: indicates the current operating status of the generator, such as normal operation, shutdown, fault, etc.;

[0081] Generator running time: record the running time of the generator from startup to now;

[0082] Generator temperature: describes the current temperature of the generator;

[0083] Other related parameters: such as voltage, current, frequency, etc.

[0084] After the original data is modeled, possible data elements include:

[0085] (1) Generator entity: represents a generator in the power system, with a generator number as a unique identifier;

[0086] Attributes corresponding to the generator entity: generator operating status, generator operating time, generator temperature, etc.;

[0087] Identifier corresponding to the generator entity: generator number.

[0088] (2) Generator output power entity: represents the output power of the generator;

[0089] Attributes corresponding to the generator output power entity: generator output power value;

[0090] The identifier corresponding to the generator output power entity: it can be a combination of the generator number and the timestamp.

[0091] (3) Other related entities: Depending on the specific situation, they may include temperature entities, operating status entities, etc.

[0092] Relationships can be established between data elements. For example, the relationship between the generator entity and the generator output power entity represents the change of the generator output power over time.

[0093] By modeling entities and the relationships between entities, the original data in the power system can be made clearer and more understandable, and facilitate subsequent data analysis and application.

[0094] In this embodiment, the power system data can be traced back by querying the knowledge graph model; for example:

[0095] Query the generator and transmission line information that a specific load relies on to understand the power supply source of the load;

[0096] Query historical fault information on the transmission line to analyze the fault cause and treatment process;

[0097] In this embodiment, the knowledge graph can be used to build a traceability model for the power system to help track the relationship between various data elements during the operation of the power system, thereby improving the operating efficiency and reliability of the power system.

[0098] like Figure 3 As shown, the S50 optimizes the knowledge graph based on the deep reinforcement learning model and fuses multi-source data; including:

[0099] S501, constructing a deep reinforcement learning model, wherein the deep reinforcement learning model includes: an environment model and an agent, wherein the agent is represented as a policy network; and the environment model is composed of a knowledge graph;

[0100] The strategy network outputs a decision result according to the data state information input by the environment model;

[0101] S502, initializing parameters of the policy network; wherein the parameters include: actions and reward values;

[0102] S503, using historical data as training samples to train the deep reinforcement learning model, and updating the parameters of the deep reinforcement learning model through interaction with the environment model to optimize the data alignment and integration effect of the knowledge graph.

[0103] like Figure 4 As shown, in this embodiment, reinforcement learning can be used to learn the association rules between different data sources during the multi-source data fusion process to achieve more accurate data alignment and integration.

[0104] Specifically, the construction of the deep reinforcement learning model may include:

[0105] S5011, Environmental Modeling, modeling the data integration environment; including: defining data sources, data formats, data quality assessment standards, etc. Environmental modeling can be used to strengthen learning algorithms to help understand the background and requirements of data integration;

[0106] S5012, state representation: representing the state of the data source; the state may include: characteristics, attributes, structure and other information of the data, which may be encoded into vectors or other forms of data representation for processing by the reinforcement learning algorithm;

[0107] In the state identification stage, the state of the data source needs to be effectively represented so that it can be processed by the reinforcement learning algorithm. This process needs to consider the relationship between data input and output and convert the data into a form suitable for reinforcement learning algorithm processing. The specific steps may include:

[0108] Data feature extraction: Extract useful features from raw data, which can describe the important attributes and characteristics of the data. For example, in the power system, features can include current, voltage, power, etc.

[0109] Data encoding: Encode the extracted features so that they can be processed in a computer. Common encoding methods include one-hot encoding, label encoding, embedded encoding, etc.

[0110] Data structure representation: Consider the structural relationship between data and represent the data as a data structure suitable for reinforcement learning algorithm processing. This may involve converting data into vectors, matrices or other forms of data representation.

[0111] Data standardization: Standardizing data to eliminate scale differences between different data and ensure that the data has a uniform dimension, which helps improve the training effect of reinforcement learning algorithms.

[0112] Clarify the relationship between data input and output: Ensure that the relationship between data input and output is clearly reflected in the state representation process; this can be done by matching input data and output data to ensure that the state representation can fully reflect the mutual influence and dependency between data.

[0113] Flexibility in data representation: Ensure flexibility in state representation to accommodate different types of data and processing scenarios; this may involve applying different representation methods to different types of data to meet the needs of different tasks.

[0114] Through the above steps, the state of the data source can be effectively represented in a form suitable for reinforcement learning algorithm processing, thus providing a good foundation for subsequent reinforcement learning model training.

[0115] S5013, Action Definition: Defines the actions that the system can take; in the context of data alignment and integration, actions can be operations such as data transformation, matching, and cleaning;

[0116] S5014, Reward Design: Design a reward function to evaluate the effectiveness of each action of the system; the reward function can be based on indicators such as the accuracy of data alignment, the quality of integrated data, and efficiency; the design of the reward function should reflect the goals and requirements of data integration as much as possible.

[0117] This embodiment also includes:

[0118] Reinforcement learning algorithm selection: Select a suitable deep reinforcement learning algorithm. The selected algorithm should be able to adapt to the complex environment of data integration and have good convergence and generalization capabilities.

[0119] In this embodiment, the reinforcement learning algorithm used by the deep reinforcement learning model should take into account the temporal nature and nonlinear relationships of the data to better capture the complex relationships between the data.

[0120] Specifically, considering the temporal and nonlinear characteristics of data to better capture the complex relationships between data is closely related to the choice of reinforcement learning algorithms. These characteristics may involve the following aspects in the process of data fusion:

[0121] Processing of time-series data: If the data is time-series, that is, the data changes over time, then in model design, it is necessary to consider how to effectively process this time-series information. A common approach is to use models such as recurrent neural networks (RNN) or long short-term memory networks (LSTM), which can capture the temporal dependencies between data and provide better performance in data fusion.

[0122] Modeling of nonlinear relationships: If there are complex nonlinear relationships between data, linear models may not be able to capture these relationships well; therefore, it is necessary to select learning algorithms that can handle nonlinear relationships, such as deep neural networks; these algorithms can model complex relationships between data through multi-level nonlinear transformations, thereby improving the performance of the fusion model.

[0123] Importance of feature engineering: In addition to choosing a suitable learning algorithm, effective feature engineering is also required to convert data into feature representations that can better reflect the relationship between data; this may involve operations such as sliding window processing of time series data and extracting time series features, so as to convert the temporal nature and nonlinear relationship of the data into a form that can be processed by the learning algorithm.

[0124] The trade-off between model complexity and generalization ability: When considering temporal and nonlinear relationships, it is also necessary to balance the model complexity and generalization ability; an overly complex model may lead to overfitting, while an overly simple model may not be able to capture the complex relationships between data well. Therefore, it is necessary to select the appropriate model complexity according to the specific situation to achieve a balance between accurate modeling of data relationships and generalization ability.

[0125] In summary, considering the temporal nature and nonlinear relationships of data requires comprehensive consideration in the selection of learning algorithms, the design of feature engineering, and the trade-off between model complexity and generalization ability, so as to ensure that the fusion model can better capture the complex relationships between data.

[0126] In this embodiment, the deep reinforcement learning model adopted is a deep reinforcement learning model based on the DDPG algorithm, taking into account the characteristics of power data.

[0127] In this embodiment, it also includes:

[0128] Model evaluation: Evaluate the trained model;

[0129] The evaluation can include the model's performance on test data, its ability to generalize to new data, etc. The evaluation results can be fed back into the model training process to further optimize the model.

[0130] Deployment and application: When the model has been evaluated and meets the expected performance requirements, it can be deployed in actual applications; the system can use the trained model to align and integrate data, thereby achieving more accurate and efficient data management and analysis.

[0131] like Figure 5 As shown, the embodiment of the present application also provides a multi-domain data fusion system based on multi-source data, including:

[0132] The data acquisition module 10 is used to acquire business data in multiple power fields; and to identify the acquired business data so that each business data has a unique identification; and to pre-process the identified business data;

[0133] The data fusion module 20 is used to construct a knowledge graph based on the preprocessed business data; optimize the knowledge graph based on the deep reinforcement learning model, and fuse multi-source data;

[0134] A data storage module 30, used to store the fused data;

[0135] The data query module 40 is used to receive an input query operation and perform a data query according to the optimized knowledge graph to obtain a query result;

[0136] The data display module 50 is used to display the query results.

[0137] Specifically, the data fusion module 20 includes:

[0138] The knowledge graph module 201 includes: modeling entities of preprocessed operation data; modeling relationships based on relationships between entities; mapping entities and relationships to entities and relationships in the knowledge graph to achieve semantic representation of business data;

[0139] The multi-source data fusion module 202 includes:

[0140] A deep reinforcement learning model 2021, wherein the deep reinforcement learning model comprises: an environment model and an agent, wherein the agent is represented by a policy network; the environment model is composed of a knowledge graph; and the policy network outputs a decision result according to data state information input by the environment model;

[0141] The training module 2022 is used to initialize the parameters of the policy network and use historical data as training samples to train the deep reinforcement learning model. By interacting with the environment model, the parameters of the deep reinforcement learning model are updated to optimize the data alignment and integration effect of the knowledge graph.

[0142] Figure 6 A specific application diagram of the multi-domain data fusion system based on multi-source data provided in the embodiment of the present application is as follows: Figure 6 As shown, the present application can realize the conversion, alignment, fusion and query of multi-source data, solve the problems of semantic differences, structural differences, quality differences and so on among multi-source data, and realize the unified management and interoperability of multi-source data; through cross-domain computing technology, it can realize the statistical analysis of power business data, and support the research and development of data products related to power business, such as power grid operation status monitoring, power quality assessment, power grid fault diagnosis, etc.

[0143] This application adopts a multi-domain data fusion method and system based on multi-source data, and uses multi-source data preprocessing, fusion and query technologies to solve the problems of semantic differences, structural differences, quality differences among multi-source data, and realize unified management and interoperability of multi-source data; the advantages brought include:

[0144] Improved data consistency and quality: Preprocessing and cleaning of multi-source data can reduce errors, redundancy and inconsistency in the data, improve the quality of the data, and make it more suitable for analysis and decision-making.

[0145] Semantic consistency and query convenience: Methods based on knowledge graphs and deep reinforcement learning help resolve semantic differences between multi-source data and achieve semantic consistency. Users can easily obtain required information through a natural language query interface, which improves query convenience.

[0146] Improved computing power of cross-source computing: By building a multi-domain cross-source computing tool for power business data, the "move computing but not data" approach is implemented, avoiding large-scale data transmission, reducing the burden of data storage and transmission, and improving computing efficiency;

[0147] Unified management and interoperability: Through multi-domain data fusion technology, unified management of multi-source data is achieved, so that data from different fields and types can work together. At the same time, data interoperability is improved, so that data from various business fields can better cooperate and promote the integration of business processes.

[0148] Support innovation and decision-making in power business through comprehensive analysis of multi-source data. For example, the development of data products in the areas of power grid operation status monitoring, power quality assessment, and power grid fault diagnosis will help improve the operating efficiency and service level of power business.

[0149] Verification of feasibility and effectiveness: Through practical applications based on power business scenarios, the feasibility and effectiveness of multi-domain data fusion applications in power business are verified. This will help to promote and apply this method in other fields.

[0150] Overall, this approach is beneficial to improving the data management and analysis capabilities of the power business, promoting digital transformation, enhancing the intelligence level of the business, and bringing more innovation and competitive advantages to the power industry.

[0151] The present application also provides an electronic device, including:

[0152] memory; processor; and computer program;

[0153] The computer program is stored in the memory and is configured to be executed by the processor to implement the method as described above.

[0154] An embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored; the computer program is executed by a processor to implement the method described above.

[0155] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods 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.

[0156] 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.

[0157] 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.

[0158] 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 The steps for the functions specified in one or more boxes.

[0159] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0160] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A multi-domain data fusion method based on multi-source data, characterized in that: The following steps are involved: S10, obtaining business data in multiple power fields; S20, performing data identification on the acquired business data so that each business data has a unique identification; S30, pre-processing the marked business data; S40, building a knowledge graph based on the preprocessed business data; S50, optimizes the knowledge graph based on the deep reinforcement learning model and integrates multi-source data; S60, performing data query according to the optimized knowledge graph to obtain query results; The S40, constructing a knowledge graph based on the preprocessed operation data, includes: S401, modeling the entity of the preprocessed operation data; S402, performing relationship modeling according to the relationship between entities; S403, mapping entities and relationships to entities and relationships in the knowledge graph to achieve semantic representation of business data; The S50, based on the deep reinforcement learning model, optimizes the knowledge graph and fuses multi-source data; includes: S501, constructing a deep reinforcement learning model, wherein the deep reinforcement learning model includes: an environment model and an agent, wherein the agent is represented as a policy network; and the environment model is composed of a knowledge graph; The strategy network outputs a decision result according to the data state information input by the environment model; S502, initializing parameters of the policy network; wherein the parameters include: actions and reward values; S503, using historical data as training samples to train the deep reinforcement learning model, and updating the parameters of the deep reinforcement learning model through interaction with the environment model to optimize the data alignment and integration effect of the knowledge graph; The construction of the deep reinforcement learning model includes: S5011, Environmental Modeling, modeling the data integration environment; including: defining data sources, data formats, and data quality assessment standards. Environmental modeling can be used to enhance learning algorithms to help understand the background and requirements of data integration; S5012, state representation: represent the state of the data source; the state includes: features, attributes, and structural information of the data, which can be encoded into vectors or other forms of data representation for processing by the reinforcement learning algorithm; In the state representation stage, the state of the data source needs to be effectively represented so that it can be processed by the reinforcement learning algorithm. The specific steps include: Data feature extraction: Extract useful features from raw data, which can describe the important attributes and characteristics of the data; Data encoding: Encode the extracted features; Data structure representation: Consider the structural relationship between data and represent the data as a data structure suitable for reinforcement learning algorithm processing; Data standardization: Standardize the data to eliminate the scale differences between different data and ensure that the data has a unified dimension; Clarify the relationship between data input and output: Ensure that the relationship between data input and output is clearly reflected in the state representation process; by matching input data and output data, ensure that the state representation can fully reflect the mutual influence and dependency between data; Flexibility of data representation: Ensure the flexibility of state representation to adapt to different types of data and processing scenarios; S5013, Action Definition: Defines the actions that the system can take; S5014, Reward Design: Design a reward function to evaluate the effectiveness of each action of the system.

2. The multi-domain data fusion method based on multi-source data according to claim 1 is characterized in that: Also includes: Establish a query interface to receive data query operations.

3. The multi-domain data fusion method based on multi-source data according to claim 1 is characterized in that: The entity is a noun with a specific meaning or strong referentiality in the operation data, including: a person's name, a place name, an event name, a product name or a professional name; each entity has attributes that describe its characteristics; The S403, mapping entities and relationships to entities and relationships in the knowledge graph to achieve semantic representation of business data, includes: S4031, mapping the identified entity to a node in the knowledge graph, and storing the attribute of the node into the attribute value of the node; S4032, mapping the relationship between entities into an edge connecting two nodes in the knowledge graph, and storing the attributes describing the relationship characteristics into the attribute values ​​of the edge; S4033, realize semantic representation of data based on the nodes and relationships mapped to the knowledge graph.

4. The multi-domain data fusion method based on multi-source data according to claim 1 is characterized in that: The deep reinforcement learning model is a deep reinforcement learning model based on the DDPG algorithm.

5. A multi-domain data fusion system based on multi-source data, characterized in that: include: A data acquisition module (10) is used to acquire business data in multiple power fields; And perform data identification on the acquired business data so that each business data has a unique identification; And pre-processing the marked business data; A data fusion module (20) is used to construct a knowledge graph based on the preprocessed business data; and optimize the knowledge graph based on a deep reinforcement learning model and fuse multi-source data; A data storage module (30), used for storing the fused data; A data query module (40) is used to receive an input query operation and perform a data query based on the optimized knowledge graph to obtain a query result; A data display module (50), used to display query results; The knowledge graph module (201) includes: modeling entities of pre-processed operation data; modeling relationships based on relationships between entities; mapping entities and relationships to entities and relationships in the knowledge graph to achieve semantic representation of business data; The multi-source data fusion module (202) includes: A deep reinforcement learning model (2021), the deep reinforcement learning model comprising: an environment model and an agent, the agent being represented by a policy network; the environment model being composed of a knowledge graph; the policy network outputting a decision result based on data state information input by the environment model; The construction of the deep reinforcement learning model includes: Environmental modeling: modeling the data integration environment, including: defining data sources, data formats, and data quality assessment standards. Environmental modeling can be used to strengthen learning algorithms to help understand the background and requirements of data integration. State representation: Represent the state of the data source; the state includes: data features, attributes, and structural information, which can be encoded into vectors or other forms of data representation for reinforcement learning algorithm processing; in the state representation stage, the state of the data source needs to be effectively represented for reinforcement learning algorithm processing; the specific steps include: Data feature extraction: Extract useful features from raw data, which can describe the important attributes and characteristics of the data; Data encoding: Encode the extracted features; Data structure representation: Consider the structural relationship between data and represent the data as a data structure suitable for reinforcement learning algorithm processing; Data standardization: Standardize the data to eliminate the scale differences between different data and ensure that the data has a unified dimension; Clarify the relationship between data input and output: Ensure that the relationship between data input and output is clearly reflected in the state representation process; by matching input data and output data, ensure that the state representation can fully reflect the mutual influence and dependency between data; Flexibility of data representation: Ensure the flexibility of state representation to adapt to different types of data and processing scenarios; Action definition: defines the actions that the system can take; Reward design: Design a reward function to evaluate the effect of each action of the system; The training module (2022) is used to initialize the parameters of the policy network and use historical data as training samples to train the deep reinforcement learning model. By interacting with the environment model, the parameters of the deep reinforcement learning model are updated to optimize the data alignment and integration effect of the knowledge graph; wherein: the parameters include: action and reward values.

6. An electronic device, characterized in that: include: Memory; processor; as well as Computer programs; The computer program is stored in the memory and is configured to be executed by the processor to implement the method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon; the computer program is executed by a processor to implement the method according to any one of claims 1 to 4.

Images