Power grid digital twinning scenarized application construction method and operation and maintenance management system

Abstract

The invention discloses a power grid digital twinning scenarized application construction method and an operation and maintenance management system, and the method comprises the steps: collecting power grid state data in real time through a sensor network disposed in a physical power grid, combining a power grid topological structure and equipment physical parameters, and inputting the data into a space-time encoder to generate a dynamic feature vector; constructing an entity relation graph based on the power grid equipment knowledge base; performing space-time alignment on the dynamic feature vector and the entity relationship map, and constructing a multi-source data lake; dynamically dividing application scene categories based on power grid service requirements, and generating an application scene label set; extracting space-time matching data from the data lake through correlation analysis, and generating corresponding twin daughters; and combining the twinborn daughters of each business demand to obtain a power grid digital twinborn scenarized application. The method has the advantages that through real-time data collection, space-time alignment and multi-source data lake construction, the power grid operation efficiency and equipment operation and maintenance management are effectively improved, potential risks are recognized in advance, and the power grid intelligence and self-adaptive capacity are optimized.

Description

Technical Field

[0001] The present invention relates to digital twin technology, and in particular to a scenario-based application construction method and operation and maintenance management system for power grid digital twins. Background Art

[0002] Digital twin technology, which creates virtual models of physical entities or systems to reflect their status, behavior, and performance in real time, has become a crucial tool in modern industry and urban management. The scenario-based application construction method for digital twins combines digital twin technology with specific application scenarios. By collecting and processing data from the physical world and leveraging technologies such as big data, cloud computing, and the Internet of Things, it provides accurate virtual simulation models for various industries.

[0003] Current market methods and systems rely on single data sources or traditional static analysis, lacking real-time dynamic data collection and intelligent spatiotemporal alignment technologies. This results in low accuracy and slow response times when handling complex grid changes, predicting equipment failures, and making operational and maintenance decisions. Furthermore, current systems lack effective entity-relationship graphs and multi-level data lakes, resulting in inefficient data fusion and storage, and an inability to fully mine and utilize historical data for accurate analysis. Furthermore, many current systems fail to effectively integrate dynamic application scenario division and real-time adjustments with grid business needs, resulting in an inability to flexibly respond to changing grid demands in different scenarios. Summary of the Invention

[0004] In order to improve existing methods and systems, a scenario-based application construction method and operation and maintenance management system for power grid digital twins are provided. This method effectively improves power grid operation efficiency and equipment operation and maintenance management through real-time data collection, spatiotemporal alignment, multi-source data lake construction and intelligent analysis, identifies potential risks in advance, and optimizes the intelligence and adaptability of the power grid.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0006] The method for building scenario-based applications of power grid digital twins includes:

[0007] The sensor network deployed on the physical power grid collects grid status data in real time, combines the grid topology and equipment physical parameters, and inputs them into a spatiotemporal encoder to generate dynamic feature vectors.

[0008] Building an entity relationship graph based on the power grid equipment knowledge base, the graph associates historical equipment failure modes with preset operation and maintenance rules;

[0009] Align dynamic feature vectors with entity relationship graphs in time and space to build a multi-source data lake;

[0010] Dynamically classify application scenarios based on power grid business needs and generate application scenario tag sets;

[0011] Based on the application scenario tag set required by power grid business, we extract spatiotemporal matching data from the data lake through correlation analysis and generate corresponding twins;

[0012] Combine the twins of various business needs to obtain scenario-based applications of power grid digital twins.

[0013] Preferably, the real-time collection of grid status data by a sensor network deployed in the physical grid, combined with the grid topology and physical parameters of the equipment, and input into a spatiotemporal encoder to generate a dynamic feature vector specifically includes:

[0014] Real-time acquisition of power grid equipment status data, including temperature, current, voltage, vibration frequency, and node connection relationship data of the power grid system;

[0015] Based on the node connection relationship data of the power grid system, a graph structure is constructed to perform topological dynamic propagation analysis and obtain the topological propagation feature vector;

[0016] Extract physical feature vectors based on the collected power grid equipment status data, and fuse them with the topology propagation feature vectors to obtain a fusion vector;

[0017] The fusion vector is updated in real time by adopting a sliding time window mechanism to generate a dynamic feature vector.

[0018] Preferably, the entity relationship graph is constructed based on the power grid equipment knowledge base, and the graph associates historical equipment failure modes with preset operation and maintenance rules, specifically including:

[0019] Obtain basic device attributes, grid topology connection relationships, and historical operation and maintenance database based on the grid equipment knowledge base;

[0020] Calculate the similarity between real-time data and historical fault characteristics through fault pattern matching, and dynamically bind conditional rules;

[0021] Modeling is performed based on each entity-relationship to generate an entity relationship graph. Entities include equipment entities, fault entities, and operation and maintenance rule entities. Relationships include physical relationships, fault relationships, and rule bindings.

[0022] Preferably, the spatiotemporal alignment of the dynamic feature vector and the entity relationship graph to build a multi-source data lake specifically includes:

[0023] The dynamic feature vectors are timestamp-aligned, and the dynamic feature vectors are fused with the entity relationship graph data through a network neural model, and then spatially and temporally aligned;

[0024] The fused data is stored in a unified format, a data lake is built to store the data, and the data is divided into multiple levels, including the original data layer, the cleaned data layer, and the processed data layer.

[0025] Preferably, the dynamically classifying application scenarios based on power grid business requirements and generating application scenario tag sets specifically includes:

[0026] Based on power grid business, obtain data from all aspects of the power grid and extract the characteristics and requirements of each application scenario;

[0027] Dynamically classify grid application scenarios based on grid business needs and data analysis;

[0028] After classifying the application scenarios, each application scenario is assigned a label that reflects the grid business needs, and the label is updated as the needs change during grid operation.

[0029] Preferably, the application scenario tag set based on the power grid business requirements extracts spatiotemporal matching data from the data lake through correlation analysis to generate corresponding twins, specifically including:

[0030] Based on the application scenario tag set required by power grid business, the association metric is quantitatively calculated according to the spatial association, temporal association, and semantic association of each data item;

[0031] Based on the quantitative calculation results of the correlation metric, obtain the matching data with the highest correlation with each application scenario;

[0032] Based on the matching data of each application scenario, twins are constructed through a five-layer twin architecture.

[0033] Furthermore, a scenario-based operation and maintenance management system for power grid digital twin applications is proposed, including:

[0034] Data perception module: The data perception layer is deployed on smart sensors, drone inspection terminals, and robot inspection terminals of the physical power grid;

[0035] Digital twin module: The digital twin module includes a precision adjustment unit and a virtual-reality interaction verification unit;

[0036] Operation and maintenance application module: The operation and maintenance application module includes an equipment health unit and a risk warning unit;

[0037] Processor: The processor is used to process the calculation process of each formula and the construction calculation process of each model.

[0038] Preferably, the digital twin module specifically includes:

[0039] Precision adjustment unit: The precision adjustment unit monitors the error rate between the twin and the actual power grid in real time. When the error rate exceeds a threshold, it triggers the retraining of the spatiotemporal encoder.

[0040] Virtual-reality interaction verification unit: The virtual-reality interaction verification unit injects fault simulation data into the twin, verifies the operation and maintenance rule response logic, sends the optimization strategy to the physical power grid, and closes the loop to verify the execution effect.

[0041] Preferably, the operation and maintenance application module specifically includes:

[0042] Equipment health unit: The equipment health unit locates the health status of equipment nodes based on the entity relationship graph and generates a degradation trend curve by matching historical failure patterns;

[0043] Risk warning unit: The risk warning unit predicts short-term power grid risks through the LSTM-Transformer hybrid model, associates the warning signals of multiple twins, and generates a cross-scenario risk topology map.

[0044] Compared with the prior art, the advantages of the present invention are:

[0045] By collecting grid status data in real time, combining it with the topology and physical parameters of the equipment, and using a spatiotemporal encoder to generate dynamic feature vectors, the accuracy of grid status monitoring and prediction is effectively improved. By constructing an entity relationship graph based on the knowledge base of power grid equipment, it is possible to associate historical equipment failure modes and preset operation and maintenance rules, providing a more accurate decision-making basis for equipment operation and maintenance. The construction of a multi-source data lake and spatiotemporal alignment technology achieve effective data integration and efficient storage, enhancing the power grid's ability to process big data. By dynamically dividing application scenarios and generating label sets based on power grid business needs, the matching of business needs and the generation of scenario-based applications are further optimized. The system has built-in precision adjustment and virtual-reality interaction verification units to ensure that the accuracy of the twin is highly consistent with the actual state of the power grid. Through risk warnings and equipment health monitoring, it can achieve early identification and response to potential risks in the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A schematic diagram of the method proposed in the present invention;

[0047] Figure 2 This is a schematic diagram of the dynamic feature vector proposed by the present invention;

[0048] Figure 3 This is a schematic diagram of the entity relationship diagram proposed in the present invention;

[0049] Figure 4 Schematic diagram of the multi-source data lake proposed in this invention;

[0050] Figure 5 This is a schematic diagram of the application scenario tag set proposed by the present invention;

[0051] Figure 6 Schematic diagram of the twin proposed by the present invention;

[0052] Figure 7 This is the digital twin module diagram proposed by the present invention;

[0053] Figure 8 This is the operation and maintenance application module diagram proposed by the present invention. DETAILED DESCRIPTION

[0054] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0055] The power grid digital twin scenario-based application operation and maintenance management system includes:

[0056] Data perception module: The data perception layer is deployed on smart sensors, drone inspection terminals, and robot inspection terminals of the physical power grid;

[0057] Digital twin module: The digital twin module includes a precision adjustment unit and a virtual-reality interaction verification unit;

[0058] Operation and maintenance application module: The operation and maintenance application module includes an equipment health unit and a risk warning unit;

[0059] Processor: The processor is used to process the calculation process of each formula and the construction calculation process of each model.

[0060] See Figure 1 As shown in the figure, the scenario-based application construction method of the power grid digital twin includes:

[0061] Step 1: A sensor network deployed on the physical power grid collects grid status data in real time, combines the grid topology and equipment physical parameters, and inputs them into a spatiotemporal encoder to generate a dynamic feature vector.

[0062] Step 2: Build an entity relationship graph based on the power grid equipment knowledge base, which associates historical equipment failure modes with preset operation and maintenance rules;

[0063] Step 3: Align the dynamic feature vectors with the entity relationship graph in time and space to build a multi-source data lake;

[0064] Step 4: Dynamically classify application scenarios based on power grid business needs and generate application scenario label sets;

[0065] Step 5: Based on the application scenario tag set required by the power grid business, extract spatiotemporal matching data from the data lake through correlation analysis to generate corresponding twins;

[0066] Step 6: Combine the twins of various business needs to obtain scenario-based applications of power grid digital twins.

[0067] See Figure 2 As shown in the figure, the sensor network deployed on the physical power grid collects grid status data in real time, combines the grid topology and equipment physical parameters, and inputs the spatiotemporal encoder to generate dynamic feature vectors, including:

[0068] Real-time acquisition of power grid equipment status data, including temperature, current, voltage, vibration frequency, and node connection relationship data of the power grid system;

[0069] Based on the node connection relationship data of the power grid system, a graph structure is constructed to perform topological dynamic propagation analysis and obtain the topological propagation feature vector;

[0070] Extract physical feature vectors based on the collected power grid equipment status data, and fuse them with the topology propagation feature vectors to obtain a fusion vector;

[0071] The fusion vector is updated in real time by adopting a sliding time window mechanism to generate a dynamic feature vector.

[0072] Specifically, the node connection relationship data of the power grid system can be represented by a topological structure as a graph G = (V, E), where V is the set of all devices in the power grid and E is the connection relationship between devices;

[0073] For the topological propagation of the power grid, a graph convolutional network can be used to model the mutual influence between nodes. The eigenvector of the topological propagation can be calculated by the following formula:

[0074]

[0075] in, is the set of neighbor nodes of node i, is the adjacency matrix element between node i and node j, is the propagation feature vector of node i at time t, is the weight matrix, is the activation function;

[0076] The extracted physical feature vectors are fused with the topological propagation feature vectors by simple splicing or weighted averaging. Based on the fused feature vectors updated in the sliding time window, a dynamic feature vector is generated for each grid node for tasks such as health monitoring and fault prediction of the grid system.

[0077] See Figure 3 As shown, an entity relationship graph is constructed based on the power grid equipment knowledge base. The graph associates historical equipment failure modes with preset operation and maintenance rules, specifically including:

[0078] Obtain basic device attributes, grid topology connection relationships, and historical operation and maintenance database based on the grid equipment knowledge base;

[0079] Calculate the similarity between real-time data and historical fault characteristics through fault pattern matching, and dynamically bind conditional rules;

[0080] Modeling is performed based on each entity-relationship to generate an entity relationship graph. Entities include equipment entities, fault entities, and operation and maintenance rule entities. Relationships include physical relationships, fault relationships, and rule bindings.

[0081] Specifically, basic attribute data of equipment, grid topology connection relationship data and historical operation and maintenance database data are extracted from the grid equipment knowledge base;

[0082] Calculate the similarity between the real-time state vector of the current device and the historical fault feature vector, and construct an entity relationship graph, forming entities and relationships into triples;

[0083] Entity types include equipment entities, such as transformers and circuit breakers, fault entities, such as overheating and short circuit, and rule entities, such as "high temperature + current fluctuation -> maintenance";

[0084] Relationship types include physical relationships, which are derived from topological connection relationships; fault association relationships, which indicate that a device has experienced or may experience a fault; and rule binding relationships, which indicate that a rule is currently activated for a device.

[0085] See Figure 4 As shown in the figure, the dynamic feature vectors are spatiotemporally aligned with the entity relationship graph to build a multi-source data lake. Specifically, the following steps are involved:

[0086] The dynamic feature vectors are timestamp-aligned, and the dynamic feature vectors are fused with the entity relationship graph data through a network neural model, and then spatially and temporally aligned;

[0087] The fused data is stored in a unified format, a data lake is built to store the data, and the data is divided into multiple levels, including the original data layer, the cleaned data layer, and the processed data layer.

[0088] Specifically, for data with different timestamps, time alignment technology is needed to ensure that the feature vectors of each device are aligned at the same timestamp;

[0089] The dynamic feature vector is fused with the entity relationships in the graph through the graph neural network model. For each device entity, in each layer of the graph neural network, it is combined with the feature information of its neighboring devices to form a new fused feature.

[0090] Graph neural network models not only need to consider the topological relationships between nodes, but also combine time information for spatiotemporal alignment. For each timestamp, time information is added to the input features of the graph neural network, allowing the model to learn the temporal dynamics of the device.

[0091] Data storage can be divided into three levels. The raw data layer stores raw data collected directly from power grid equipment and topology, including sensor data such as temperature, current, voltage, and vibration frequency, as well as equipment status information. The cleaned data layer performs pre-processing on the raw data, including denoising, gap filling, and outlier handling. The cleaned data is assumed to be the result of processing the raw data. The processed data layer contains data processed through feature extraction and feature fusion, suitable for subsequent analysis and modeling. This layer includes fused dynamic feature vectors and graph fusion data.

[0092] See Figure 5 As shown in the figure, based on the power grid business needs, the application scenario categories are dynamically divided and the generated application scenario tag set specifically includes:

[0093] Based on power grid business, obtain data from all aspects of the power grid and extract the characteristics and requirements of each application scenario;

[0094] Dynamically classify grid application scenarios based on grid business needs and data analysis;

[0095] After classifying the application scenarios, each application scenario is assigned a label that reflects the grid business needs, and the label is updated as the needs change during grid operation.

[0096] Specifically, based on the operation of the power grid, multiple application scenarios are defined, such as:

[0097] Equipment monitoring scenario: used to monitor the health status of equipment and detect faults in a timely manner;

[0098] Load forecasting scenario: used to predict future power demand based on historical load data;

[0099] Power grid dispatching scenario: Optimize power flow and conduct dispatching based on power grid load conditions;

[0100] Emergency response scenario: Automatically activate the emergency plan when a power grid failure occurs;

[0101] By analyzing data from all aspects of the power grid, extracting the characteristics of each application scenario, and using machine learning or clustering algorithms to analyze the historical data of the power grid, the patterns in power grid operation are automatically identified, and the data is divided into different application scenarios according to different needs. As the needs change during the power grid operation process, the scenario categories are dynamically adjusted.

[0102] See Figure 6 As shown in the figure, based on the application scenario tag set of power grid business requirements, spatiotemporal matching data is extracted from the data lake through correlation analysis to generate the corresponding twins, specifically including:

[0103] Based on the application scenario tag set required by power grid business, the association metric is quantitatively calculated according to the spatial association, temporal association, and semantic association of each data item;

[0104] Based on the quantitative calculation results of the correlation metric, obtain the matching data with the highest correlation with each application scenario;

[0105] Based on the matching data of each application scenario, twins are constructed through a five-layer twin architecture.

[0106] Specifically, spatial correlation measures the similarity between different devices or different areas. It usually considers the grid topology and the physical distance between devices and uses Euclidean distance, Manhattan distance, etc. to measure spatial similarity;

[0107] Time correlation reflects the changing trend of equipment or load at different time points and is calculated through time series analysis;

[0108] For each application scenario label, calculate the relevance of all data items to the scenario and select the data item with the highest relevance;

[0109] The five-layer twin architecture usually includes the following five layers:

[0110] Physical layer: physical entities of the power grid, such as equipment, lines, and transformers;

[0111] Perception layer: obtains real-time data of equipment and power grid through sensors and monitoring devices;

[0112] Data layer: data storage, cleaning, processing, and generating feature vectors that can be used for analysis;

[0113] Modeling layer: In this layer, a machine learning model is used to build a twin model of the device to simulate the status and behavior of the device;

[0114] Application layer: Based on the output of the model, business decisions such as power grid optimization scheduling and fault diagnosis are made.

[0115] See Figure 7 As shown in the figure, the digital twin module specifically includes:

[0116] Precision adjustment unit: The precision adjustment unit monitors the error rate between the twin and the actual power grid in real time. When the error rate exceeds a threshold, it triggers the retraining of the spatiotemporal encoder.

[0117] Virtual-reality interaction verification unit: The virtual-reality interaction verification unit injects fault simulation data into the twin, verifies the operation and maintenance rule response logic, sends the optimization strategy to the physical power grid, and closes the loop to verify the execution effect.

[0118] Specifically, the error rate is a measure of the difference between the state of the twin and the actual power grid at a certain moment. The output of the twin at time t is , and the actual state of the power grid is , then the error rate can be expressed as:

[0119]

[0120] in, is the error rate;

[0121] The purpose of the spatiotemporal encoder is to generate more accurate twin model outputs by processing the spatiotemporal data of the power grid;

[0122] The virtual-reality interaction verification unit first simulates a fault scenario to test the twin's response capability under abnormal circumstances. After receiving the fault data, the twin will respond according to the predetermined operation and maintenance rules. During this process, the virtual-reality interaction verification unit evaluates the twin's operation and maintenance rule response logic to ensure its rationality and accuracy.

[0123] Once the operation and maintenance rules of the twin are successfully verified, the virtual-reality interaction verification unit will send the optimization strategy to the physical power grid. After the optimization strategy is sent to the physical power grid and executed, the virtual-reality interaction verification unit will monitor the operating status of the physical power grid in real time and compare it with the predicted status of the twin.

[0124] See Figure 8 As shown in the figure, the operation and maintenance application module specifically includes:

[0125] Equipment health unit: The equipment health unit locates the health status of equipment nodes based on the entity relationship graph and generates a degradation trend curve by matching historical failure patterns;

[0126] Risk warning unit: The risk warning unit predicts short-term power grid risks through the LSTM-Transformer hybrid model, associates the warning signals of multiple twins, and generates a cross-scenario risk topology map.

[0127] Specifically, by matching historical failure patterns, we can generate equipment degradation trends and predict future equipment degradation paths. Degradation trends describe the health change trends of equipment. Typically, linear regression, exponential decay models, or machine learning models can be used to fit degradation curves. Based on the degradation trend curves, the equipment health unit can determine whether the equipment needs repair or replacement.

[0128] The risk warning unit not only considers risk predictions for individual devices but also correlates warning signals from multiple twins to comprehensively predict the overall risk of the power grid system. By correlating warning signals from multiple twins, it generates a cross-scenario risk topology map. The risk topology map is a graphical representation in which nodes represent devices or areas in the power grid, and edges represent dependencies or risk propagation paths between them.

[0129] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0130] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0131] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A scenario-based application construction method for power grid digital twins, characterized by: include: The sensor network deployed on the physical power grid collects grid status data in real time, combines the grid topology and equipment physical parameters, and inputs them into a spatiotemporal encoder to generate dynamic feature vectors. Building an entity relationship graph based on the power grid equipment knowledge base, the graph associates historical equipment failure modes with preset operation and maintenance rules; Align dynamic feature vectors with entity relationship graphs in time and space to build a multi-source data lake; Dynamically classify application scenarios based on power grid business needs and generate application scenario tag sets; Based on the application scenario tag set required by power grid business, we extract spatiotemporal matching data from the data lake through correlation analysis and generate corresponding twins; Combine the twins of various business needs to obtain scenario-based applications of power grid digital twins.

2. The method for constructing scenario-based applications of power grid digital twins according to claim 1, characterized in that: The method of collecting grid status data in real time through a sensor network deployed on the physical grid, combining the grid topology and physical parameters of the equipment, and inputting the data into a spatiotemporal encoder to generate a dynamic feature vector specifically includes: Real-time acquisition of power grid equipment status data, including temperature, current, voltage, vibration frequency, and node connection relationship data of the power grid system; Based on the node connection relationship data of the power grid system, a graph structure is constructed to perform topological dynamic propagation analysis and obtain the topological propagation feature vector; Extract physical feature vectors based on the collected power grid equipment status data, and fuse them with the topology propagation feature vectors to obtain a fusion vector; The fusion vector is updated in real time by adopting a sliding time window mechanism to generate a dynamic feature vector.

3. The method for constructing scenario-based applications of power grid digital twins according to claim 1, characterized in that: The entity relationship graph is constructed based on the power grid equipment knowledge base, and the graph associates historical equipment failure modes with preset operation and maintenance rules, specifically including: Obtain basic device attributes, grid topology connection relationships, and historical operation and maintenance database based on the grid equipment knowledge base; Calculate the similarity between real-time data and historical fault characteristics through fault pattern matching, and dynamically bind conditional rules; Modeling is performed based on each entity-relationship to generate an entity relationship graph. Entities include equipment entities, fault entities, and operation and maintenance rule entities. Relationships include physical relationships, fault relationships, and rule bindings.

4. The method for constructing scenario-based applications of power grid digital twins according to claim 1, characterized in that: The spatiotemporal alignment of dynamic feature vectors with entity relationship graphs to build a multi-source data lake specifically includes: The dynamic feature vectors are timestamp-aligned, and the dynamic feature vectors are fused with the entity relationship graph data through a network neural model, and then spatially and temporally aligned; The fused data is stored in a unified format, a data lake is built to store the data, and the data is divided into multiple levels, including the original data layer, the cleaned data layer, and the processed data layer.

5. The method for constructing scenario-based applications of power grid digital twins according to claim 1, characterized in that: Dynamically classifying application scenarios based on power grid business requirements and generating application scenario tag sets specifically includes: Based on power grid business, obtain data from all aspects of the power grid and extract the characteristics and requirements of each application scenario; Dynamically classify grid application scenarios based on grid business needs and data analysis; After classifying the application scenarios, each application scenario is assigned a label that reflects the grid business needs, and the label is updated as the needs change during grid operation.

6. The method for constructing scenario-based applications of power grid digital twins according to claim 1, characterized in that: The application scenario tag set based on power grid business requirements extracts spatiotemporal matching data from the data lake through correlation analysis to generate corresponding twins, specifically including: Based on the application scenario tag set required by power grid business, the association metric is quantitatively calculated according to the spatial association, temporal association, and semantic association of each data item; Based on the quantitative calculation results of the correlation metric, obtain the matching data with the highest correlation with each application scenario; Based on the matching data of each application scenario, twins are constructed through a five-layer twin architecture.

7. A power grid digital twin scenario-based application operation and maintenance management system, configured to implement the power grid digital twin scenario-based application construction method according to any one of claims 1 to 6, characterized in that: include: Data perception module: The data perception layer is deployed on smart sensors, drone inspection terminals, and robot inspection terminals of the physical power grid; Digital twin module: The digital twin module includes a precision adjustment unit and a virtual-reality interaction verification unit; Operation and maintenance application module: The operation and maintenance application module includes an equipment health unit and a risk warning unit; Processor: The processor is used to process the calculation process of each formula and the construction calculation process of each model.

8. The power grid digital twin scenario-based application operation and maintenance management system according to claim 7, characterized in that: The digital twin module specifically includes: Precision adjustment unit: The precision adjustment unit monitors the error rate between the twin and the actual power grid in real time. When the error rate exceeds a threshold, it triggers the retraining of the spatiotemporal encoder. Virtual-reality interaction verification unit: The virtual-reality interaction verification unit injects fault simulation data into the twin, verifies the operation and maintenance rule response logic, sends the optimization strategy to the physical power grid, and closes the loop to verify the execution effect.

9. The power grid digital twin scenario-based application operation and maintenance management system according to claim 7, characterized in that: The operation and maintenance application module specifically includes: Equipment health unit: The equipment health unit locates the health status of equipment nodes based on the entity relationship graph and generates a degradation trend curve by matching historical failure patterns; Risk warning unit: The risk warning unit predicts short-term power grid risks through the LSTM-Transformer hybrid model, associates the warning signals of multiple twins, and generates a cross-scenario risk topology map.

Images