Power grid integration state rapid intelligent calculation method based on high-dimensional heterogeneous graph data

Abstract

The invention provides a rapid intelligent calculation method for a power grid integration state based on high-dimensional heterogeneous graph data, and belongs to the technical field of power grids. The method comprises the steps of carrying out enterprise-level power grid high-dimensional heterogeneous graph data fusion, carrying out integrated intelligent calculation on a power grid state based on calculation normal form fusion, and accelerating power grid integrated calculation based on high-throughput graph calculation. The method can effectively solve the problem of non-convergence of large-scale load flow calculation, and improves the speed and accuracy of power grid operation state calculation.

Description

Technical Field

[0001] The present invention relates to the field of power grid technology, and in particular to a method for fast intelligent calculation of power grid integration status based on high-dimensional heterogeneous graph data. Background Art

[0002] In graph databases, graph computing is data-centric and adopts a new computing model that closely combines analytical computing with data storage. It has natural advantages in data structure organization and complex association analysis, and has become a disruptive technology that breaks through data management and analytical computing. The graph model naturally supports storage queries of complex association relationships, which can effectively avoid the loading of redundant data in multiple relationship queries and support deep association analysis of entities up to 10 layers or more. In terms of storage and computing mode, the query language of the graph database can seamlessly connect data query and computing, eliminate the time overhead caused by data migration and conversion, and support node parallelism, layered parallelism and other storage and computing integrated computing methods, significantly improving the efficiency of analytical computing. In terms of application ecology, graph computing supports the relative separation of data source layer, data management and data application layer, and promotes the transformation of software development model from application-centric to data-centric, from isolated and closed to open and shared, and has achieved significant success in many fields such as the Internet.

[0003] Large-scale distributed graph computing can be divided into three major factions: MapReduc faction, Pregel faction and GraphLab faction. Most of the existing large-scale graph computing systems are global and batch processing systems, and are oriented to static graph structures. They have good scalability, high fault tolerance and flexibility for processing massive graph data. In terms of system operation mode, synchronous control systems have high accuracy of operation results, while restricted asynchronous systems have fast execution speed. At present, graph computing systems have a large development space for different computing models, key technologies and research fields. The advantages of graph computing technology have been fully demonstrated in social networks, finance, smart navigation, biopharmaceuticals and other industries.

[0004] At the same time, in the field of parallel computing, there are more and more parallel computing hardware and many parallel computing optimization technologies and algorithms, such as load balancing algorithms, communication optimization technologies, fault-tolerant technologies, etc.

[0005] However, at the same time, there is a problem of non-convergence of large-scale power flow calculations in the existing power grid operation status calculations, which makes it difficult to break through the bottleneck of speed and efficiency of power grid integration calculation deduction, and it is impossible to perform high-speed and accurate power grid operation status calculations. Summary of the invention

[0006] The purpose of the present invention is to provide a fast intelligent calculation method for the integrated state of a power grid based on high-dimensional heterogeneous graph data, which can effectively solve the problem of non-convergence of large-scale power flow calculations and improve the speed and accuracy of power grid operation state calculations.

[0007] In order to achieve the above object, the present invention provides a method for fast intelligent calculation of power grid integration status based on high-dimensional heterogeneous graph data, the method comprising: Carry out enterprise-level power grid high-dimensional heterogeneous graph data fusion; Based on the integrated intelligent computing of computing paradigm, the grid status is calculated; Accelerating grid integration computing based on high-throughput graph computing.

[0008] Preferably, performing enterprise-level power grid high-dimensional heterogeneous graph data fusion includes: Based on the public information model of the power grid, the ontology unified modeling paradigm is constructed based on the ontology modeling method of the multi-source relational power grid model, and the physical equipment and abstract logical data model of the power grid that integrates multiple business fields is established. Through attribute classification, device node definition, abstract node aggregation and node optimization, the automatic construction of the model from relational database to graph database is studied; Preprocessing to integrate and coordinate image data from different sources, modalities, or spatial-temporal resolutions; Based on the matching alignment and graph data fusion of high-dimensional heterogeneous graphs in enterprise-level power grids, an integrated spatiotemporal data platform for the entire graph status of the generation, transmission, transformation and distribution network is constructed. A holographic model of the power system for generation, transmission and distribution is constructed with holographic equipment components covering equipment asset attributes, geographic location attributes, and electrical parameter attributes as nodes, and topological connection relationships, subordinate relationships, combination relationships, and management relationships between equipment components as edges.

[0009] Preferably, the attribute classification in the enterprise-level power grid high-dimensional heterogeneous graph data fusion is to classify the nodes or edges in the graph structure based on their attributes, including machine learning classification and graph neural network classification, wherein: Machine learning classification uses the attribute information of nodes and edges and applies supervised learning models to classify nodes and edges; Graph neural network GNN uses graph structure and its attribute information to classify nodes, and identifies nodes of different categories in the graph by propagating the attribute information of nodes.

[0010] Preferably, subsequent analysis tasks include node classification, edge classification and community detection, and supervised learning models include decision trees, support vector machines and logistic regression.

[0011] Preferably, the device node definition in the enterprise-level power grid high-dimensional heterogeneous graph data fusion is used to model various devices in the Internet of Things as nodes in the Internet of Things IoT scenario, and define the characteristics and attributes of the nodes; wherein, Devices are classified and defined according to their type, purpose, and communication protocol to determine which device nodes can interact with data. At the same time, useful feature information is extracted for each device node and used as node attributes for subsequent analysis. Device nodes and their relationships are stored and managed through a graph database.

[0012] Preferably, the abstract node aggregation in the enterprise-level power grid high-dimensional heterogeneous graph data fusion is performed by merging multiple similar nodes or nodes with the same attributes into one abstract node, including: Attribute-based aggregation, which aggregates nodes based on their similar attributes; Community structure-based aggregation uses a community detection algorithm to identify the community structure in the graph and treats each community as an abstract node; Graph convolution aggregation, in the graph neural network GNN, the information of neighbor nodes is aggregated to the target node through convolution operations to learn the abstract representation of the node.

[0013] Preferably, the node optimization in the enterprise-level power grid high-dimensional heterogeneous graph data fusion is performed by improving the node representation in the graph or adjusting the relationship between nodes through an algorithm or strategy, including: Using graph embedding optimization algorithms, we project nodes in the graph into a low-dimensional space by learning the embedding representation of nodes. Optimization problem modeling: modeling some graph optimization tasks as optimization problems and using heuristic algorithms to optimize the relationships between nodes; Node importance analysis, optimizes the graph structure by analyzing the importance of nodes.

[0014] Preferably, the integrated intelligent computing grid status based on computing paradigm fusion includes: Based on the high-dimensional mixed graph data of the power grid, four types of storage-computing integrated meta-computing functions are developed: A self-calculated function that only considers the characteristics of the node itself, without considering the characteristics of its surrounding nodes; A first-order neighbor calculation function that only considers the feature calculation of nodes directly adjacent to the node; A second-order neighbor calculation function that considers not only the nodes directly adjacent to a node, but also the neighboring nodes of the node's neighbors for feature calculation; and a subgraph calculation function that calculates features of a subgraph consisting of a node and its adjacent nodes as a whole; Based on the meta-computing function, the graph node parallel computing mechanism and graph layered distributed computing are studied, a graph computing acceleration function module library for grid integration computing applications is developed, and a power-specific graph computing module is constructed. The part related to matrix LU decomposition is solved by graph layered parallel computing, and the part unrelated to matrix decomposition is solved by graph node distributed parallel computing; the distributed parallel solution steps are as follows: S1, local computing, an independent computing process using computing logic; S2, message communication, after the local calculation is completed, each computing node exchanges its own data information through message transmission; S3, roadblock synchronization, ensures that each computing unit completes the calculation of each unit in the current superstep before the next superstep starts, thus achieving synchronization between blocks; Sort out different computing paradigms that can be used for joint computing of main and distribution systems, and develop integrated intelligent computing of power grid status based on computing paradigm fusion, where computing paradigm fusion includes the fusion of graph computing and graph learning as well as the fusion of distributed computing and parallel computing.

[0015] Preferably, the acceleration of grid integration calculation based on high-throughput graph calculation includes: Expert knowledge is organized into constraints of the graph neural network and added to the model loss function. The interpretability of graph machine learning is improved by integrating the graph machine learning function library that combines traditional model analysis with data-driven. The power graph learning fusion calculation is realized through reinforcement learning and spatiotemporal graph convolutional network, and the accelerated calculation of power graph flow is realized. Based on the massive distributed resources on the user side, a power grid subgraph containing distributed resources is established according to their access information; based on the graph database, a distributed resource access association relationship including access points, aggregation points and topology is established, and a characteristic logical relationship is established; based on the rapid aggregation of ultra-large-scale distributed resources, the search and parallel algorithms of graph computing are used to parallel aggregate ultra-large-scale distributed resources in a power grid model with tens of thousands of nodes, and the spatiotemporal dynamic adjustment of the aggregate is completed based on the association relationship analysis.

[0016] According to the above technical scheme, the present invention constructs an enterprise-level high-dimensional heterogeneous graph of the power grid with multi-source data fusion, and designs a method for rapid and accurate state estimation based on computing paradigm fusion and high-throughput graph computing. At the same time, an integrated intelligent computing component of the power grid state is formed, and a method for rapid intelligent computing of the integrated power grid state based on high-dimensional heterogeneous graph data is designed. By introducing a graph database and a graph computing model, significant advantages are produced in terms of unified knowledge expression of power system data, "storage and computing in one" analysis, deep correlation mining, expansion of data processing scale, and rapid deduction of power grid state. It can effectively solve the problem of non-convergence of large-scale power flow calculations, break through the speed and efficiency bottleneck of integrated power grid computing deduction, and support high-speed and accurate calculation of power grid operation status.

[0017] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings: Figure 1 It is a flow chart of a method for fast intelligent calculation of power grid integration status based on high-dimensional heterogeneous graph data provided by the present invention. DETAILED DESCRIPTION

[0019] The specific implementation of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.

[0020] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application are in compliance with the relevant provisions of national laws and regulations. In the embodiments of this application, some existing solutions in the industry such as certain software, components, and models may be mentioned, which should be considered as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of this application, but it does not mean that the applicant has or will necessarily use the solution.

[0021] See also Figure 1 The present invention provides a method for fast intelligent calculation of the integrated state of a power grid based on high-dimensional heterogeneous graph data. The method includes fusion of enterprise-level high-dimensional heterogeneous graph data of the power grid, integrated intelligent calculation of the power grid state based on computing paradigm fusion, and acceleration of the integrated power grid calculation based on high-throughput graph computing.

[0022] In this embodiment, the enterprise-level power grid high-dimensional heterogeneous graph data fusion includes: First, based on the power grid public information model and the ontology modeling method of the multi-source relational power grid model, an ontology unified modeling paradigm is constructed to establish a power grid physical equipment and abstract logical data model that integrates multiple business fields. Through attribute classification, equipment node definition, abstract node aggregation, node optimization and other methods, the automatic construction technology of the pattern from relational database to graph database is studied.

[0023] Secondly, in order to address the problems of diverse data structures of equipment entities between different business systems, inconsistent sets of common attributes, inconsistent values ​​of the same attribute of the same entity in different systems, and difficulty in entity matching and data fusion, a preprocessing method is developed to integrate and coordinate graph data from different sources, different modalities, or different spatial-temporal resolutions. The preprocessing of multi-source heterogeneous data includes steps such as dimensional unification, data cleaning, data normalization, data sampling, and data interpolation to improve data quality and consistency and reduce data noise and redundancy.

[0024] Finally, based on the enterprise-level high-dimensional heterogeneous graph matching alignment and graph data fusion technology, an integrated spatiotemporal data platform for the entire graph status of the generation, transmission, transformation and distribution network is constructed, and a holographic model of the power system for generation, transmission and distribution is constructed with holographic equipment components covering equipment asset attributes, geographic location attributes, and electrical parameter attributes as nodes, and topological connection relationships, subordinate relationships, combination relationships, and management relationships between equipment components as edges.

[0025] Specifically, the above attribute classification refers to classifying nodes or edges based on their attributes in a graph structure, thereby providing a basis for subsequent analysis tasks (such as node classification, edge classification, community detection, etc.). Common attribute classification methods include machine learning classification and graph neural networks. Machine learning classification uses the attribute information of nodes and edges and applies supervised learning models (such as decision trees, support vector machines, and logistic regression) to classify nodes and edges. For example, in a knowledge graph, the contextual features and attributes of nodes are used to classify entities. Graph neural networks (GNNs) can effectively use graph structures and their attribute information to classify nodes. By propagating the attribute information of nodes, GNNs can identify nodes of different categories in a graph.

[0026] Device node definition is used in the Internet of Things (IoT) scenario. It refers to modeling various devices in the IoT (such as sensors, smart devices) as nodes and defining the characteristics and properties of these nodes. Device node definition helps to establish a relationship network between devices for better data analysis, fault detection, collaborative optimization between devices, etc. Classify and define devices according to their type, purpose, communication protocol, etc., and determine which device nodes can interact with data. Extract useful feature information for each device node, including the operating status of the device, the value read by the sensor, etc., which can be used as node attributes for subsequent analysis. Device nodes and their relationships are stored and managed through graph databases (such as Neo4j and JanusGraph) to support real-time analysis and query.

[0027] Abstract node aggregation refers to reducing the complexity of the graph by merging multiple similar nodes or nodes with the same attributes into one abstract node. This method can significantly reduce the computational cost and improve the representation ability of the graph when processing large-scale graph data. The node aggregation methods are: 1. Attribute-based aggregation: nodes are aggregated according to their similar attributes; for example, in a social network, users with similar interests can be aggregated into one node.

[0028] 2. Aggregation based on community structure, using community detection algorithms (such as Louvain algorithm, spectral clustering) to identify the community structure in the graph and treat each community as an abstract node; 3. Graph Convolutional Aggregation: In the graph neural network GNN, the information of neighbor nodes is aggregated to the target node through convolution operations to learn the abstract representation of the node.

[0029] In this embodiment, node optimization refers to improving the node representation in the graph or adjusting the relationship between nodes through algorithms or strategies to improve the performance and accuracy of graph analysis tasks. Node optimization can be used to enhance the embedded representation of nodes, improve the connectivity between nodes, reduce unnecessary node noise, etc. In the process of node embedding, by optimizing the feature representation of nodes, the nodes in the graph can better represent their context information in a low-dimensional space. The methods of node optimization are: Graph embedding optimization algorithms: such as Node2Vec, DeepWalk, GraphSAGE, etc. These algorithms project the nodes in the graph into a low-dimensional space by learning the embedding representation of the nodes for downstream tasks (such as classification and clustering).

[0030] Optimization problem modeling: Some graph optimization tasks can be modeled as optimization problems, using heuristic algorithms (such as simulated annealing, genetic algorithms) to optimize the relationship between nodes.

[0031] Node importance analysis: By analyzing the importance of nodes (such as using PageRank or node centrality indicators), the structure of the graph can be optimized and the analysis efficiency can be improved.

[0032] In addition, in the integrated intelligent computing of power grid status based on computing paradigm fusion, firstly, four types of storage-computing-integrated meta-computing functions are developed based on the high-dimensional mixed graph data of the power grid, including: SelfCalculation function that only considers the characteristics of the node itself, not the characteristics of its surrounding nodes; The first-order neighbor calculation (FirstOrderNeighborhood) function only considers the feature calculation of the nodes directly adjacent to the node; The second order neighbor calculation function considers not only the nodes directly adjacent to a node, but also the neighboring nodes of the node's neighbors. And the SubgraphCalculation function that calculates the features of the subgraph composed of a node and its adjacent nodes as a whole; Secondly, based on the meta-computing function, the graph node parallel computing mechanism and graph hierarchical distributed computing are studied, and a graph computing acceleration function module library for grid integration computing applications is developed. A power-specific graph computing module is constructed, including Jacobian matrix formation, gain matrix formation, node active injection power calculation, node reactive injection power calculation, branch active power flow calculation, branch reactive power flow calculation, node voltage amplitude correction, and node voltage phase angle correction. The part related to the matrix LU decomposition is solved by graph hierarchical parallel computing, and the part unrelated to the matrix decomposition is solved by graph node distributed parallel computing. Among them, the distributed parallel solution steps are as follows: S1, local computing, an independent computing process using computing logic; S2, message communication, after the local calculation is completed, each computing node exchanges its own data information through message transmission; S3, roadblock synchronization, ensures that each computing unit completes the calculation of each unit in the current superstep before the next superstep starts, thus achieving synchronization between blocks; Finally, we sorted out different computing paradigms that can be used for joint computing of main and distribution systems, including graph computing, graph neural networks, and graph distributed computing, and developed integrated intelligent computing of power grid status based on the fusion of computing paradigms. The fusion of computing paradigms includes the fusion of graph computing and graph learning, as well as the fusion of distributed computing and parallel computing.

[0033] In this embodiment, the acceleration of grid integration calculation based on high-throughput graph calculation includes: First, based on the application characteristics of power grid state estimation and power flow calculation, expert knowledge is organized into constraints of graph neural networks and added to the model loss function. The interpretability of graph machine learning is improved by integrating the graph machine learning function library that combines traditional model analysis with data-driven. Based on the interactive fusion of power graph calculation and graph learning, the fusion calculation of power graph learning is realized through reinforcement learning and spatiotemporal graph convolutional networks, which realizes the accelerated calculation of power graph power flow and provides an accurate power flow calculation result for the subsequent analytical method with initial value dependence. Secondly, based on the massive distributed resources on the user side, including distributed new energy and flexible loads, a power grid subgraph containing distributed resources is established according to their access information; based on the graph database, a distributed resource access association relationship including access points, aggregation points and topology is established, and characteristic logical relationships are established, including distributed resource response and feasible domain; based on the rapid aggregation of ultra-large-scale distributed resources, the search and parallel algorithms of graph computing are used to parallel aggregate ultra-large-scale distributed resources in the tens of thousands of nodes power grid model, and the spatiotemporal dynamic adjustment of the aggregate is completed based on the association relationship analysis.

[0034] Through the above technical solutions, a high-dimensional heterogeneous graph of enterprise-level power grids with multi-source data fusion was constructed, and a fast and accurate state estimation method based on computing paradigm fusion and high-throughput graph computing was designed. At the same time, an integrated intelligent computing component for power grid state was formed, and a fast intelligent computing method for integrated power grid state based on high-dimensional heterogeneous graph data was designed. By introducing graph databases and graph computing models, significant advantages are achieved in the unified expression of knowledge of power system data, "storage and computing in one" analysis, deep correlation mining, expansion of data processing scale, and rapid deduction of power grid state. It can effectively solve the problem of non-convergence of large-scale power flow calculations, break through the speed and efficiency bottleneck of integrated power grid computing deduction, and support high-speed and accurate calculation of power grid operation status.

[0035] 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 combining 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.) containing computer-usable program codes.

[0036] 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.

[0037] 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.

[0038] 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.

[0039] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0040] The memory may include non-permanent memory 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.

[0041] 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 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.

[0042] 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.

[0043] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A fast intelligent calculation method for power grid integration status based on high-dimensional heterogeneous graph data, characterized in that: The method comprises: Carry out enterprise-level power grid high-dimensional heterogeneous graph data fusion; Based on the integrated intelligent computing of computing paradigm, the grid status is calculated; Accelerating grid integration computing based on high-throughput graph computing.

2. The method for fast intelligent calculation of power grid integration status based on high-dimensional heterogeneous graph data according to claim 1 is characterized in that: The enterprise-level power grid high-dimensional heterogeneous graph data fusion includes: Based on the public information model of the power grid, the ontology unified modeling paradigm is constructed based on the ontology modeling method of the multi-source relational power grid model, and the physical equipment and abstract logical data model of the power grid that integrates multiple business fields is established. Through attribute classification, device node definition, abstract node aggregation and node optimization, the automatic construction of the model from relational database to graph database is studied; Preprocessing to integrate and coordinate image data from different sources, modalities, or spatial-temporal resolutions; Based on the matching alignment and graph data fusion of high-dimensional heterogeneous graphs in enterprise-level power grids, an integrated spatiotemporal data platform for the entire graph status of the generation, transmission, transformation and distribution network is constructed. A holographic model of the power system for generation, transmission and distribution is constructed with holographic equipment components covering equipment asset attributes, geographic location attributes, and electrical parameter attributes as nodes, and topological connection relationships, subordinate relationships, combination relationships, and management relationships between equipment components as edges.

3. The method for fast intelligent calculation of power grid integration status based on high-dimensional heterogeneous graph data according to claim 2 is characterized in that: The attribute classification in the enterprise-level power grid high-dimensional heterogeneous graph data fusion is to classify the nodes or edges in the graph structure based on their attributes, including machine learning classification and graph neural network classification, where: Machine learning classification involves using the attribute information of nodes and edges and applying supervised learning models to classify nodes and edges; Graph neural network GNN involves using graph structure and its attribute information to classify nodes, and identifying different categories of nodes in the graph by propagating the attribute information of the nodes.

4. The method for fast intelligent calculation of power grid integration status based on high-dimensional heterogeneous graph data according to claim 3 is characterized in that: Subsequent analysis tasks include node classification, edge classification, and community detection, and supervised learning models include decision trees, support vector machines, and logistic regression.

5. The method for fast intelligent calculation of power grid integration status based on high-dimensional heterogeneous graph data according to claim 2 is characterized in that: The device node definition in the enterprise-level power grid high-dimensional heterogeneous graph data fusion is used to model various devices in the Internet of Things as nodes in the Internet of Things (IoT) scenario and define the characteristics and attributes of the nodes; among them, Devices are classified and defined according to their type, purpose, and communication protocol to determine which device nodes can interact with data. At the same time, useful feature information is extracted for each device node and used as node attributes for subsequent analysis. Device nodes and their relationships are stored and managed through a graph database.

6. The method for fast intelligent calculation of power grid integration status based on high-dimensional heterogeneous graph data according to claim 2 is characterized in that: Abstract node aggregation in enterprise-level power grid high-dimensional heterogeneous graph data fusion is performed by merging multiple similar nodes or nodes with the same attributes into one abstract node, including: Attribute-based aggregation, which aggregates nodes based on their similar attributes; Community structure-based aggregation uses a community detection algorithm to identify the community structure in the graph and treats each community as an abstract node; Graph convolution aggregation, in the graph neural network GNN, the information of neighbor nodes is aggregated to the target node through convolution operations to learn the abstract representation of the node.

7. The method for fast intelligent calculation of power grid integration status based on high-dimensional heterogeneous graph data according to claim 2 is characterized in that: Optimize nodes in enterprise-level power grid high-dimensional heterogeneous graph data fusion to improve the node representation in the graph or adjust the relationship between nodes through algorithms or strategies, including: Using graph embedding optimization algorithms, we project nodes in the graph into a low-dimensional space by learning the embedding representation of nodes. Optimization problem modeling: modeling some graph optimization tasks as optimization problems and using heuristic algorithms to optimize the relationships between nodes; Node importance analysis, optimizes the graph structure by analyzing the importance of nodes.

8. The method for fast intelligent calculation of power grid integration status based on high-dimensional heterogeneous graph data according to claim 2 is characterized in that: Based on the integrated intelligent computing paradigm, the grid status includes: Based on the high-dimensional mixed graph data of the power grid, four types of storage-computing integrated meta-computing functions are developed: A self-calculated function that only considers the characteristics of the node itself, without considering the characteristics of its surrounding nodes; A first-order neighbor calculation function that only considers the feature calculation of nodes directly adjacent to the node; A second-order neighbor calculation function that considers not only the nodes directly adjacent to a node, but also the neighboring nodes of the node's neighbors for feature calculation; and a subgraph calculation function that calculates features of a subgraph consisting of a node and its adjacent nodes as a whole; Based on the meta-computing function, the graph node parallel computing mechanism and graph layered distributed computing are studied, a graph computing acceleration function module library for grid integration computing applications is developed, and a power-specific graph computing module is constructed. The part related to matrix LU decomposition is solved by graph layered parallel computing, and the part unrelated to matrix decomposition is solved by graph node distributed parallel computing; the distributed parallel solution steps are as follows: S1, local computing, an independent computing process using computing logic; S2, message communication, after the local calculation is completed, each computing node exchanges its own data information through message transmission; S3, roadblock synchronization, ensures that each computing unit completes the calculation of each unit in the current superstep before the next superstep starts, thus achieving synchronization between blocks; Sort out different computing paradigms that can be used for joint computing of main and distribution systems, and develop integrated intelligent computing of power grid status based on computing paradigm fusion, where computing paradigm fusion includes the fusion of graph computing and graph learning as well as the fusion of distributed computing and parallel computing.

9. The method for fast intelligent calculation of power grid integration status based on high-dimensional heterogeneous graph data according to claim 1, characterized in that: Acceleration of grid integration computing based on high-throughput graph computing includes: Expert knowledge is organized into constraints of the graph neural network and added to the model loss function. The interpretability of graph machine learning is improved by integrating the graph machine learning function library that combines traditional model analysis with data-driven. The power graph learning fusion calculation is realized through reinforcement learning and spatiotemporal graph convolutional network, and the accelerated calculation of power graph flow is realized. Based on the massive distributed resources on the user side, a power grid subgraph containing distributed resources is established according to their access information; based on the graph database, a distributed resource access association relationship including access points, aggregation points and topology is established, and a characteristic logical relationship is established; based on the rapid aggregation of ultra-large-scale distributed resources, the search and parallel algorithms of graph computing are used to parallel aggregate ultra-large-scale distributed resources in a power grid model with tens of thousands of nodes, and the spatiotemporal dynamic adjustment of the aggregate is completed based on the association relationship analysis.

Images