A power grid future state deduction method and system

Abstract

The application provides a power grid future state deduction method and system, wherein the method comprises: collecting first data, preprocessing the first data to obtain a sample data set; constructing a graph structure of the power grid according to the sample data set, the graph structure comprising nodes and edges; grouping and clustering the nodes and assigning a category label to each node to obtain clustered nodes; performing multi-layer feature propagation on the graph structure through a graph neural network model according to the clustered nodes to extract node embedding features; the node embedding features are used to reflect topological dependence, device coupling and operation semantics; mapping the power grid topological relationship and the node embedding features to structured semantics; inputting the structured semantics into a large language model to obtain a future state power grid model through a retrieval enhancement mechanism of the large language model. In the above manner, the application can identify potential risks in advance, provide reliable basis for safety checking and dispatching decision-making, and reduce the probability of accidents that may occur in power grid operation.

Description

Technical Field

[0001] This application relates to the field of power system analysis technology, and in particular to a method and system for predicting the future state of a power grid. Background Technology

[0002] In modern power system planning and operation, to meet the multiple demands of load growth, renewable energy integration, and maintenance, power grid modeling involves not only real-time operational states but also the construction and management of planned states, future states, and multiple scenario versions. Future-state power grid models are typically generated based on master equipment data from the planning model, combined with equipment, topology, and operational rules, resulting in annual, monthly, and maintenance-specific versions. These multi-version power grid models play a crucial supporting role in safety verification, power flow calculation, fault analysis, and planning decisions.

[0003] However, the construction and maintenance of actual power grid models face multiple challenges. The topological relationships of power grid equipment are complex, involving various devices such as main transformers, switches, disconnectors, busbars, and branches. Their multi-layered, multi-level, and cross-voltage-level interconnections make it difficult for traditional manual or rule-based methods to comprehensively and accurately generate a complete topology. Multi-temporal and multi-scenario power grid models are prone to omissions, redundancies, or conflicts during updates, completions, and discrepancy comparisons, making it difficult to guarantee consistency and reliability between versions. Furthermore, existing methods rely heavily on manual verification or single rules for checking power grid equipment parameters, topological constraints, and identification information, lacking intelligent analysis tools and failing to efficiently apply complex rules and historical experience.

[0004] Existing research often employs statistical learning or deep learning-based methods to predict power grid states, but these typically focus on power flow or power flow prediction, failing to clearly define the topological integrity, missing equipment, and structural anomalies in the output power grid model. This results in insufficient model interpretability, making it difficult to meet the needs of engineering decision-making and safety verification. In recent years, Graph Neural Networks (GNNs) have been increasingly introduced into the field of power grid analysis to characterize the structural relationships between power grid nodes and lines. These methods can, to some extent, integrate power grid topology information and improve the ability to model spatial correlations. However, existing GNNs are only used for node state feature extraction and deduction, not for structural completion or multi-version consistency analysis. Furthermore, this method lacks rule enhancement and historical experience support, relying on manual or static rules for equipment completion, topology verification, and parameter rationality analysis, making it difficult to adapt to complex, multi-temporal power grid modeling scenarios.

[0005] Therefore, there is an urgent need for a modeling method and system that can intelligently generate topology completion, automatically verify parameters and connection relationships, support version management and difference analysis, and provide interpretable analysis by combining rule knowledge and historical experience for multi-version models of power grid planning, operation and future states, so as to provide high-quality and usable future state power grid model technical support for safety verification, planning decision and operation and maintenance scheduling. Summary of the Invention

[0006] This application provides a method and system for predicting the future state of a power grid, which can realize the dual functions of model prediction and structural verification, and ensure stability and reliability under complex operating conditions.

[0007] Firstly, this application provides a method for predicting the future state of a power grid, which is executed by a computing device. The computing device can be understood as a computer or similar device, and is not limited thereto in this application. The method includes:

[0008] First data is collected and preprocessed to obtain a sample dataset. This first data represents historical model data and real-time operational data required for the future state projection of the target power grid. Based on the sample dataset, a graph structure of the power grid is constructed, including nodes and edges. Nodes are grouped and clustered, and each node is assigned a category label to obtain clustered nodes. Based on the clustered nodes, a graph neural network model is used to perform multi-layer feature propagation on the graph structure to extract node embedding features. Node embedding features are used to reflect topological dependencies, equipment coupling, and operational semantics. The power grid topology and node embedding features are mapped to structured semantics. The structured semantics are input into a large language model, and through the retrieval enhancement mechanism of the large language model, a future power grid model is obtained.

[0009] Through the above methods, this application first collects first data and preprocesses it to obtain a preprocessed sample dataset. Then, it constructs a power grid diagram structure for the sample dataset to ensure the accuracy and rationality of the collected sample data. At the same time, it groups and clusters the nodes and assigns a category label to each node to obtain the clustered nodes. Through graph neural networks, it performs multi-layer feature extraction on the power grid topology and equipment characteristics, which can fully capture the coupling relationship and topological dependence between nodes, significantly reduce the inference error caused by input data fluctuations or abnormal equipment status, and use historical operating data and planning rules to perform knowledge enhancement, so as to achieve robust inference of future multi-version models and improve the applicability of the model in complex topology and multi-node coupling situations.

[0010] The aforementioned method for projecting the future state of power grids also includes:

[0011] The inference results of the graph neural network model are verified. The inference results are at least one future state power grid model. Each future state power grid model includes at least one of the following: complete power grid topology, equipment parameters and operating status, model generation path, and inference basis identifier. The verification includes at least one of the following: topology integrity verification, parameter rationality verification, and rule consistency verification. Among them, topology integrity verification is used to determine whether there are broken links, islands, or illegal connections; parameter rationality verification is used to verify whether the equipment parameters meet the rated range constraints; and rule consistency verification is used to verify whether the model conforms to the planning and scheduling rules.

[0012] Through the above methods, after obtaining the future power grid model, this application verifies, filters, and standardizes the output of the power grid model's projection results, ensuring that the future power grid model can be directly used to regulate the cloud business system, while significantly reducing projection errors caused by fluctuations in input data or abnormal equipment status.

[0013] In the aforementioned method for predicting the future state of the power grid, the first data is collected, and the first data is preprocessed, including:

[0014] Collect primary data, which includes at least one of the following: main equipment model data, topology information and connection relationship data, operating status data, equipment parameters, and historical operating and planning data; add data traceability tags to all primary data, which are used to mark the data source and time of the primary data; perform data cleaning and standardization on the primary data, and remove outliers and missing values; unify and normalize the dimensions of equipment parameters, and perform consistency checks on the topology structure.

[0015] Through the above methods, this application collects first data, obtains multi-source basic data required for the future power grid model deduction, and forms standardized and structured graph computation input through unified preprocessing, providing a reliable data foundation for subsequent graph construction and graph neural network modeling.

[0016] In the aforementioned method for predicting the future state of the power grid, the graph structure of the power grid is constructed, including:

[0017] Based on the power grid topology and equipment characteristics, the power grid is abstracted into a graph structure; power equipment or connection points are mapped as graph nodes, and lines or topological connections are mapped as edges.

[0018] By using the above method, the power grid is abstracted into a graph structure based on the power grid topology and equipment characteristics. The graph structure contains graph nodes mapped by power equipment or connection points and edges mapped by line or topology connection relationships, which reduces the modeling complexity of large-scale power grids and improves the stability and generalization ability of feature learning.

[0019] In the aforementioned method for predicting the future state of the power grid, the nodes are grouped and clustered, including:

[0020] The nodes are clustered using either the K-means clustering algorithm based on node static attributes or the spectral clustering algorithm based on power grid topology similarity.

[0021] The objective function for clustering is as follows:

[0022]

[0023] Where min represents the clustering result of the nodes that minimizes the objective function. Indicates the number of cluster categories. Describes the set of nodes of the kth class. This represents the feature center vector of the k-th class. This represents the i-th node of the k-th class. This represents the original feature vector of the i-th node.

[0024] By using the above methods, this application performs clustering processing on nodes, which not only reduces the computational complexity of large-scale power grids but also enhances the characteristic consistency of similar equipment. At the same time, the use of K-means clustering algorithm or spectral clustering algorithm improves the credibility of clustering results and can enhance the stability, interpretability and generalization ability of the model.

[0025] In the aforementioned method for predicting the future state of the power grid, structured semantics are input into a large language model. Through the retrieval enhancement mechanism of the large language model, a future power grid model is obtained, including:

[0026] The node embedding features are mapped to topological relationships as structured semantics. This structured semantics is then input into a large language model for training in a power grid scenario. Training includes incremental pre-training using power system planning and design regulations and power grid dispatching and operation regulations as power professional corpora. Training also includes low-rank adaptation fine-tuning using historical power grid simulation cases, topology completion cases, and multi-temporal verification cases. A retrieval enhancement mechanism is used to retrieve relevant rules, historical cases, and operational specifications from a knowledge base. Based on the trained structured semantics and the relevant rules, historical cases, and operational specifications obtained through the retrieval enhancement mechanism, a future power grid model is derived.

[0027] In this way, by combining the structured features output by the graph neural network with retrieval enhancement knowledge, this application drives a large language model to complete topology completion, state inference and multi-version future state model generation, which not only enhances the model's semantic understanding ability in the power grid domain, but also enhances the safety and reliability of the power grid.

[0028] In a second aspect, this application provides a future state prediction system for power grids, used to execute the method in the first aspect of this application, including: a data acquisition module, a construction module, a clustering module, an extraction module, a mapping module, and an acquisition module;

[0029] The system comprises the following modules: a data acquisition module for acquiring first data and preprocessing it to obtain a sample dataset; the first data represents historical model data and real-time operational data required for the future state projection of the target power grid; a construction module for constructing a graph structure of the power grid based on the sample dataset, the graph structure including nodes and edges; a clustering module for grouping and clustering nodes and assigning category labels to each node to obtain clustered nodes; an extraction module for extracting node embedding features based on the clustered nodes through multi-layer feature propagation of the graph structure using a graph neural network model; the node embedding features are used to reflect topological dependencies, equipment coupling, and operational semantics; a mapping module for mapping the power grid topology and node embedding features into structured semantics; and an acquisition module for inputting the structured semantics into a large language model and obtaining the future state power grid model through the retrieval enhancement mechanism of the large language model.

[0030] Thirdly, this application also provides a computing device, comprising: a memory for storing program instructions; and a processor for calling the program instructions stored in the memory and executing the method described in the first aspect according to the obtained program instructions.

[0031] Fourthly, this application also provides a computer-readable storage medium storing computer-readable instructions, which, when read and executed by a computer, implement the method of the first aspect described above.

[0032] Fifthly, this application provides a computer program product including a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform the method described in the first aspect.

[0033] Beneficial effects:

[0034] Through the above methods, this application first collects first data and preprocesses it to obtain a preprocessed sample dataset. Then, it constructs a power grid diagram structure for the sample dataset to ensure the accuracy and rationality of the collected sample data. At the same time, it groups and clusters the nodes and assigns a category label to each node to obtain the clustered nodes. Through graph neural networks, it performs multi-layer feature extraction on the power grid topology and equipment characteristics, which can fully capture the coupling relationship and topological dependence between nodes, significantly reduce the inference error caused by input data fluctuations or abnormal equipment status, and use historical operating data and planning rules to perform knowledge enhancement, so as to achieve robust inference of future multi-version models and improve the applicability of the model in complex topology and multi-node coupling situations. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0036] Figure 1 It is a schematic flowchart of a method for future state deduction of a power grid provided in Embodiment 1 of this application;

[0037] Figure 2 It is a schematic flowchart of a method for future state deduction of a power grid provided in Embodiment 2 of this application;

[0038] Figure 3 It is a schematic structural diagram of a system for future state deduction of a power grid provided in Embodiment 3 of this application;

[0039] Figure 4 It is a schematic structural diagram of a computing device provided in the embodiments of this application. Specific embodiments

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention.

[0041] In the following embodiments of this application, "and / or" describes the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone, where A and B can be singular or plural. The character " / " generally indicates that the associated objects before and after are in an "or" relationship. "At least one (item) below" or similar expressions refer to any combination of these items, including any combination of single item (item) or multiple items (items). For example, at least one (item) of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, c can be single or multiple. The singular expression forms "a", "one kind", "the", "above-mentioned", "this", and "this one" are also intended to include expressions such as "one or more" unless there is a clear opposite indication in the context. Also, unless there is a contrary statement, the ordinal numbers such as "first", "second", etc. mentioned in the embodiments of this application are used to distinguish multiple objects and are not used to limit the order, time sequence, priority, or importance of multiple objects.

[0042] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0043] Example 1

[0044] Embodiment 1 of this application provides a method for predicting the future state of a power grid. This method is executed by a computing device, which can be understood as a computer or other similar device, but is not limited thereto in this application. The method flow is as follows: Figure 1 As shown, it includes:

[0045] Step 101: Collect the first data and preprocess it to obtain a sample dataset. The first data represents the historical model data and real-time operational data required for the future state projection of the target power grid.

[0046] Specifically, first data for future state projection is collected from the existing model library and operation database of the control cloud. The first data includes at least one of the following: main equipment model data, topology information and connection relationship data, operation status data, equipment parameters, and historical operation and planning data.

[0047] For example, the main equipment model data can include the type, rated parameters, equipment attributes, and voltage level of equipment such as generators, transformers, transmission lines, busbars, switches, and disconnectors. For instance, the main equipment model data in the model library may include the following: Equipment type: steam turbine synchronous generator; Equipment model: QFSN-300-2-20; Rated power: 300mW; Rated voltage: 20kV; Equipment attribute: synchronous generator. It should be noted that the above description of main equipment model data is merely an example. Main equipment model data can also include other model data containing main equipment data, such as the number of phases and operating status. This application does not limit this. The above steam turbine synchronous generator and its main equipment model data are merely an example, and this application does not limit this.

[0048] Topology information and connection relationship data can include electrical connections between devices, route routing, node associations, etc. It should be noted that the above-mentioned topology information and connection relationship data is only an example, and may also include other data used to describe topology information and connection relationships, such as topology node identifiers, etc. This application does not limit this.

[0049] Operational status data may include the open / closed status of switches and disconnectors, the commissioning or decommissioning status of equipment, etc. It should be noted that the above-mentioned operational status data is only an example, and may also include other content used to represent operational status data, which is not limited in this application.

[0050] Historical operation and planning data may include at least one of the following: historical operation mode, typical wiring mode, maintenance plan, and main equipment information given in the planning model. It should be noted that the above-mentioned content of historical operation and planning data is only an example, and may also include other content used to represent historical operation and planning data. This application does not limit this.

[0051] Furthermore, data source tags are added to all primary data sets to indicate their data origin and timestamp. The primary data undergoes data cleaning and standardization to remove outliers and missing values. Equipment parameters are standardized and normalized, and the topology structure is checked for consistency. Specifically, missing fields are filled in, and obviously abnormal or conflicting data is removed or marked. Field mapping and naming standardization are performed on model data from different sources and in different formats. Equipment parameters are standardized and normalized to form standardized input features. A consistency check is performed on the topology structure to ensure the integrity of nodes and connections.

[0052] In one possible implementation, this application also includes unstructured data. The processing of unstructured data includes, for example, performing word segmentation, entity recognition, and relation extraction on text data, converting the recognized information such as power equipment, maintenance time, and operating status into structured key-value pairs, and embedding them into a power grid equipment feature library, etc. This application does not limit the scope of this application.

[0053] Step 102: Based on the sample dataset, construct the graph structure of the power grid, which includes nodes and edges.

[0054] Specifically, after preprocessing, the power grid model is converted into an input format suitable for graph computation. Based on the power grid topology and equipment characteristics, the power grid is abstracted into a graph structure; power equipment or connection points are mapped as graph nodes, and lines or topological connections are mapped as edges. Simultaneously, node feature matrices and edge feature matrices are constructed to provide input for subsequent graph neural network computations.

[0055] In one possible implementation, the graph structure of the power grid is constructed based on the sample dataset, including the following steps:

[0056] Abstract the power grid as an undirected or directed graph. .

[0057] in, V represents the set of nodes, where each node represents a power grid device or electrical connection point; N represents the total number of nodes; This represents a set of edges, where each edge represents a node. With nodes The electrical connection between them.

[0058] For each node in the node set V Constructing the original feature vector of the node :

[0059] ;

[0060] In the formula, Indicates the device type. Indicates voltage level, Indicates capacity parameter, This indicates the running status information.

[0061] For nodes With nodes The edge between Constructing edge feature vectors:

[0062] ;

[0063] In the formula, Indicates the connection type. Indicates the line structure parameters. This indicates the constraints of the operational planning.

[0064] The above method is used to construct the graph structure of the power grid and to build the node feature matrix and edge feature matrix, providing input for subsequent graph neural network calculations.

[0065] Step 103: Group and cluster the nodes, and assign a category label to each node to obtain the clustered nodes.

[0066] In one possible implementation, the nodes are clustered using either a K-means clustering algorithm based on node static attributes or a spectral clustering algorithm based on grid topology similarity.

[0067] The objective function for clustering is as follows:

[0068]

[0069] Where min represents the clustering result of the nodes that minimizes the objective function. Indicates the number of cluster categories. Describes the set of nodes of the kth class. This represents the feature center vector of the k-th class. This represents the i-th node of the k-th class. This represents the original feature vector of the i-th node.

[0070] During the clustering process, the dimensions of clustering features are expanded by adding the geographical region, operating status, and planning attributes of the equipment as core clustering features on the basis of equipment type and voltage level. At the same time, a hierarchical clustering strategy is adopted for large-scale power grids, first by voltage level, then by geographical region, and finally by equipment type to avoid confusion of equipment features across voltage levels and regions.

[0071] After clustering is completed, a category label is assigned to each node. This label is used in subsequent graph neural networks to limit the range of parameter sharing, constrain message passing patterns, and group inference strategies, thereby reducing computational complexity and enhancing the consistency of features of similar devices.

[0072] Step 104: Based on the clustered nodes, perform multi-layer feature propagation on the graph structure using a graph neural network model to extract node embedding features.

[0073] Among them, node embedding features are used to reflect topological dependencies, device coupling, and operational semantics.

[0074] Specifically, under the constraint of node clustering results, a graph neural network is used to perform multi-layer feature propagation on the power grid graph to extract high-dimensional node embedding features that reflect topological dependencies, equipment coupling, and operational semantics. The steps are as follows:

[0075] Initialize node features by using the original feature vectors of the nodes as the initial input to the graph neural network.

[0076] ;

[0077] in, For nodes Embedded features in the initial layer, This represents the original feature vector of the node.

[0078] After initialization, neighborhood feature aggregation and updating are performed, specifically using a message-passing-based graph neural network structure. In a layered graph neural network, the node feature update process is as follows:

[0079] ;

[0080] in, Represents a node In the Layer embedding features; Represents a non-linear activation function; Indicates the first Layer weight matrix; Represents a neighborhood feature aggregation function; Represents a node neighboring nodes In the The set of embedded features of the layer; Represents a node The set of adjacent nodes; Indicates the first The bias vector of the layer.

[0081] Furthermore, temporal and structural features are fused. To address future state projection needs, and in conjunction with historical operational status and planning information, node features are expanded to include:

[0082] ;

[0083] in, The structural feature vector is used to characterize the structural location and electrical connection of a node in the power grid topology. It is learned by a graph neural network during the message passing process in a multi-layer neighborhood.

[0084] This represents the historical operating feature vector, which is obtained by encoding historical operating data and is used to characterize the operating behavior of equipment under historical operating conditions.

[0085] The feature vector representing planning and future state constraints is obtained by encoding planning model data and future operational constraint information.

[0086] Joint modeling can be performed through splicing, weighted fusion, or attention mechanisms to embed nodes into vectors. It also includes information on power grid topology semantics, historical operational behavior characteristics, and future planning constraints. The final node embedding features are obtained through multi-layer graph neural network propagation. This is used for subsequent semantic mapping and large-scale model derivation.

[0087] Step 105: Map the power grid topology and node embedding features into structured semantics.

[0088] Specifically, node embedding features are mapped to topological relationships as structured semantics; a graph neural network model is used as the basis for training in power grid scenarios; training includes incremental pre-training, which uses power system planning and design procedures and power grid dispatching and operation procedures as power professional corpora; training also includes low-rank adaptation fine-tuning, which uses power grid historical simulation cases, topology completion cases and multi-temporal verification cases for fine-tuning.

[0089] In one possible implementation, by combining the structured features output by the graph neural network with retrieval-enhanced knowledge, a large language model is driven to complete topology completion, state inference, and generation of multiple versions of future state models. The specific steps are as follows:

[0090] Mapping structured features to semantic input, and mapping node embedding features and topological relationships to structured semantic descriptions:

[0091] ;

[0092] In the formula, For nodes The semantic representation is used for language models to understand the semantics of power grid structure.

[0093] Step 106: Input the structured semantics into the large language model, and obtain the future power grid model through the retrieval enhancement mechanism of the large language model.

[0094] Specifically, it includes the following processes:

[0095] 1. Based on a large model, lightweight adaptation training is carried out for power grid scenarios. The training content specifically includes incremental pre-training and low-rank adaptation fine-tuning.

[0096] The power grid scenario refers to the overall environment and business context formed by the interaction and coordinated operation of physical entities such as power plants, substations, transmission lines, distribution networks and end-user equipment within a specific time and space, with the power system's production, transmission, distribution and consumption as the core.

[0097] Among them, incremental pre-training includes using power system planning and design procedures, power grid dispatching and operation procedures and other power professional corpora to complete pre-training, thereby enhancing the model’s semantic understanding ability in the power grid domain;

[0098] Low-rank adaptation fine-tuning includes fine-tuning using historical power grid simulation cases, topology completion cases, and multi-temporal verification cases.

[0099] 2. Enhanced retrieval knowledge introduction

[0100] Based on the current inference objectives, relevant rules and experiences are retrieved from the knowledge base to construct a comprehensive set of contextual information input to the large model. :

[0101] ;

[0102] in, This is a collection of relevant rules, historical cases, and operational specifications obtained from external knowledge bases through a retrieval enhancement mechanism.

[0103] The knowledge base involved in this application adopts a dual-structure construction. The power grid planning rules, dispatching procedures, and equipment parameter database are constructed into a structured knowledge graph, where nodes represent power equipment or rules, and edges represent relationships. Historical cases and maintenance plans are constructed into an unstructured text database. A hybrid retrieval strategy is employed: first, candidate knowledge is filtered from the text database using keywords; then, the candidate knowledge is combined with the structured semantic representation of the power grid. Perform similarity calculations and select the knowledge with the highest similarity as... By incorporating the context of deduction, the accuracy of retrieval can be improved.

[0104] 3. Perform future state model deduction and generate multiple versions of future state power grid models.

[0105] Large language models are based on contextual information sets Perform the following deduction tasks: generate missing switches, disconnectors and their connection relationships according to equipment and topology completion rules; infer the reasonable operating state of the equipment under future states; generate multiple candidate future state model versions under different planning or operating constraints.

[0106] Example 2

[0107] In one possible implementation, this application also provides a multi-temporal verification method for the future state projection method of a power grid. Based on Embodiment 1 of this application, the method further includes verifying the projection results of a graph neural network model.

[0108] The deduction result is at least one future power grid model. Each future power grid model includes at least one of the following: complete power grid topology, equipment parameters and operating status, model generation path and deduction basis identifier.

[0109] The verification includes at least one of the following: topology integrity verification, parameter and rationality verification, and rule consistency verification;

[0110] Among them, topology integrity verification is used to determine whether there are broken links, isolated islands or illegal connections; parameter rationality verification is used to verify whether the equipment parameters meet the rated range constraints; and rule consistency verification is used to verify whether the model conforms to the planning and scheduling rules.

[0111] For example, if the system generates three future state models, where scheme A has a bus island in a 220kV substation, scheme B has a main transformer load rate exceeding the planning limit, and scheme C simultaneously satisfies topology integrity and operational constraints, then the system automatically marks scheme C as a "recommended future state model" and records the topology completion rules, operational constraints, and historical case reference sources on which its generation was based. It should be noted that the above example is merely an example of a specific implementation of Embodiment 2 of this application, and this application does not limit it.

[0112] The verified future state model is output to the control cloud platform for security verification, operation mode generation and multi-version model management.

[0113] Example 3

[0114] Based on Embodiments 1 and 2, Embodiment 3 of this application provides a specific implementation process for a method of predicting the future state of a power grid. The steps of this method are as follows: Figure 2 As shown, it includes:

[0115] Step 201: Calculate the number obtained by the device and preprocess it;

[0116] Acquire historical model data and real-time operational data required for the future state projection of the target power grid. The historical model data includes main equipment models, topology information, switch and disconnector states, equipment parameters, and historical operation records. Clean and standardize the raw data, remove outliers and missing values, and encode the topology structure as graph nodes and edges to prepare data for input to the graph neural network.

[0117] Main equipment model data: type, rated parameters, equipment attributes, and voltage level of equipment such as generators, transformers, transmission lines, busbars, switches, and disconnectors; topology and connection data: electrical connection relationships between equipment, line routes, and node relationships; operating status data: open / closed status of switches and disconnectors, and equipment commissioning or decommissioning status; historical operation and planning data: historical operating modes, typical wiring methods, maintenance plans, and main equipment information given in the planning model; during data acquisition, data traceability markers are added to all data items, indicating the data source and timestamp.

[0118] Step 202: Construct the power grid diagram structure;

[0119] Based on the power grid topology and equipment characteristics, the power grid is abstracted into a graph structure, where nodes represent power equipment or connection points, and edges represent lines or topological connections.

[0120] Step 203, node clustering;

[0121] Clustering nodes into groups allows nodes of the same type to share information in the graph neural network, improving the ability to represent features.

[0122] Step 204, Graph Neural Network Feature Extraction;

[0123] The power grid graph structure is input into a graph neural network. Through multi-layer graph convolution or message passing mechanisms, the structural information, operating status characteristics and equipment parameters in the neighborhood of nodes are aggregated and updated. High-dimensional embedding vectors reflecting the topological dependencies and local coupling characteristics between nodes are extracted and used as structured inputs for subsequent future state model deduction.

[0124] Step 205, Knowledge Base;

[0125] Based on the current inference objective, relevant rules and experiences are retrieved from the knowledge base to construct a comprehensive set of contextual information input to the large model.

[0126] The context information set is a collection of relevant rules, historical cases, and operational specifications obtained from external knowledge bases through retrieval enhancement mechanisms.

[0127] The knowledge base is constructed using a dual structure. The power grid planning rules, dispatching procedures, and equipment parameter database are built into a structured knowledge graph, where nodes are power equipment or rules and edges are relationships. Historical cases and maintenance plans are built into an unstructured text database.

[0128] Step 206, Large Language Model Deduction;

[0129] The graph structure is propagated through a graph neural network model to extract node embedding features. The output node embedding features and topological relationships are mapped into a structured semantic representation and input into a large language model. Through the retrieval enhancement mechanism of the large language model, the comprehensive contextual information set input to the large model is constructed based on the knowledge base. Together with the structured semantic input, it forms the inference context, realizing the inference of multiple versions of the future power grid model and obtaining the future power grid model.

[0130] Accurate future state projection and dynamic parameter update mechanisms enable the early identification of potential risks such as equipment anomalies, topology changes, and multiple operating conditions, providing a reliable basis for safety verification and dispatch decisions, and reducing the probability of accidents that may occur during power grid operation.

[0131] Step 207, Model Validation and Consistency Verification;

[0132] The simulation results are verified multiple times, including parameter rationality, topology integrity and equipment constraint checks, to ensure that the model can be used for application scenarios such as planning verification, operation mode adjustment and safety analysis.

[0133] Perform multi-layered verification on the output model, including: topology integrity verification: determine whether there are broken links, isolated links, or illegal connections; parameter rationality verification: verify whether the equipment parameters meet the rated range constraints; rule consistency verification: verify whether the model conforms to the planning and scheduling rules.

[0134] Step 208: Output the deduction results.

[0135] Output the derived future state multi-version power grid model, including topology, equipment status and parameters, and provide the model generation path and credibility assessment results.

[0136] Through the above methods, this application specifically achieves the following functions:

[0137] Building intelligent topology: For multi-version power grid models, future-state planning models and complex equipment topologies, establish automatic completion, unified management and node coupling relationship modeling mechanisms to enable power grid simulation to maintain stable and accurate computing capabilities when facing topology changes, node anomalies or input data fluctuations.

[0138] Parameter verification: Intelligent verification of nodes, equipment and related parameter information, identification of potential redundancy or abnormal information, automatic data organization and standardization, thereby improving the interpretability of model inference results and providing clear and usable inference outputs for power grid safety verification and operation and maintenance scheduling.

[0139] Integrating historical knowledge mechanisms: The graph neural network is enhanced with planning rules, historical operating data and topological constraints, and a dynamic parameter update mechanism for different operating conditions is established. This enables the model to maintain high reliability and generalization ability under large-scale power grid, multi-scenario and real-time operation conditions, and achieve highly available power grid simulation.

[0140] Meanwhile, this application has the following technical effects, which improve the accuracy and stability of the future state model of the power grid: by extracting multi-layer features of the power grid topology and equipment characteristics through graph neural networks, it can fully capture the coupling relationship and topological dependence between nodes, significantly reduce the inference error caused by input data fluctuations or abnormal equipment status, and use historical operating data and planning rules to enhance knowledge, realize robust inference of multiple versions of the future state model, and improve the applicability of the model in complex topology and multi-node coupling situations.

[0141] Enhance the interpretability and multi-version management capabilities of the model: perform automatic topology completion, equipment parameter verification and version management for multiple versions of the power grid model, generate directly understandable model outputs through node feature vectors and topology information, including equipment status, topology integrity and anomaly prompts, ensure that the simulation results can be used for safety verification, operation analysis and operation and maintenance scheduling, identify the differences between different versions, and provide traceable basis for model management and auditing.

[0142] Enhance the dynamic adaptability and generalization ability of the model: Dynamically update the power grid model parameters based on real-time data and planning information to adapt to different operating conditions, topology changes and equipment scheduling strategies. By integrating historical knowledge, planning rules and real-time data, it can maintain high reliability and high availability in large-scale, multi-scenario and multi-version environments, and significantly improve the generalization ability of power grid simulation results.

[0143] Finally, this application also has the following socio-economic benefits:

[0144] Enhancing power grid safety and reliability: Accurate future state projection and dynamic parameter update mechanisms enable early identification of potential risks such as equipment anomalies, topology changes, and multiple operating conditions, providing a reliable basis for safety verification and dispatch decisions, and reducing the probability of accidents that may occur during power grid operation.

[0145] Supports intelligent decision-making across multiple versions and scenarios: The system can provide available future state models for different planning schemes, maintenance plans, and operation strategies, assisting operators in conducting multi-scenario simulations and decision analysis, and improving the scientific and economical nature of power grid dispatch.

[0146] Example 4

[0147] Embodiment 2 of this application provides a future state prediction system for a power grid. This system is used to execute the method in the first aspect of this application, as follows: Figure 3 As shown; the system includes: a data acquisition module, a construction module, a clustering module, an extraction module, a mapping module, and an acquisition module;

[0148] The system comprises the following modules: a data acquisition module for acquiring first data and preprocessing it to obtain a sample dataset; the first data represents historical model data and real-time operational data required for the future state projection of the target power grid; a construction module for constructing a graph structure of the power grid based on the sample dataset, the graph structure including nodes and edges; a clustering module for grouping and clustering nodes and assigning a category label to each node to obtain clustered nodes; an extraction module for extracting node embedding features by performing multi-layer feature propagation on the graph structure using a graph neural network model based on the clustered nodes; the node embedding features are used to reflect topological dependencies, equipment coupling, and operational semantics; a mapping module for mapping the power grid topology and node embedding features into structured semantics; and an acquisition module for inputting the structured semantics into the graph neural network model and obtaining the future state power grid model through a retrieval enhancement mechanism.

[0149] Example 5

[0150] Having introduced the future state projection system of the power grid in an exemplary embodiment of this application, we will now introduce a computing device in another exemplary embodiment of this application.

[0151] Those skilled in the art will understand that various aspects of this application can be implemented as a system, method, or program product. Therefore, various aspects of this application can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, collectively referred to herein as a "circuit," "module," or "system."

[0152] In some possible implementations, the computing device according to this application may include at least one processor and at least one memory. The memory stores a computer program that, when executed by the processor, causes the processor to perform the steps in the future state prediction method for a power grid according to various exemplary embodiments of this application described above.

[0153] The following reference Figure 4 To describe a computing device 130 according to this embodiment of the present application. Figure 4 The computing device 130 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this application. Figure 4 As shown, the computing device 130 is presented in the form of a general-purpose smart terminal (or Bluetooth headset). The components of the computing device 130 may include, but are not limited to: at least one processor 131, at least one memory 132, and a bus 133 connecting different system components (including memory 132 and processor 131).

[0154] Bus 133 represents one or more of several bus architectures, including a memory bus or memory controller, peripheral bus, processor, or local bus using any of the various bus architectures. Memory 132 may include readable media in the form of volatile memory, such as random access memory (RAM) 1321 and / or cache memory 1322, and may further include read-only memory (ROM) 1323. Memory 132 may also include a program / utility 1325 having a set (at least one) of program modules 1324, including but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment.

[0155] The computing device 130 can also communicate with one or more external devices 134 (e.g., keyboard, pointing device, etc.), and / or with any device that enables the computing device 130 to communicate with one or more other smart terminals (e.g., router, modem, etc.). This communication can be performed via input / output (I / O) interface 135. Furthermore, the computing device 130 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 136. As shown, network adapter 136 communicates with other modules used in the computing device 130 via bus 133. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with the computing device 130, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0156] In some possible implementations, various aspects of the power grid future state prediction method provided in this application can also be implemented in the form of a program product, which includes a computer program that, when run on a computer device, causes the computer device to perform the steps in the power grid future state prediction method according to various exemplary embodiments of this application as described above.

[0157] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0158] The program product for time-domain noise processing according to the embodiments of this application may employ a portable compact disc read-only memory (CD-ROM) and include a computer program, and may run on a smart terminal. However, the program product of this application is not limited thereto. In this document, the readable storage medium may be any tangible medium that contains or stores a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0159] A readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying a readable computer program. Such propagated data signals may take various forms, including—but not limited to—electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0160] It should be noted that although several units or sub-units of the device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of this application, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be further divided and embodied by multiple units.

[0161] Furthermore, although the operations of the method of this application are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0162] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable access frequency prediction device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable access frequency prediction device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0163] These computer program instructions may also be stored in a computer-readable storage medium capable of directing a computer or other programmable access predictive device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0164] These computer program instructions can also be loaded onto a computer or other programmable access predictive device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0165] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0166] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for predicting the future state of a power grid, characterized in that, include: First data is collected and preprocessed to obtain a sample dataset; wherein, the first data represents the historical model data and real-time operation data required for the future state projection of the target power grid; Based on the sample dataset, a graph structure of the power grid is constructed, which includes nodes and edges; The nodes are grouped and clustered, and each node is assigned a category label to obtain the clustered nodes; Based on the clustered nodes, a graph neural network model is used to perform multi-layer feature propagation on the graph structure to extract node embedding features; the node embedding features are used to reflect topological dependencies, device coupling, and operational semantics. Map the power grid topology and the node embedding features into structured semantics; The structured semantics are input into a large language model, and the future power grid model is obtained through the retrieval enhancement mechanism of the large language model.

2. The method according to claim 1, characterized in that, The method further includes: The derivation results of the graph neural network model are verified; The simulation result is at least one future power grid model, and each future power grid model includes at least one of the following: complete power grid topology, equipment parameters and operating status, model generation path and simulation basis identifier; The verification includes at least one of topology integrity verification, parameter and rationality verification, and rule consistency verification; The topology integrity check is used to determine whether there are broken links, isolated islands, or illegal connections; the parameter rationality check is used to check whether the device parameters meet the rated range constraints; and the rule consistency check is used to verify whether the model conforms to the planning and scheduling rules.

3. The method according to claim 1, characterized in that, The process of collecting the first data and preprocessing the first data includes: Collect first data, which includes at least one of the following: main equipment model data, topology information and connection relationship data, operating status data, equipment parameters, and historical operating and planning data; Add a data traceability tag to all first data, the data traceability tag being used to mark the data source and time of the first data; The first data is cleaned and standardized to remove outliers and missing values; the device parameters are standardized and normalized; and the topology is checked for consistency.

4. The method according to claim 1, characterized in that, The graph structure for constructing the power grid includes: Based on the power grid topology and equipment characteristics, the power grid is abstracted into a graph structure; Map power equipment or connection points to graph nodes, and map lines or topological connections to edges.

5. The method according to claim 1, characterized in that, The process of grouping and clustering the nodes includes: The nodes are clustered using either the K-means clustering algorithm based on node static attributes or the spectral clustering algorithm based on power grid topology similarity. The objective function for clustering is as follows. ； Where min represents the clustering result of the nodes that minimizes the objective function. Indicates the number of cluster categories. Describes the set of nodes of the kth class. This represents the feature center vector of the k-th class. Represents the i-th node of the k-th class, the This represents the original feature vector of the i-th node.

6. The method according to claim 1, characterized in that, The step of inputting the structured semantics into a large language model and obtaining a future-state power grid model through the retrieval enhancement mechanism of the large language model includes: Map node embedding features to topological relationships into structured semantics; The structured semantics are input into a large language model and trained for power grid scenarios; The training includes incremental pre-training, which uses power system planning and design procedures and power grid dispatching and operation procedures as power professional corpora. The training also includes low-rank adaptation fine-tuning, which uses power grid historical simulation cases, topology completion cases and multi-temporal verification cases for fine-tuning. Through the retrieval enhancement mechanism, relevant rules, historical cases, and operational specifications are retrieved from the knowledge base; Based on the trained structured semantics and the relevant rules, historical cases, and operational specifications obtained through the retrieval enhancement mechanism, a future power grid model is derived.

7. A future state prediction system for a power grid, characterized in that, include: The acquisition module is used to acquire first data, preprocess the first data, and obtain a sample dataset; wherein, the first data represents the historical model data and real-time operation data required for the future state projection of the target power grid; A construction module is used to construct a graph structure of the power grid based on the sample dataset, the graph structure including nodes and edges; The clustering module is used to group and cluster the nodes, and assign a category label to each node to obtain the clustered nodes; The extraction module is used to extract node embedding features based on the clustered nodes by performing multi-layer feature propagation on the graph structure through a graph neural network model; the node embedding features are used to reflect topological dependencies, device coupling, and operational semantics. The mapping module is used to map the power grid topology and the node embedding features into structured semantics; The acquisition module is used to input the structured semantics into a large language model and obtain the future power grid model through the retrieval enhancement mechanism of the large language model.

8. A computing device, characterized in that, Its features include: Memory, used to store program instructions; A processor is configured to invoke program instructions stored in the memory and execute the method as described in any one of claims 1-6 according to the obtained program instructions.

9. A computer-readable storage medium, characterized in that, Includes computer-readable instructions that, when read and executed by a computer, cause the method as described in any one of claims 1 to 6 to be implemented.

10. A computer program product, characterized in that, It includes a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform the steps of the method according to any one of claims 1 to 6.

Images