Method and apparatus for constructing high-order tensor network of large-scale power grid, and device and medium

Abstract

The present disclosure relates to the technical field of smart power grids. Disclosed are a method and apparatus for constructing a high-order tensor network of a large-scale power grid, and a device and a medium. The method comprises: acquiring multiple category attribute sets of heterogeneous nodes in a large-scale power grid; performing feature extraction on the multiple category attribute sets, so as to obtain attribute features of the heterogeneous nodes; on the basis of the attribute features of the heterogeneous nodes, using a deep hash mapping model to establish multiple feature sub-spaces; using a breadth learning strategy to align the multiple feature sub-spaces to a unified feature dimension; on the basis of an attribute feature vector of the unified feature dimension, using a distance measurement method to calculate a distance measurement value between the heterogeneous nodes; on the basis of the attribute feature vector of the unified feature dimension and the weight of a heterogeneous node interaction relationship, determining a basis tensor for representing the heterogeneous node interaction relationship; and on the basis of the basis tensor and the weight of the heterogeneous node interaction relationship, using a tensor multiplication operation rule to construct a high-order tensor network of a large-scale power grid.

Description

Method, device, equipment and medium for constructing high-order tensor network of large-scale power grid

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This disclosure is based on Chinese patent application number 202410081021.4, application date January 19, 2024, and invention name “Method, device, equipment and medium for constructing high-order tensor network of large-scale power grid”, and claims the priority of the Chinese patent application. The entire content of the Chinese patent application is hereby introduced into the present disclosure as a reference. Technical Field

[0003] The present disclosure relates to the field of smart grid technology, and in particular to a method, device, equipment, and medium for constructing a large-scale power grid high-order tensor network. Background Art

[0004] With the access of large-scale distributed resources, the power grid has gradually developed into a high-order network with complex entities, rapid time-varying, and heterogeneous interactions. The power grid data is characterized by complex and diverse types, different representation forms, heterogeneous node attributes, and frequent node interactions. Large-scale power grids are complex digital coupling spaces of high order, high dimension, heterogeneous, and incomplete. They cover power grids at all voltage levels, including generation, transmission, distribution, and consumption. They are composed of power plants, substations, primary and secondary equipment at all voltage levels, and their connecting networks. The computing node types cover various entities such as power equipment, power loads, and power users. Due to the large number of physical nodes in large-scale power grids, their different types, the difficulty of data fusion, and the large computational scale, current model construction methods for large-scale power grids have difficulty in efficiently representing large-scale power grids. Problems such as slow analysis and calculation speed and high computing power consumption hinder the development of power grids. Therefore, the related technologies lack models that can efficiently represent large-scale power grids.

[0005] Summary of the Invention

[0006] In view of this, the present disclosure provides a method, device, equipment and medium for constructing a large-scale power grid high-order tensor network to solve the problem of lack of efficient characterization of large-scale power grid models in related technologies.

[0007] In a first aspect, the present disclosure provides a method for constructing a large-scale power grid high-order tensor network, including: obtaining a plurality of category attribute sets corresponding to heterogeneous nodes in a large-scale power grid; performing feature extraction on the plurality of category attribute sets to obtain heterogeneous node attribute features; using a deep hash mapping model to establish multiple feature subspaces based on the heterogeneous node attribute features, and any feature subspace corresponds to a type of heterogeneous node interaction relationship; using a breadth learning strategy to align the multiple feature subspaces to a unified feature dimension to obtain an attribute feature vector of the unified feature dimension; using a distance measurement method to calculate the distance measurement value between heterogeneous nodes based on the attribute feature vector of the unified feature dimension, and the distance measurement value is used to determine the weight of the heterogeneous node interaction relationship; determining a basis tensor for representing the heterogeneous node interaction relationship based on the attribute feature vector of the unified feature dimension and the weight of the heterogeneous node interaction relationship; and using a tensor multiplication operation rule to construct a large-scale power grid high-order tensor network based on the basis tensor and the weight of the heterogeneous node interaction relationship.

[0008] In the disclosed embodiment, a tensor network is introduced to characterize a large-scale power grid. A deep hash mapping model is used to establish multiple feature subspaces based on the extracted heterogeneous node attribute features to represent the interaction relationship between heterogeneous nodes. A distance metric method is used to calculate the weight of the interaction relationship between heterogeneous nodes based on the aligned multiple feature subspaces. Then, a base tensor representing the interaction relationship is determined, and the base tensor is used to construct a high-order tensor network to characterize the large-scale power grid. Because the tensor network has the distributed storage and parallel processing capabilities for high-dimensional heterogeneous features and complex associations, the use of the tensor network to model the large-scale power grid achieves the effective integration of high-dimensional and complex features, thereby improving the scalability of the large-scale power grid model and reducing the computational complexity, solving the problem of the lack of an efficient model for characterizing large-scale power grids in the related art.

[0009] In an optional embodiment, the method also includes: acquiring heterogeneous power grid data; compressing a large-scale power grid high-order tensor network based on a high-order singular value analysis framework, bit planes, run-length coding, and arithmetic coding; constructing a nonlinear representation model of the spatiotemporal mechanism corresponding to the heterogeneous power grid data; and generating a large-scale power grid high-order tensor network containing a dynamic evolution mechanism based on the nonlinear representation model of the spatiotemporal mechanism and the compressed large-scale power grid high-order tensor network.

[0010] In the embodiments of the present disclosure, by constructing a nonlinear representation model of the spatiotemporal mechanism, accurate modeling of the dynamic evolution mechanism of a large-scale power grid is achieved, thereby achieving the purpose of further improving the high-order tensor network characterization capability of a large-scale power grid.

[0011] In an optional embodiment, feature extraction is performed on multiple category attribute sets to obtain heterogeneous node attribute features, including: extracting features corresponding to multiple category attribute sets; storing the features corresponding to any extracted category attribute set as a second-order tensor; determining multiple second-order tensors corresponding to the multiple category attribute sets; and using the multiple second-order tensors as heterogeneous node attribute features.

[0012] In the embodiment of the present disclosure, a second-order tensor is used to store a set of multiple categories of attributes of heterogeneous nodes, thereby achieving the purpose of merging the attributes of heterogeneous nodes in a large-scale power grid and realizing efficient representation of the attribute characteristics of heterogeneous nodes.

[0013] In an optional embodiment, the deep hash mapping model includes a common hash coding module and multiple hash coding sub-modules. The deep hash mapping model is used to establish multiple feature subspaces based on the attribute characteristics of heterogeneous nodes. Any feature subspace corresponds to a type of heterogeneous node interaction relationship, including: using the common hash coding module to map the attribute characteristics of heterogeneous nodes to the common attribute feature space; using each hash coding sub-module to establish a feature subspace based on the common attribute feature space.

[0014] In the embodiment of the present disclosure, deep hash mapping is used to establish multiple feature subspaces, which achieves the alignment and separation of the attribute space of heterogeneous nodes in a large-scale power grid, thereby using feature subspaces to characterize the interactive relationship of a class of heterogeneous nodes.

[0015] In an optional embodiment, the attribute feature vector of the unified feature dimension includes the attribute feature vector of the unified feature dimension corresponding to the heterogeneous node and its neighboring nodes, and a distance measurement method is used to calculate the distance measurement value between the heterogeneous nodes based on the attribute feature vector of the unified feature dimension. The distance measurement value is used to determine the weight of the interaction relationship between the heterogeneous nodes, including: using a distance measurement method to calculate the distance measurement value between the heterogeneous nodes based on the attribute feature vector of the unified feature dimension; using graph convolution to merge the attribute feature vector of the unified feature dimension corresponding to the heterogeneous node and its neighboring nodes based on the distance measurement value to obtain a new heterogeneous node attribute feature vector; and determining the weight of the heterogeneous node interaction relationship based on the new heterogeneous node attribute feature vector.

[0016] In the embodiment of the present disclosure, the purpose of unifying the heterogeneous nodes and the heterogeneous node interaction relationships is achieved by determining the weights of the heterogeneous node interaction relationships.

[0017] In an optional embodiment, a nonlinear representation model of the spatiotemporal mechanism corresponding to heterogeneous power grid data is constructed, including: obtaining multi-factor spatiotemporal neighborhood characteristics; using a regularization method to jointly model the spatiotemporal neighborhood information of the heterogeneous power grid data according to the multi-factor spatiotemporal neighborhood characteristics to obtain a power grid topology map in the spatiotemporal dimension; using a nonlinear Kalman filter state transition estimation method to analyze the power grid topology map in the spatiotemporal dimension to determine the state transition mode of the power grid topology map; and constructing a nonlinear representation model of the spatiotemporal mechanism corresponding to the heterogeneous power grid data based on the state transition mode of the power grid topology map.

[0018] In the disclosed embodiment, a nonlinear representation model of the spatiotemporal mechanism corresponding to heterogeneous power grid data is constructed based on the multi-factor spatiotemporal neighborhood characteristics, thereby achieving the purpose of analyzing the potential patterns of large-scale power grid time series and integrating spatiotemporal characteristics into large-scale power grid high-order tensor networks.

[0019] In an optional embodiment, a regularization method is used to jointly model the spatiotemporal neighborhood information of heterogeneous power grid data based on multi-factor spatiotemporal neighborhood characteristics to obtain a power grid topology map in the spatiotemporal dimension, including: constructing a corresponding undirected weighted graph based on the heterogeneous power grid data; using the Laplace operator and time difference to generate a single spatiotemporal neighborhood information based on the undirected weighted graph; using a multi-index joint constraint regularization method to jointly model the spatiotemporal neighborhood information of the heterogeneous power grid data based on the single spatiotemporal neighborhood information to obtain a power grid topology map in the spatiotemporal dimension.

[0020] In the embodiment of the present disclosure, by generating single spatiotemporal neighborhood information and then adopting a multi-index joint constraint regularization method for joint modeling, the problem of strong and weak differences in single time or space characteristics is solved, and the high-order tensor network characterization capability of large-scale power grids is further improved.

[0021] In the second aspect, the present disclosure provides a large-scale power grid high-order tensor network construction device, including: an attribute set acquisition module, configured to obtain a plurality of category attribute sets corresponding to heterogeneous nodes in a large-scale power grid; a feature extraction module, configured to extract features from a plurality of category attribute sets to obtain heterogeneous node attribute features; a multiple feature subspace establishment module, configured to use a deep hash mapping model to establish multiple feature subspaces according to the heterogeneous node attribute features, and any feature subspace corresponds to a type of heterogeneous node interaction relationship; a feature alignment module, configured to use a breadth learning strategy to align multiple feature subspaces to a unified feature dimension, and obtain the attribute feature vector of the unified feature dimension; a distance metric value calculation module is configured to use a distance metric method to calculate the distance metric value between heterogeneous nodes according to the attribute feature vector of the unified feature dimension, and the distance metric value is used to determine the weight of the interaction relationship between heterogeneous nodes; a base tensor determination module is configured to determine the base tensor used to represent the interaction relationship between heterogeneous nodes according to the attribute feature vector of the unified feature dimension and the weight of the interaction relationship between heterogeneous nodes; a power grid high-order tensor network construction module is configured to use a tensor multiplication operation rule to construct a large-scale power grid high-order tensor network according to the base tensor and the weight of the interaction relationship between heterogeneous nodes.

[0022] In an optional embodiment, the device also includes: a power grid data acquisition module, configured to acquire heterogeneous power grid data; a compression module, configured to compress large-scale power grid high-order tensor networks based on a high-order singular value analysis framework, bit planes, run-length coding, and arithmetic coding; a mechanism model construction module, configured to construct a spatiotemporal mechanism nonlinear representation model corresponding to the heterogeneous power grid data; and a dynamic network generation module, configured to generate a large-scale power grid high-order tensor network containing a dynamic evolution mechanism based on the spatiotemporal mechanism nonlinear representation model and the compressed large-scale power grid high-order tensor network.

[0023] In an optional embodiment, the feature extraction module includes: a feature extraction unit, configured to extract features corresponding to multiple category attribute sets; a feature storage unit, configured to store the features corresponding to any extracted category attribute set as a second-order tensor; a tensor determination unit, configured to determine multiple second-order tensors corresponding to multiple category attribute sets; and a node attribute feature generation unit, configured to use multiple second-order tensors as heterogeneous node attribute features.

[0024] In an optional embodiment, the deep hash mapping model includes a common hash coding module and multiple hash coding sub-modules, and the multiple feature subspace establishment module includes: a feature mapping unit, configured to use the common hash coding module to map heterogeneous node attribute features to a common attribute feature space; a subspace establishment unit, configured to use each hash coding sub-module to establish a feature subspace based on the common attribute feature space.

[0025] In an optional embodiment, the attribute feature vector of the unified feature dimension includes the attribute feature vector of the unified feature dimension corresponding to the heterogeneous node and its neighboring nodes, and the distance measurement value calculation module includes: a calculation unit, configured to use a distance measurement method to calculate the distance measurement value between the heterogeneous nodes based on the attribute feature vector of the unified feature dimension; a merging unit, configured to use graph convolution to merge the attribute feature vector of the unified feature dimension corresponding to the heterogeneous node and its neighboring nodes based on the distance measurement value to obtain a new heterogeneous node attribute feature vector; a determination unit, configured to determine the weight of the heterogeneous node interaction relationship based on the new heterogeneous node attribute feature vector.

[0026] In an optional embodiment, the mechanism model construction module includes: a spatiotemporal feature acquisition unit, configured to acquire multi-factor spatiotemporal neighborhood features; a joint modeling unit, configured to use a regularization method to perform joint modeling of spatiotemporal neighborhood information of heterogeneous power grid data according to multi-factor spatiotemporal neighborhood features to obtain a power grid topology map in the spatiotemporal dimension; a parsing unit, configured to use a nonlinear Kalman filter state transition estimation method to parse the power grid topology map in the spatiotemporal dimension to determine the state transition mode of the power grid topology map; a mechanism model construction unit, configured to construct a spatiotemporal mechanism nonlinear representation model corresponding to the heterogeneous power grid data based on the state transition mode of the power grid topology map.

[0027] In an optional embodiment, the joint modeling unit includes: a graph construction subunit, configured to construct a corresponding undirected weighted graph based on heterogeneous power grid data; a spatiotemporal neighborhood information generation subunit, configured to generate a single spatiotemporal neighborhood information based on the undirected weighted graph using the Laplace operator and time difference; a joint modeling subunit, configured to use a multi-index joint constraint regularization method to perform joint modeling of spatiotemporal neighborhood information on the heterogeneous power grid data based on the single spatiotemporal neighborhood information, and obtain a power grid topology map in the spatiotemporal dimension.

[0028] In a third aspect, the present disclosure provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to execute the large-scale power grid high-order tensor network construction method of the first aspect or any corresponding embodiment thereof.

[0029] In a fourth aspect, the present disclosure provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method for constructing a large-scale power grid high-order tensor network according to the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the specific embodiments of the present disclosure or the technical solutions in the related technologies, the following briefly introduces the drawings required for use in the specific embodiments or related technical descriptions. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0031] FIG1 is a flow chart of a method for constructing a large-scale power grid high-order tensor network according to an embodiment of the present disclosure;

[0032] FIG2 is a flow chart of another method for constructing a large-scale power grid high-order tensor network according to an embodiment of the present disclosure;

[0033] FIG3 is a schematic diagram of an undirected weighted graph corresponding to heterogeneous power grid data according to an embodiment of the present disclosure;

[0034] FIG4 is a schematic diagram of the overall framework of a method for constructing a large-scale power grid high-order tensor network according to an embodiment of the present disclosure;

[0035] FIG5 is a schematic flow chart of a large-scale high-order tensor network feature learning method according to an embodiment of the present disclosure;

[0036] FIG6 is a schematic diagram of heterogeneous space merging according to an embodiment of the present disclosure;

[0037] FIG7 is a structural block diagram of a device for constructing a large-scale power grid high-order tensor network according to an embodiment of the present disclosure;

[0038] FIG8 is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0039] To make the purpose, technical solutions, and advantages of the embodiments of the present disclosure more clear, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present disclosure.

[0040] It should be noted that, in the description of the present disclosure, the terms "first", "second", etc. are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices. The directions or positional relationships indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside", etc. are based on the directions or positional relationships shown in the accompanying drawings and are only for the convenience of describing the present disclosure and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operate in a specific direction, and therefore should not be understood as limiting the present disclosure. The terms "mounted," "connected," and "connected" should be interpreted broadly. For example, they can refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediary; internal communication between two components; and wireless or wired connections. Those skilled in the art will understand the specific meanings of these terms in this disclosure based on specific circumstances.

[0041] In addition, if "and / or" appears in this disclosure, it includes three parallel solutions. For example, "A and / or B" includes solution A, solution B, or solutions that meet both A and B. In addition, the technical solutions of various embodiments may be combined with each other, but this must be based on the fact that they can be implemented by ordinary technicians in this field. If the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection claimed by this disclosure.

[0042] With the instant access to large-scale distributed resources, power grids are exhibiting new characteristics such as complex topological structures, increased operational fluctuations, and bidirectional interactions between main and distribution networks. They are gradually evolving into high-order networks with complex entities, rapid time-varying, and heterogeneous interactions. Large-scale power grids are complex digital coupling spaces with high order, high dimensionality, heterogeneity, and incompleteness. They are characterized by complex and diverse data types, different representation forms, heterogeneous node attributes, and frequent node interactions. The rapid development of power grids has brought multiple technical challenges, such as difficult data fusion, large computational scale, time-sensitive analysis, and high optimization requirements. For example, large-scale power grids can cover multiple power grids at the generation, transmission, distribution, and consumption voltage levels. They are composed of power plants (power plants and substations), primary and secondary equipment at each voltage level, and their connecting networks. Computing node types include various entities such as power equipment, power loads, and power users. Large-scale power grid computations involve a vast number of physical nodes of varying types. For example, in a production management system, there are thousands of transformer models available, and the number of grid computational objects is even greater. Traditional graph computing approaches, when faced with tens of millions of topological nodes and hundreds of levels of data fusion analysis, still face significant challenges, with response delays reaching 5 seconds per 10,000 nodes and poor analytical model scalability. Current technical approaches have significant shortcomings in expressing and utilizing the high-order, complex, heterogeneous, and incomplete panoramic power grid coupling space, resulting in high model complexity and slow analysis speeds.

[0043] Accurate characterization of large-scale power grids is the basic support for intelligent tasks such as load forecasting and fault warning. Existing characterization methods such as matrix decomposition and graph neural networks have the following shortcomings: (1) The continuous network space needs to be segmented, the logical mapping is complex, and the time sequence and topology information are easily lost, resulting in damage to the complete spatiotemporal power grid structure; (2) The incompleteness, computational complexity and scalability of large-scale power grid data are not considered, and the analysis and calculation speed is slow; (3) The increase in network order leads to decentralized data storage, and heterogeneous data makes modeling more difficult and the model representation efficiency is low; (4) The description data of the equipment involved in power grid calculation is complex and diverse, the data description forms of different equipment are different, and there is no unified description of the relationship between different equipment nodes; (5) There are many types of power grid topology nodes, which makes network topology maintenance difficult. Therefore, how to use a simpler and more efficient high-order network model to describe the complex power grid structure with multi-agent interaction, so as to meet the application requirements of better models, faster calculations, more accurate analysis, and less computing power and energy consumption, is a key technical problem that needs to be solved in the future development of power grids. Considering that tensor networks have good scalability and can completely preserve the spatiotemporal patterns of power grids, how to use high-order tensor networks to fully represent the heterogeneous node attributes, heterogeneous node interactions and complex mapping logic of large-scale power grids, and to comprehensively model the spatial structure information, heterogeneous information and evolution mechanism of large-scale power grids to achieve accurate panoramic modeling representation of large-scale power grids is a key difficulty that needs to be overcome in establishing large-scale power grid characterization methods.

[0044] According to an embodiment of the present disclosure, an embodiment of a method for constructing a high-order tensor network of a large-scale power grid is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a terminal such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0045] In this embodiment, a method for constructing a large-scale power grid high-order tensor network is provided, which can be used in the above-mentioned terminals, such as central processing units, servers, etc. FIG1 is a flow chart of the method for constructing a large-scale power grid high-order tensor network according to an embodiment of the present disclosure. As shown in FIG1 , the flow chart includes the following steps:

[0046] S101: Obtain multiple attribute sets corresponding to heterogeneous nodes in a large-scale power grid. Optionally, a large-scale power grid includes a large number of heterogeneous nodes, i.e., nodes of different types and attributes. Heterogeneous nodes may be various types of equipment constituting the power grid, such as generators, buses, loads, transformers, and lines. Heterogeneous nodes may include multiple attributes, such as geographic location, voltage level, and construction age. The attribute set is a set consisting of multiple attributes.

[0047] S102: Extract features from the multiple attribute sets to obtain heterogeneous node attribute features. Optionally, extract features from the multiple attribute sets obtained in S101 using machine learning or other methods to obtain heterogeneous node attribute features. The extracted heterogeneous node attribute features can be stored in the form of tensors or matrices.

[0048] S103: Using a deep hash mapping model to establish multiple feature subspaces based on the attribute features of heterogeneous nodes, any one feature subspace corresponds to a class of heterogeneous node interaction relationships. Optionally, a deep hash mapping model is used to map the heterogeneous node attribute features of S102 to a low-dimensional common feature space. Feature extraction is then performed in this common feature space to obtain multiple feature subspaces. The attribute features on each feature subspace are fed back to the heterogeneous node relationship to represent a specific class of interaction, thereby achieving the purpose of using multiple feature subspaces to represent the interaction logic of heterogeneous nodes in large-scale power grids.

[0049] S104: Utilize a breadth learning strategy to align multiple feature subspaces to a unified feature dimension, thereby obtaining an attribute feature vector of the unified feature dimension. Optionally, the core consideration of the breadth learning strategy is to extract more valuable information by fusing different types of data. For example, an autoencoder can be employed based on the breadth learning strategy to align multiple subspaces to a unified feature dimension, thereby achieving a unified dimension for the attribute feature vector.

[0050] S105, using a distance measurement method to calculate the distance measurement value between heterogeneous nodes based on the attribute feature vector of the unified feature dimension, the distance measurement value is used to determine the weight of the interaction relationship between heterogeneous nodes. Optionally, the distance measurement method includes information entropy, mutual information, correlation coefficient, etc., and the distance measurement method is used to calculate the distance measurement value between heterogeneous nodes based on the attribute feature vector of the unified feature dimension in S104, that is, to quantify the specific relationship between heterogeneous nodes, thereby converting the attribute information of heterogeneous nodes into a heterogeneous relationship between feature vectors. Among them, the weight of the interaction relationship between heterogeneous nodes can be calculated based on the distance measurement value, thereby achieving the purpose of determining the connection relationship between heterogeneous nodes.

[0051] S106: Determine a basis tensor for representing heterogeneous node interaction relationships based on the attribute feature vectors of the unified feature dimension and the weights of the heterogeneous node interaction relationships. Optionally, the interaction relationship can be a single interaction relationship or multiple interaction relationships, determined based on the analysis task requirements. Taking a single interaction relationship as an example, a set of heterogeneous node relationships with a single relationship can be determined based on the attribute feature vectors obtained in S104 and the weights obtained in S105, thereby obtaining a set of basis tensors representing the characteristics of a single relationship.

[0052] S107, using tensor multiplication operation rules to construct a large-scale power grid high-order tensor network according to the basis tensor and the weights of the interaction relationship between heterogeneous nodes. Optionally, based on the basis tensor determined in S106 and the weights calculated according to the distance metric value, a high-order tensor network is constructed using tensor multiplication operation rules. It should be noted that based on different analysis task requirements, different orders of basis tensors can be used to construct high-order tensor networks of different forms. For example, a third-order basis tensor can describe a ternary relationship, and a high-order tensor network is constructed using tensor multiplication operation rules such as tensor modular product according to the weights of the large-scale power grid node association relationship, i.e., the interaction relationship, corresponding to different orders of tensors.

[0053] In the disclosed embodiment, a tensor network is introduced to characterize a large-scale power grid. A deep hash mapping model is used to establish multiple feature subspaces based on the extracted heterogeneous node attribute features to represent the interaction relationship between heterogeneous nodes. A distance metric method is used to calculate the weight of the interaction relationship between heterogeneous nodes based on the aligned multiple feature subspaces. Then, a base tensor representing the interaction relationship is determined, and the base tensor is used to construct a high-order tensor network to characterize the large-scale power grid. Because the tensor network has the distributed storage and parallel processing capabilities for high-dimensional heterogeneous features and complex associations, the use of the tensor network to model the large-scale power grid achieves the effective integration of high-dimensional and complex features, thereby improving the scalability of the large-scale power grid model and reducing the computational complexity, solving the problem of the lack of an efficient model for characterizing large-scale power grids in the related art.

[0054] In this embodiment, a method for constructing a large-scale power grid high-order tensor network is provided, which can be used in the above-mentioned terminals, such as central processing units, servers, etc. FIG2 is a flow chart of another method for constructing a large-scale power grid high-order tensor network according to an embodiment of the present disclosure. As shown in FIG2 , the flow chart includes the following steps:

[0055] S201: Acquire multiple attribute sets corresponding to heterogeneous nodes in a large-scale power grid. For details, please refer to S101 in the embodiment shown in FIG1 , which will not be described in detail here.

[0056] S202, extracting features from multiple attribute sets to obtain heterogeneous node attribute features. Optionally, S202 includes:

[0057] a1, extract features corresponding to multiple category attribute sets.

[0058] a2, stores the features corresponding to any extracted category attribute set as a second-order tensor.

[0059] a3, determine multiple second-order tensors corresponding to multiple category attribute sets.

[0060] a4, taking multiple second-order tensors as heterogeneous node attribute features.

[0061] For example, for a large-scale grid with heterogeneous nodes with different characteristics, the method extracts the features corresponding to the various attribute sets corresponding to the various attribute sets, taking into account the differences in node attribute data, such as the substation's geographic location, voltage level, and construction age. These features are then classified and stored as a second-order tensor (matrix network). Each attribute set is stored as a separate second-order tensor, thus achieving a categorized representation of the attributes of the heterogeneous nodes.

[0062] S203, using a deep hash mapping model to establish multiple feature subspaces based on the attribute features of heterogeneous nodes, where any feature subspace corresponds to a type of heterogeneous node interaction relationship. Optionally, the deep hash mapping model includes a common hash coding module and multiple hash coding submodules, and the above S203 includes:

[0063] b1, use the public hash coding module to map the heterogeneous node attribute features into the public attribute feature space.

[0064] b2, using each hash coding submodule to establish a feature subspace based on the common attribute feature space.

[0065] Specifically, the different second-order tensors obtained in 202 are used as the input of the deep hash mapping model, and the heterogeneous attribute feature spaces corresponding to different category attributes are mapped (stored) into a common attribute feature space through a common hash coding module. Then, multiple (different) hash coding sub-modules are used to extract multiple feature subspaces of heterogeneous node attributes, where the attribute features on each subspace represent a specific type of interaction relationship between heterogeneous nodes. Therefore, the heterogeneous attribute space (multiple feature subspaces) in the entire large-scale power grid corresponds to a set of high-order heterogeneous node interaction logic (interaction relationship).

[0066] S204: Utilize a breadth learning strategy to align multiple feature subspaces to a unified feature dimension, thereby obtaining an attribute feature vector of the unified feature dimension. For details, please refer to S104 in the embodiment shown in FIG1 , and will not be repeated here. It should be noted that alignment allows for more accurate learning of the relationships between features, thereby improving generalization and performance.

[0067] S205, using a distance metric method to calculate the distance metric value between heterogeneous nodes based on the attribute feature vector of the unified feature dimension, the distance metric value is used to determine the weight of the interaction relationship between the heterogeneous nodes. Optionally, the attribute feature vector of the unified feature dimension includes the attribute feature vector of the unified feature dimension corresponding to the heterogeneous node and its neighboring nodes, S205 includes:

[0068] c1, uses the distance measurement method to calculate the distance measurement value between heterogeneous nodes based on the attribute feature vector of the unified feature dimension.

[0069] c2, using graph convolution to merge the attribute feature vectors of the unified feature dimension corresponding to the heterogeneous node and its neighboring nodes based on the distance metric value to obtain a new heterogeneous node attribute feature vector.

[0070] c3, determines the weight of the interaction relationship between heterogeneous nodes based on the new heterogeneous node attribute feature vector.

[0071] As an example, distance measurement methods such as information entropy, mutual information, and correlation coefficient are used to calculate the distance measurement value (index) between nodes using the attribute feature vector of the unified feature dimension after the heterogeneous nodes are aligned, and the specific relationship between the heterogeneous nodes is quantified, thereby converting the attribute information of the heterogeneous nodes into the heterogeneous relationship of the edge in the vector, that is, converting the attribute information of the heterogeneous nodes into the heterogeneous relationship of the attribute feature vector. The aligned attribute features of the heterogeneous nodes and their neighboring nodes are merged using the graph convolution operation, and the weight (relationship weight) of the interaction relationship between the heterogeneous nodes is determined based on the new heterogeneous node attribute features. Specifically, the distance measurement value between heterogeneous nodes is calculated one by one using the attribute feature vector of the unified feature dimension after the heterogeneous nodes are aligned in the large-scale power grid topology, as the weight to measure a certain specific relationship between the heterogeneous node pairs, and then converted into the connection relationship between the heterogeneous nodes, and finally the unification of heterogeneous nodes and heterogeneous relationships is achieved.

[0072] S206: Determine a basis tensor for representing the heterogeneous node interaction relationship based on the attribute feature vector of the unified feature dimension and the weight of the heterogeneous node interaction relationship. For details, please refer to S106 in the embodiment shown in FIG1 , which will not be repeated here.

[0073] S207: Build a large-scale power grid high-order tensor network based on the basis tensors and the weights of the heterogeneous node interaction relationships using tensor multiplication operation rules. For details, please refer to S107 of the embodiment shown in FIG1 , which will not be described in detail here.

[0074] In an optional implementation manner, after S207, the following steps are further included:

[0075] S208, obtaining heterogeneous power grid data. Optionally, the heterogeneous power grid data includes not only heterogeneous node data, such as multiple attribute sets corresponding to large-scale power grid heterogeneous nodes, but also heterogeneous power grid operation data, which can be collected using wireless sensor nodes (power grid information collection nodes).

[0076] S209, compressing the high-order tensor network of the large-scale power grid based on the high-order singular value analysis framework, bit plane, run-length coding and arithmetic coding. Specifically, although a high-order (heterogeneous) tensor network representing the large-scale power grid has been constructed, in some specific problems, it is not necessary to have a tensor network first and then decompose it, but to use a tensor network to describe the state of the entire system from the beginning, and the internal dimension of the high-order tensor network of the large-scale power grid will increase rapidly with the influence of the outside world. Therefore, it is necessary to compress the high-order tensor network of the large-scale power grid so that the construction of the tensor network can continue to run. This embodiment is based on the high-order singular value analysis framework (Singular Value Decomposition, HOSVD) and combines bit planes, run-length and arithmetic coding methods to achieve compression and optimization of multi-spatiotemporal, multi-scale high-order tensor networks. Specifically, let T be the input tensor, is the result of compression and decompression. Each network data sampling pipeline accepts a primary compression parameter, the error target ε, which can be expressed as RMSE (Root Mean Squared Error) or MAE (Mean Absolute Error). The specified error target is then converted to SSE (Sum of Squares Due to Error) using the following equivalent method:

[0077] Here, C is the total number of bit-plane grid points I1…IN, and PSNR is a preset metric. First, a full, non-truncated HOSVD is run on the input dataset T, resulting in N orthogonal square factor matrices and an N-dimensional kernel of this matrix. The N-dimensional kernel of the HOSVD is flattened into a one-dimensional vector of size C, which is then scaled and converted to 64-bit integers and sorted using C, traversing the dimensions in the kernel from right to left. Theoretically, this integer sequence is processed into a C×64 binary matrix M. Next, the number of leftmost columns (bit planes) of this matrix is ​​losslessly compressed, i.e., such that the overall L2 error (mean square error) falls to the minimum under a given target. This compression is achieved using run-length encoding (RLE) and arithmetic coding (AC). Finally, the orthogonal square factor matrix is ​​compressed using a preset cost-effectiveness budget criterion, resulting in a compressed high-order tensor network for large-scale power grids.

[0078] S210: Construct a nonlinear representation model of the spatiotemporal mechanism corresponding to heterogeneous power grid data. Optionally, due to the different geographical locations of nodes in a large-scale power grid, and the fact that heterogeneous power grid data possesses certain potential patterns in time series, this embodiment models the dynamic evolution mechanism to achieve dynamic evolution deduction of the power grid topology. Specifically, S210 includes:

[0079] d1, obtain multi-factor spatiotemporal neighborhood features. Optionally, the multi-factor spatiotemporal neighborhood features are temporal and spatial features corresponding to multiple factors, such as information about how a certain attribute changes over time and space.

[0080] d2. Jointly modeling the spatiotemporal neighborhood information of the heterogeneous power grid data using a regularization method based on multi-factor spatiotemporal neighborhood characteristics to obtain a power grid topology map in the spatiotemporal dimension. Optionally, d2 includes: constructing a corresponding undirected weighted graph based on the heterogeneous power grid data; generating a single spatiotemporal neighborhood information based on the undirected weighted graph using a Laplace operator and time difference; and jointly modeling the spatiotemporal neighborhood information of the heterogeneous power grid data based on the single spatiotemporal neighborhood information using a multi-index joint constraint regularization method to obtain a power grid topology map in the spatiotemporal dimension.

[0081] In view of the situation where there are strong and weak differences in a single time or space characteristic, a basic modeling method of a single neighborhood spatiotemporal mechanism is designed. Specifically, due to the different geographical distribution of heterogeneous nodes in large-scale power grids, data from real application scenarios generally have spatial correlation. At the same time, the time series of heterogeneous power grid data can be generated by a finite number of patterns, and show strong correlation with its adjacent data in a certain time slot. Therefore, in view of spatial correlation, an undirected weighted graph G(V,ε,W) as shown in Figure 3 is constructed, where V represents the information collection node in the power grid, that is, nodes A, B...G in Figure 3, such as the set of wireless sensor nodes for data sampling, ε represents the edge, that is, the similarity relationship between wireless sensors, and W represents the weight change between associated sensors, that is, the value on the edge in Figure 3. Based on the graph structure representation, the Laplace operator is defined as follows: L=diag(d1,...,d N )-W (2)

[0082] Among them, L is the Laplace operator, W is the weight matrix, d N is the sum of the Nth row of the weight matrix W, and diag(·) is a diagonal matrix constructor. Regarding time correlation, since adjacent data show strong correlation in time, the time difference operator is introduced to represent the time correlation of the data, which is specifically expressed as follows:

[0083] Among them, T is the total time series number of data from power grid information collection nodes. Comprehensively considering the spatiotemporal neighborhood characteristics of each factor, a multi-index joint constraint regularization method is adopted to realize the joint modeling representation of spatiotemporal neighborhood information. Specifically, due to the different dimensions of the spatiotemporal neighborhood feature spaces of different factors, in order to achieve subsequent unified modeling, it is necessary to perform variable-order processing on the spatiotemporal neighborhood feature spaces of different factors to achieve the alignment of spatiotemporal neighborhoods. Assume that the spatiotemporal domain matrix corresponding to factor a is A, with a size of M×N, the feature matrix P corresponding to its spatial dimension is M×r, and the feature matrix Q corresponding to its time dimension is N×r; Assume that the spatiotemporal domain matrix corresponding to factor b is B, with a size of U×I, the feature matrix X corresponding to its spatial dimension is U×s, and the feature matrix Y corresponding to its time dimension is I×s. Use the kernel function to perform variable-order mapping on the above feature matrix according to the following formula:

[0084] Among them, κ(·) represents the mapping using kernel function, and The sizes of are M×f, N×f, U×f and I×f respectively, where f represents the unified feature space dimension. This embodiment integrates the spatiotemporal neighborhood feature space of each factor and establishes a multi-index composite regularization constraint through joint constraints, thereby realizing the joint modeling of spatiotemporal neighborhood information and achieving the purpose of improving the generalization ability of the model in different application backgrounds. Specifically, the kernel function is post-processed. and Conduct joint modeling The mathematical expression is as follows:

[0085] Among them, β1 and β2 represent the control parameters of the regularization terms constructed by different indicators, reg(·) represents the regularization function, Represents the loss function. It should be noted that regularization can adopt L1 regularization, L2 regularization, L1 and L2 mixed regularization, graph regularization and other schemes. This embodiment adaptively adjusts the control parameters β1 and β2 of different regularization terms based on particle swarm optimization to achieve adaptive fusion of multi-factor spatiotemporal neighborhood information. Specifically, adjustment is performed according to the following formula:

[0086] Among them, s = [β1, β2] is a two-dimensional vector composed of parameters that need to be adaptively adjusted, v represents the update step size, s pbest and s gbest where represents the individual and population optima, respectively; w represents the inertia factor; c1 and c2 represent the acceleration coefficients; and r1 and r2 represent random numbers between [0, 1]. To improve the adaptive effect, a fitness function can be specifically designed in conjunction with composite regularization constraints.

[0087] d3, using the nonlinear Kalman filter state transition estimation method to analyze the grid topology in the time and space dimensions to determine the state transition mode of the grid topology. Specifically, based on the nonlinear Kalman filter state transition estimation method, the state transition mode of the grid topology in the time and space dimensions generated by d2 is analyzed, thereby realizing the deduction of the dynamic evolution of the grid topology. For example, using the nonlinear Kalman filter dynamic estimation method in automatic control theory, a state transition space is constructed to determine the state transition mode of the grid topology:

[0088] Among them, f(*) and h(*) represent different nonlinear functions, x k and x k-1 Respectively represent the grid topology data corresponding to the kth moment and the k-1th moment, w k represents the state transition noise, v k represents the measurement noise, z k Represents known sampled data in spatiotemporal neighborhood information.

[0089] d4, constructing a nonlinear representation model of the spatiotemporal mechanism corresponding to the heterogeneous power grid data based on the state transition pattern of the power grid topology graph. Optionally, constructing a nonlinear representation model of the spatiotemporal mechanism corresponding to the heterogeneous power grid data based on the state transition pattern of the power grid topology graph.

[0090] S211: Generate a large-scale power grid high-order tensor network that includes a dynamic evolution mechanism based on the spatiotemporal mechanism nonlinear representation model and the compressed large-scale power grid high-order tensor network. Optionally, the spatiotemporal mechanism nonlinear representation model is combined with the compressed large-scale power grid high-order tensor network to obtain a large-scale power grid high-order tensor network that includes a dynamic evolution mechanism, thereby accurately modeling the dynamic evolution mechanism of the large-scale power grid and solving the problem of efficient and accurate panoramic modeling of the large-scale power grid.

[0091] This embodiment innovatively adopts a high-order tensor network model to characterize large-scale power grids, merges, aligns and separates the attribute space of heterogeneous nodes in large-scale power grids to obtain different-order basis tensors representing the relationship between heterogeneous nodes, and constructs a high-order tensor network of data grids with high scalability based on task orientation, thereby achieving the purpose of comprehensive modeling and representation of large-scale power grid topological structure information and heterogeneous information; by jointly modeling the spatiotemporal domain information of digital networks and constructing a nonlinear representation model of spatiotemporal mechanisms, the purpose of accurately modeling the dynamic evolution mechanism of large-scale power grids is achieved, thereby solving the problem of efficient and accurate panoramic modeling and representation of large-scale power grids.

[0092] In an optional embodiment, FIG4 is a schematic diagram of the overall framework of the method for constructing a large-scale power grid high-order tensor network according to an embodiment of the present disclosure. As shown in FIG4, first, heterogeneous power grid data is obtained. Secondly, a digital power grid (large-scale power grid) high-order tensor network model is performed based on the heterogeneous power grid data, specifically including: spatial merging of heterogeneous node attributes of large-scale power grid, interactive transformation of heterogeneous node attributes into heterogeneous nodes, and spatial separation of heterogeneous node attributes. For details, please refer to S202 to S207 of the embodiment shown in FIG2, which will not be repeated here. Thirdly, a compressed large-scale power grid high-order tensor is obtained by a digital power grid high-order tensor network compression method, specifically including: feature compression using high-order singular value analysis, bit plane coding and other methods. For details, please refer to S208 to S209 of the embodiment shown in FIG2, which will not be repeated here. Then, a digital power grid dynamic evolution mechanism based on a high-order tensor network model is established, specifically including: basic modeling of a single neighborhood spatiotemporal mechanism, joint modeling of spatiotemporal neighborhood information, and construction of a spatiotemporal mechanism nonlinear representation model. For details, please refer to S210 to S211 of the embodiment shown in FIG2, which will not be repeated here. Finally, a high-order tensor network representing the large-scale power grid is obtained.

[0093] In an optional embodiment, FIG5 is a flowchart of a large-scale high-order tensor network feature learning method according to an embodiment of the present disclosure, corresponding to the digital power grid high-order tensor network modeling based on heterogeneous power grid data in FIG4. As shown in FIG5, the large-scale high-order tensor network feature learning method includes the following steps:

[0094] S501: Extract features corresponding to attribute sets of different categories. Please refer to S202 of the embodiment shown in FIG2 for details, which will not be repeated here.

[0095] S502: Using a deep hash mapping model to take the extracted node features as input. For details, please refer to S203 of the embodiment shown in FIG2 , which will not be described in detail here.

[0096] S503, heterogeneous attribute feature space transformation. For details, please refer to S203 of the embodiment shown in FIG2 , which will not be described in detail here.

[0097] S504: Obtaining multiple feature subspaces of heterogeneous node attributes. For details, please refer to S203 in the embodiment shown in FIG2 , which will not be described in detail here.

[0098] S505: Feature merging. Please refer to S204 in the embodiment shown in FIG2 for details, which will not be repeated here.

[0099] S506: Calculate the node and relationship weights. Please refer to S205 in the embodiment shown in FIG2 for details, which will not be repeated here.

[0100] S507: Obtain a basis tensor representing a single relationship feature. For details, please refer to S206 in the embodiment shown in FIG2 , which will not be described in detail here.

[0101] S508: Constructing a high-order tensor network. For details, please refer to S207 of the embodiment shown in FIG2 , which will not be described in detail here.

[0102] In an optional embodiment, FIG6 is a schematic diagram of heterogeneous space merging according to an embodiment of the present disclosure, as shown in FIG6 , including three stages: feature extraction, feature merging, and subspace alignment. First, a set of latent features is extracted from a set of multiple category attributes (attribute A...attribute F) of heterogeneous nodes of the digital power grid, and then input into a deep hash mapping model, and a common hash coding module is used to transform the heterogeneous attribute feature space corresponding to different category attributes into a common attribute feature space. Multiple feature subspaces of heterogeneous node attributes are extracted according to different hash coding modules, where the attribute features on each subspace are fed back to the heterogeneous node relationship to represent a specific type of interaction, so that the entire digital power grid heterogeneous attribute space corresponds to a set of high-order heterogeneous node interaction logics, and finally the multiple subspaces are aligned to a unified feature dimension through a breadth learning strategy.

[0103] This embodiment innovatively adopts a high-order tensor network model to characterize large-scale power grids. First, the attribute space of heterogeneous nodes in the large-scale power grid is merged, aligned and separated to obtain different-order basis tensors representing the relationship between heterogeneous nodes, thereby constructing a high-order tensor network for power grids with high scalability based on task orientation, and realizing comprehensive modeling representation of large-scale power grid topological structure information and heterogeneous information; then, based on the high-order singular value analysis framework and combined with bit plane, run-length and arithmetic coding methods, a high-order tensor network compression method for power grids is adopted to achieve the effect of reducing the internal dimension of the high-order tensor network; finally, by jointly modeling the spatiotemporal domain information of digital networks and constructing a nonlinear representation model of spatiotemporal mechanisms, accurate modeling of the dynamic evolution mechanism of large-scale power grids is achieved, thereby solving the problem of efficient and accurate panoramic modeling representation of large-scale power grids.

[0104] This embodiment also provides a large-scale power grid high-order tensor network construction device, which is used to implement the above-mentioned embodiments and preferred implementations. Details that have already been described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and contemplated.

[0105] This embodiment provides a large-scale power grid high-order tensor network construction device, as shown in Figure 7, including: an attribute set acquisition module 701, configured to obtain multiple categories of attribute sets corresponding to heterogeneous nodes in a large-scale power grid; a feature extraction module 702, configured to extract features from multiple categories of attribute sets to obtain heterogeneous node attribute features; a multiple feature subspace establishment module 703, configured to use a deep hash mapping model to establish multiple feature subspaces based on heterogeneous node attribute features, and any multiple feature subspace corresponds to a type of heterogeneous node interaction relationship; a feature alignment module 704, configured to use a breadth learning strategy to align multiple feature subspaces to a unified Feature dimension, obtain the attribute feature vector of the unified feature dimension; the distance metric value calculation module 705 is configured to use the distance metric method to calculate the distance metric value between heterogeneous nodes according to the attribute feature vector of the unified feature dimension, and the distance metric value is used to determine the weight of the heterogeneous node interaction relationship; the base tensor determination module 706 is configured to determine the base tensor used to represent the heterogeneous node interaction relationship according to the attribute feature vector of the unified feature dimension and the weight of the heterogeneous node interaction relationship; the power grid high-order tensor network construction module 707 is configured to use the tensor multiplication operation rule to construct a large-scale power grid high-order tensor network according to the base tensor and the weight of the heterogeneous node interaction relationship.

[0106] In an optional embodiment, the device also includes: a power grid data acquisition module, configured to acquire heterogeneous power grid data; a compression module, configured to compress large-scale power grid high-order tensor networks based on a high-order singular value analysis framework, bit planes, run-length coding, and arithmetic coding; a mechanism model construction module, configured to construct a spatiotemporal mechanism nonlinear representation model corresponding to the heterogeneous power grid data; and a dynamic network generation module, configured to generate a large-scale power grid high-order tensor network containing a dynamic evolution mechanism based on the spatiotemporal mechanism nonlinear representation model and the compressed large-scale power grid high-order tensor network.

[0107] In an optional embodiment, the feature extraction module includes: a feature extraction unit, configured to extract features corresponding to multiple category attribute sets; a feature storage unit, configured to store the features corresponding to any extracted category attribute set as a second-order tensor; a tensor determination unit, configured to determine multiple second-order tensors corresponding to multiple category attribute sets; and a node attribute feature generation unit, configured to use multiple second-order tensors as heterogeneous node attribute features.

[0108] In an optional embodiment, the deep hash mapping model includes a common hash coding module and multiple hash coding sub-modules, and the multiple feature subspace establishment module includes: a feature mapping unit, configured to use the common hash coding module to map heterogeneous node attribute features to a common attribute feature space; a subspace establishment unit, configured to use each hash coding sub-module to establish a feature subspace based on the common attribute feature space.

[0109] In an optional embodiment, the attribute feature vector of the unified feature dimension includes the attribute feature vector of the unified feature dimension corresponding to the heterogeneous node and its neighboring nodes, and the distance measurement value calculation module includes: a calculation unit, configured to use a distance measurement method to calculate the distance measurement value between the heterogeneous nodes based on the attribute feature vector of the unified feature dimension; a merging unit, configured to use graph convolution to merge the attribute feature vector of the unified feature dimension corresponding to the heterogeneous node and its neighboring nodes based on the distance measurement value to obtain a new heterogeneous node attribute feature vector; a determination unit, configured to determine the weight of the heterogeneous node interaction relationship based on the new heterogeneous node attribute feature vector.

[0110] In an optional embodiment, the mechanism model construction module includes: a spatiotemporal feature acquisition unit, configured to acquire multi-factor spatiotemporal neighborhood features; a joint modeling unit, configured to use a regularization method to perform joint modeling of spatiotemporal neighborhood information of heterogeneous power grid data according to multi-factor spatiotemporal neighborhood features to obtain a power grid topology map in the spatiotemporal dimension; a parsing unit, configured to use a nonlinear Kalman filter state transition estimation method to parse the power grid topology map in the spatiotemporal dimension to determine the state transition mode of the power grid topology map; a mechanism model construction unit, configured to construct a spatiotemporal mechanism nonlinear representation model corresponding to the heterogeneous power grid data based on the state transition mode of the power grid topology map.

[0111] In an optional embodiment, the joint modeling unit includes: a graph construction subunit, configured to construct a corresponding undirected weighted graph based on heterogeneous power grid data; a spatiotemporal neighborhood information generation subunit, configured to generate a single spatiotemporal neighborhood information based on the undirected weighted graph using the Laplace operator and time difference; a joint modeling subunit, configured to use a multi-index joint constraint regularization method to perform joint modeling of spatiotemporal neighborhood information on the heterogeneous power grid data based on the single spatiotemporal neighborhood information, and obtain a power grid topology map in the spatiotemporal dimension.

[0112] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0113] The large-scale power grid high-order tensor network construction device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0114] The embodiment of the present disclosure also provides a computer device having the large-scale power grid high-order tensor network construction device shown in FIG7 above.

[0115] Please refer to Figure 8, which is a structural diagram of a computer device provided by an optional embodiment of the present disclosure. As shown in Figure 8, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. The various components are connected to each other using different buses and can be installed on a common motherboard or installed in other ways as needed. The processor can process instructions executed within the computer device, including instructions stored in or on the memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 8 takes a processor 10 as an example.

[0116] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0117] The memory 20 stores instructions that can be executed by at least one processor 10, so as to enable at least one processor 10 to execute the method shown in the above embodiment.

[0118] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created based on the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0119] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0120] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0121] The embodiments of the present disclosure also provide a computer-readable storage medium. The above-mentioned method according to the embodiments of the present disclosure can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0122] Although the embodiments of the present disclosure have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present disclosure, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for constructing a high - order tensor network of a large - scale power grid, where the large - scale power grid includes heterogeneous nodes, and the method includes: Obtaining a set of multiple category attributes corresponding to the heterogeneous nodes in the large - scale power grid; Performing feature extraction on the set of multiple category attributes to obtain heterogeneous node attribute features; Using a deep hash mapping model to establish multiple feature sub - spaces according to the heterogeneous node attribute features, where any one - fold feature sub - space corresponds to a type of heterogeneous node interaction relationship; Using a breadth - first learning strategy to align the multiple feature sub - spaces to a unified feature dimension to obtain an attribute feature vector with a unified feature dimension; Adopting a distance metric method to calculate the distance metric value between heterogeneous nodes according to the attribute feature vector with a unified feature dimension, and the distance metric value is used to determine the weight of the heterogeneous node interaction relationship; Determining a base tensor for representing the heterogeneous node interaction relationship according to the attribute feature vector with a unified feature dimension and the weight of the heterogeneous node interaction relationship; Using the tensor multiplication operation rule to construct a high - order tensor network of the large - scale power grid according to the base tensor and the weight of the heterogeneous node interaction relationship.

2. The method for constructing a large-scale power grid high-order tensor network according to claim 1, wherein, The method further includes: Obtaining heterogeneous power grid data; Compressing the high - order tensor network of the large - scale power grid based on a high - order singular value analysis framework, bit - plane, run - length encoding, and arithmetic encoding; Constructing a spatio - temporal mechanism non - linear representation model corresponding to the heterogeneous power grid data; Generating a high - order tensor network of the large - scale power grid containing dynamic evolution mechanisms based on the spatio - temporal mechanism non - linear representation model and the compressed high - order tensor network of the large - scale power grid.

3. The method for constructing a large-scale power grid high-order tensor network according to claim 1, wherein, The performing feature extraction on the set of multiple category attributes to obtain heterogeneous node attribute features includes: Extracting features corresponding to the set of multiple category attributes; Storing the features corresponding to any one - fold category attribute set as a second - order tensor; Determining multiple second - order tensors corresponding to the set of multiple category attributes; Using the multiple second - order tensors as heterogeneous node attribute features.

4. The method for constructing a large-scale power grid high-order tensor network according to claim 1, wherein, The deep hash mapping model includes a common hash encoding module and multiple hash encoding sub - modules. The using a deep hash mapping model to establish multiple feature sub - spaces according to the heterogeneous node attribute features, where any one - fold feature sub - space corresponds to a type of heterogeneous node interaction relationship, includes: Using the common hash encoding module to map the heterogeneous node attribute features to a common attribute feature space; Using each hash encoding sub - module to establish a one - fold feature sub - space according to the common attribute feature space.

5. The method for constructing a large-scale power grid high-order tensor network according to claim 1, wherein, The attribute feature vector with a unified feature dimension includes the attribute feature vectors with a unified feature dimension corresponding to the heterogeneous nodes and their neighbor nodes. The adopting a distance metric method to calculate the distance metric value between heterogeneous nodes according to the attribute feature vector with a unified feature dimension, and the distance metric value is used to determine the weight of the heterogeneous node interaction relationship, includes: Adopting a distance metric method to calculate the distance metric value between heterogeneous nodes according to the attribute feature vector with a unified feature dimension; Using graph convolution to merge the attribute feature vectors with a unified feature dimension corresponding to the heterogeneous nodes and their neighbor nodes based on the distance metric value to obtain a new heterogeneous node attribute feature vector; Determining the weight of the heterogeneous node interaction relationship according to the new heterogeneous node attribute feature vector.

6. The method for constructing a large-scale power grid high-order tensor network according to claim 2, wherein The construction of the spatio-temporal mechanism non-linear representation model corresponding to heterogeneous power grid data includes: Obtain multi-factor spatio-temporal neighborhood features; Use the regularization method to jointly model the spatio-temporal neighborhood information of heterogeneous power grid data based on the multi-factor spatio-temporal neighborhood features, and obtain a power grid topology graph in the spatio-temporal dimension; Use the non-linear Kalman filter state transition estimation method to analyze the power grid topology graph in the spatio-temporal dimension, and determine the power grid topology graph state transition mode; Construct a spatio-temporal mechanism non-linear representation model corresponding to heterogeneous power grid data based on the power grid topology graph state transition mode.

7. The method for constructing a large-scale power grid high-order tensor network according to claim 6, wherein, The use of the regularization method to jointly model the spatio-temporal neighborhood information of heterogeneous power grid data based on the multi-factor spatio-temporal neighborhood features, and obtain a power grid topology graph in the spatio-temporal dimension, includes: Construct a corresponding undirected weighted graph according to the heterogeneous power grid data; Use the Laplacian operator and time difference to generate single spatio-temporal neighborhood information based on the undirected weighted graph; Adopt a multi-index joint constraint regularization method to jointly model the spatio-temporal neighborhood information of heterogeneous power grid data based on the single spatio-temporal neighborhood information, and obtain a power grid topology graph in the spatio-temporal dimension.

8. A device for constructing a large-scale power grid high-order tensor network. The large-scale power grid includes heterogeneous nodes. The device includes: An attribute set acquisition module configured to acquire a variety of category attribute sets corresponding to heterogeneous nodes in the large-scale power grid; A feature extraction module configured to extract features from the variety of category attribute sets to obtain heterogeneous node attribute features; A multi-feature subspace establishment module configured to establish a multi-feature subspace according to the heterogeneous node attribute features by using a deep hashing mapping model. Any one of the multi-feature subspaces corresponds to a type of heterogeneous node interaction relationship; A feature alignment module configured to align the multi-feature subspaces to a unified feature dimension by using a breadth learning strategy to obtain an attribute feature vector with a unified feature dimension; A distance metric value calculation module configured to calculate the distance metric value between heterogeneous nodes according to the attribute feature vector with a unified feature dimension by using a distance metric method. The distance metric value is used to determine the weight of the heterogeneous node interaction relationship; A base tensor determination module configured to determine a base tensor for representing the heterogeneous node interaction relationship according to the attribute feature vector with a unified feature dimension and the weight of the heterogeneous node interaction relationship; A power grid high-order tensor network construction module configured to construct a large-scale power grid high-order tensor network according to the base tensor and the weight of the heterogeneous node interaction relationship by using the tensor multiplication operation rule.

9. The large-scale power grid high-order tensor network construction device according to claim 8, wherein, The device further includes: A power grid data acquisition module configured to acquire heterogeneous power grid data; A compression module configured to compress the large-scale power grid high-order tensor network based on a high-order singular value analysis framework, bit plane, run-length encoding, and arithmetic encoding; A mechanism model construction module configured to construct a spatio-temporal mechanism non-linear representation model corresponding to heterogeneous power grid data; A dynamic network generation module configured to generate a large-scale power grid high-order tensor network containing dynamic evolution mechanisms based on the spatio-temporal mechanism non-linear representation model and the compressed large-scale power grid high-order tensor network.

10. The large-scale power grid high-order tensor network construction device according to claim 8, wherein, The feature extraction module includes: A feature extraction unit configured to extract features corresponding to a variety of category attribute sets; A feature storage unit configured to store the features corresponding to any set of category attributes extracted as a second-order tensor; A tensor determination unit configured to determine multiple second-order tensors corresponding to multiple sets of category attributes; A node attribute feature generation unit configured to use the multiple second-order tensors as heterogeneous node attribute features. The depth hash mapping model includes a common hash encoding module and multiple hash encoding sub-modules. The multiple feature subspace establishment module includes:

11. The large-scale power grid high-order tensor network construction device according to claim 8, wherein, A feature mapping unit configured to map heterogeneous node attribute features to a common attribute feature space using the common hash encoding module; A subspace establishment unit configured to establish a single feature subspace based on the common attribute feature space using each hash encoding sub-module. The attribute feature vectors with unified feature dimensions include the attribute feature vectors with unified feature dimensions corresponding to heterogeneous nodes and their neighbor nodes. The distance metric value calculation module includes:

12. The large-scale power grid high-order tensor network construction device according to claim 8, wherein, A calculation unit configured to calculate the distance metric value between heterogeneous nodes according to the attribute feature vectors with unified feature dimensions using a distance metric method; A merging unit configured to merge the attribute feature vectors with unified feature dimensions corresponding to heterogeneous nodes and their neighbor nodes based on the distance metric value using graph convolution to obtain new heterogeneous node attribute feature vectors; A determination unit configured to determine the weights of the interaction relationships of heterogeneous nodes according to the new heterogeneous node attribute feature vectors. The mechanism model construction module includes:

13. The large-scale power grid high-order tensor network construction device according to claim 9, wherein, A spatio-temporal feature acquisition unit configured to acquire multi-factor spatio-temporal neighborhood features; A joint modeling unit configured to perform joint modeling of spatio-temporal neighborhood information on heterogeneous power grid data according to multi-factor spatio-temporal neighborhood features using a regularization method to obtain a power grid topology graph in the spatio-temporal dimension; An analysis unit configured to analyze the power grid topology graph in the spatio-temporal dimension using a non-linear Kalman filter state transition estimation method to determine the state transition mode of the power grid topology graph; A mechanism model construction unit configured to construct a spatio-temporal mechanism non-linear representation model corresponding to the heterogeneous power grid data based on the state transition mode of the power grid topology graph. The joint modeling unit includes:

14. The large-scale power grid high-order tensor network construction device according to claim 13, wherein, A graph construction sub-unit configured to construct an undirected weighted graph corresponding to the heterogeneous power grid data; A spatio-temporal neighborhood information generation sub-unit configured to generate single spatio-temporal neighborhood information according to the undirected weighted graph using a Laplacian operator and time difference; A joint modeling sub-unit configured to perform joint modeling of spatio-temporal neighborhood information on heterogeneous power grid data according to the single spatio-temporal neighborhood information using a multi-index joint constraint regularization method to obtain a power grid topology graph in the spatio-temporal dimension.

15. A computer device, comprising: A memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to execute the large-scale power grid high-order tensor network construction method according to any one of claims 1 to 7.

16. A computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to cause a computer to execute the large-scale power grid high-order tensor network construction method according to any one of claims 1 to 7. ​

Images