Power grid security situation awareness platform and method
By using the dynamic fault propagation factor algorithm, combined with spatiotemporal graph convolutional networks and entropy quantization, a power grid security situation awareness platform is constructed. This solves the problem of insufficient latent fault perception in existing technologies, enables early identification and risk assessment of latent power grid faults, and improves the decision-making efficiency and accuracy of power grid security control.
Patent Information
- Application Number
- CN202511651486.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-12
- Publication Date
- 2026-02-10
AI Technical Summary
Existing power grid safety monitoring systems are unable to effectively detect latent faults that have not reached alarm thresholds, such as early equipment degradation, weak oscillation propagation, and latent faults induced by environmental factors, resulting in insufficient early warning response.
By employing the dynamic fault propagation factor algorithm, and through data acquisition, spatiotemporal alignment, fault feature extraction, spatiotemporal graph convolutional network, and fault propagation path entropy quantification, a power grid safety situation awareness platform is constructed to achieve early and accurate capture and risk assessment of latent faults.
It enables early and accurate detection of latent faults in the power grid, provides a longer early warning and response window, improves the decision-making efficiency of power grid security control and the reliability of risk assessment, and reduces decision-making delays and misjudgments.
Smart Images
Figure CN121507730A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of power grid sensing technology, and in particular relates to a power grid security situation sensing platform and method. Background Technology
[0002] With the rapid growth of power generation from new energy sources such as wind and solar power, the era of smart energy and the Internet of Things (IoT) for electricity is rapidly approaching. The energy system is transforming into an era of fragmented energy, where fragmented energy will exist in a highly intelligent, interconnected form, maximizing its value. The IoT environment is a network environment for collecting, storing, analyzing, computing, and sharing big data. It is a massive, nonlinear, and complex system. Its complexity is mainly manifested in the huge number of nodes, node diversity, connection diversity, information diversity, dynamic complexity, complex and variable network structure, and the fusion of multiple complexities. Therefore, the network in the IoT environment faces more security risks.
[0003] Existing power grid safety monitoring systems mainly rely on threshold alarm mechanisms based on SCADA / PMU measurement data. This method can only respond to explicit faults that have exceeded preset thresholds (such as voltage over-limits and sudden current surges), but it lacks effective detection capabilities for implicit faults that have not reached alarm thresholds (such as early equipment degradation, weak oscillation propagation, and latent faults induced by environmental factors). To address these issues, the following solutions are proposed. Summary of the Invention
[0004] The purpose of this invention is to provide a power grid security situation awareness platform and method. Through the dynamic fault propagation factor algorithm, it achieves relatively accurate early detection of latent faults in the power grid, solving the problem that existing methods can only respond to obvious faults that have exceeded the preset threshold, while lacking effective detection capabilities for latent faults that have not reached the alarm threshold.
[0005] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:
[0006] This invention relates to a power grid security situation awareness method, which includes the following steps:
[0007] Step S1, Data Acquisition and Spatiotemporal Alignment: Acquire power grid topology, real-time measurement and meteorological data, and construct a spatiotemporally aligned feature matrix;
[0008] Step S2, Fault Feature Extraction: Calculate the node power anomaly degree and line meteorological vulnerability factor as dynamic features based on the spatiotemporally aligned feature matrix;
[0009] Step S3: Construct a spatiotemporal graph convolutional network: Utilize dynamic features to construct a spatiotemporal graph neural network by designing spatiotemporal convolutional kernels, and extract node association features hierarchically through the spatiotemporal graph neural network;
[0010] Step S4, Fault Propagation Factor Calculation: Calculate the fault penetration rate through node association features, and calculate the dynamic fault propagation factor based on the gradient field defined by the fault penetration rate.
[0011] Step S5, Fault Propagation Path Entropy Quantification: Construct a fault propagation probability matrix through dynamic fault propagation factors, calculate the path selection entropy value based on the fault propagation probability matrix, and generate a risk heat map;
[0012] Step S6, Situational Awareness Output: Overlay the risk heat map onto the GIS system to dynamically render the fault propagation path.
[0013] Preferably, step S1 includes:
[0014] S11: Collect power grid topology data ,in, For a set of nodes, Let be a set of edges, each edge containing electrical parameters. ;
[0015] S12: Synchronously acquire SCADA system measurement data ,in, For timestamps, Number the nodes. For nodes exist Voltage amplitude at time 10:00 For the line arrive exist Current amplitude at time , The lines are respectively arrive exist Active or reactive power at any given moment;
[0016] S13: Access meteorological data And it is matched with the geographical coordinates of the power grid equipment, where, For device geolocation tags;
[0017] Step S14: Construct the spatiotemporal alignment matrix: ;
[0018] In the formula, For the spatiotemporal alignment matrix, The total dimension of the features. For topological characteristic submatrices, This is a geospatial feature submatrix.
[0019] Preferably, step S2 includes:
[0020] S21: Based on spatiotemporally aligned power grid data, generate a node anomaly degree that comprehensively reflects abrupt changes in electrical state, used to capture dynamic distortions in node operating states. ,
[0021] In the formula, For nodes At any moment Anomaly at any given moment For nodes Power change gradient, The standard deviation of power across the entire network. The node's rated voltage, These are the upper and lower limits of the voltage safety limit, respectively.
[0022] S22: For transmission lines, vulnerability factors are calculated by combining real-time meteorological impacts to assess the intensity of line fault risk. ,
[0023] In the formula, Let i be the vulnerability factor of line i to j at time t. The rated current of the line. This is the short-circuit current threshold of the line. This is the meteorological impact coefficient. Let be the Euclidean norm of the meteorological vector at the location of the line.
[0024] Preferably, step S3 includes:
[0025] S31: Construct a graph node feature matrix based on the node anomaly degree and line vulnerability factor extracted in step S2. As network input;
[0026] S32: Design cascaded spatiotemporal convolution kernels to capture the dynamic evolution of node states: , ,
[0027] In the formula, For the first Layer network output, It is a linear rectified activation function. For convolution operations, For spatial convolution kernels, For temporal convolution kernels, For trainable parameter vectors, For diagonal matrix operators, Given an adjacency matrix with self-loops, for The degree matrix;
[0028] S33: Set up a three-layer ST-GCN, with each layer having an output dimension of 64.
[0029] Preferably, step S4 includes:
[0030] S41: Based on the node feature vectors output by the third layer of the spatiotemporal graph neural network, calculate the fault penetration rate between nodes, representing the probability strength of a fault propagating from node i to node j. ,
[0031] In the formula, Let i be the fault penetration rate from node i to j. For the Sigmoid function, For trainable weight matrix, This is a vector concatenation operation. For node i, the third layer ST-GCN output, Let $\frac{i}{j}$ be the time delay for the fault to propagate from $i$ to $j$.
[0032] S42: Defines the fault gradient field, reflecting electromagnetic transient characteristics. ,
[0033] In the formula, Let be the fault gradient field vector. For the second-order mixed partial derivative of fault penetration rate with respect to node coordinates, This represents the topological connection direction vector;
[0034] S43: Generate dynamic propagation factors to quantify fault propagation strength, path stability, and node criticality. ,
[0035] In the formula, As a dynamic fault propagation factor, Let j be the degree of node j. It is the sum of the degrees of all nodes in the network.
[0036] Preferably, step S5 includes:
[0037] S51: Constructing a fault propagation probability matrix using dynamic fault propagation factors: ,
[0038] In the formula, This is the fault propagation probability matrix;
[0039] S52: Calculate the path selection entropy value to quantify the uncertainty of the fault propagation path: ,
[0040] In the formula, Choose an entropy value for the path of node i. Let i be the set of neighbors of node i;
[0041] S53: Calculate node-level risk heat values to generate a global risk propagation heat map of the power grid. ,
[0042] In the formula, Let i be the risk heat value of node i. The maximum DFPF value pointing to node i. Let be the infinite norm of the entropy vector.
[0043] Preferably, step S6 includes:
[0044] S61: Using a heatmap to depict risk propagation Overlay onto a GIS geographic information system;
[0045] S62: Real-time rendering of the entire network risk distribution through chroma gradient mapping rules;
[0046] S63: Based on the propagation probability matrix Dynamically draw the fault propagation path: ,
[0047] In the formula, The most likely path for fault propagation. For the set of all feasible paths, It is a product of path probabilities. This is an operation to find the path with the highest probability.
[0048] S64: Periodically update the global situational awareness interface that integrates heatmaps and propagation paths, and automatically push it to the dispatch center console.
[0049] A power grid security situation awareness platform includes: a multi-source data acquisition module, a dynamic feature engine module, a spatiotemporal graph neural network module, a fault propagation inference module, and a situation visualization module, wherein the multi-source data acquisition module, the dynamic feature engine module, the spatiotemporal graph neural network module, the fault propagation inference module, and the situation visualization module are connected in sequence.
[0050] The multi-source data acquisition module is used to integrate power grid topology data, real-time measurement data from the SCADA system, and meteorological data, and to construct a spatiotemporally aligned feature matrix based on timestamp synchronization and matching with the geographical coordinates of power grid equipment.
[0051] The dynamic feature engine module is used to calculate the node power anomaly degree reflecting the sudden change in node electrical state and the line meteorological vulnerability factor combined with the influence of real-time meteorology based on the spatiotemporally aligned feature matrix, and generate a dynamic feature matrix.
[0052] The spatiotemporal graph neural network module is used to extract spatiotemporal correlation features between power grid nodes by using the dynamic feature matrix as input, constructing a spatiotemporal graph neural network through the design of cascaded spatiotemporal convolution kernels, and extracting spatiotemporal correlation features between power grid nodes in a hierarchical manner.
[0053] The fault propagation inference module is used to calculate the fault penetration rate based on the node association features, obtain the dynamic fault propagation factor by combining the gradient field defined by the fault penetration rate, execute the dynamic fault propagation factor algorithm, construct the fault propagation probability matrix through the dynamic fault propagation factor, and quantify the propagation path selection entropy value based on the matrix to generate a risk heat map.
[0054] The situation visualization module is used to overlay the risk heat map onto the GIS geographic information system, render the risk distribution of the entire network in real time through color gradient mapping rules, dynamically draw the most likely fault propagation path based on the fault propagation probability matrix, periodically update the global situation awareness interface that integrates the heat map and the propagation path, and push the real-time situation information to the dispatch center console.
[0055] The present invention has the following beneficial effects:
[0056] 1. This invention achieves relatively accurate early detection of latent faults in the power grid through the Dynamic Fault Propagation Factor (DFPF) algorithm. This scheme integrates multi-source spatiotemporal data and uses a spatiotemporal graph convolutional network to extract the dynamic coupling characteristics of node anomaly degree and line vulnerability factor. It quantifies the latent correlation influence between equipment through a fault gradient field model. The gradient suppression term in DFPF can effectively identify abnormal fluctuations that have not reached the alarm threshold during transient processes, while the fault penetration rate calculation can reflect the potential propagation trend of faults along the topology. This design can identify weak links before the fault becomes explicit, providing dispatchers with a longer early warning response window.
[0057] 2. This invention transforms traditional binary judgment into probabilistic risk assessment by constructing a propagation probability matrix based on entropy quantification. The path selection entropy value comprehensively considers topological connection strength, real-time electrical status, and environmental interference factors, and can dynamically reflect the uncertainty of fault propagation direction. Combined with the node influence weight term in the dynamic propagation factor, the system can adaptively distinguish the differences in propagation risk between trunk lines and branches. This design can avoid misjudgments caused by neglecting network structure characteristics in traditional methods, and provide a more reliable decision-making basis for preventing cascading failures.
[0058] 3. This invention unifies the spatiotemporal benchmarks of topology, measurement, and meteorological data through the design of a spatiotemporal alignment matrix, and the cascaded structure of spatiotemporal convolutional kernels enables joint extraction of cross-modal features. This design simultaneously completes data association and feature enhancement within a single computational framework, eliminating decision delays caused by data silos. The generation process of dynamic propagation factors is directly embedded in the feature interaction results, avoiding redundant data conversion steps. This design improves the processing efficiency of high-dimensional heterogeneous data and meets the timeliness requirements of real-time situational awareness.
[0059] 4. This invention achieves a visual representation of complex risks through entropy heatmaps and dynamic path deduction; the risk propagation heatmap transforms the abstract probability of fault propagation into a visually intuitive spatial distribution, and the entropy intensity color gradient mechanism identifies key risk areas; the dynamic path drawing algorithm is based on the path search of the maximum propagation probability, which can trace the dominant direction of fault spread; this design can help operators quickly locate the source of systemic risks and their scope of influence, and improve the decision-making efficiency of large power grid safety control.
[0060] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description
[0061] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0062] Figure 1 This is a flowchart illustrating the power grid security situation awareness method of the present invention.
[0063] Figure 2 This is a schematic diagram of the power grid security situation awareness platform of the present invention. Detailed Implementation
[0064] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0065] Please see Figure 1 As shown, the present invention is a power grid security situation awareness method, comprising:
[0066] Step S1, Data Acquisition and Spatiotemporal Alignment: First, collect power grid topology data, including the electrical parameters of all nodes and connecting edges; simultaneously acquire real-time measurement data from the power monitoring system, including node voltage, line current, and power values; also access wind speed, temperature, and humidity data from the meteorological monitoring system and accurately match them with the geographical locations of power grid equipment; finally, integrate the three types of data into a unified spatiotemporal alignment matrix according to timestamps and spatial coordinates, specifically:
[0067] Step S11: Collect power grid topology data ,in, For a set of nodes, Let be a set of edges, each edge containing electrical parameters. ;
[0068] Step S12: Synchronously acquire SCADA system measurement data ,in, For timestamps, Number the nodes. For nodes exist Voltage amplitude at time 10:00 For the line arrive exist Current amplitude at time , The lines are respectively arrive exist Active or reactive power at any given moment;
[0069] Step S13: Access meteorological data And it is matched with the geographical coordinates of the power grid equipment, where, For device geolocation tags;
[0070] Step S14: Construct the spatiotemporal alignment matrix: ;
[0071] In the formula, For the spatiotemporal alignment matrix, The total dimension of the features. For topological characteristic submatrices, This is a geospatial feature submatrix.
[0072] Step S2, Fault Feature Extraction: Based on the spatiotemporal matrix generated in Step S1, calculate the anomaly index for each power node, which integrates the power change gradient and voltage deviation. Simultaneously, calculate the vulnerability factor of the transmission line, combining real-time line current load and meteorological environmental data. The output is a time-stamped set of node anomalies and a set of line vulnerability factors, specifically:
[0073] Step S21: Based on the spatiotemporally aligned power grid data, generate a node anomaly degree that comprehensively reflects abrupt changes in electrical state, used to capture dynamic distortions in node operating states: ,
[0074] In the formula, For nodes At any moment Anomaly at any given moment For nodes Power change gradient, The standard deviation of power across the entire network. The node's rated voltage, These are the upper and lower limits of the voltage safety limit, respectively.
[0075] Step S22: For transmission lines, calculate vulnerability factors based on real-time meteorological impacts to assess the intensity of line fault risk. ,
[0076] In the formula, Let i be the vulnerability factor of line i to j at time t. The rated current of the line. This is the short-circuit current threshold of the line. This is the meteorological impact coefficient. Let be the Euclidean norm of the meteorological vector at the location of the line.
[0077] Step S3: Construct a spatiotemporal graph convolutional network: Using the node anomaly degree and line vulnerability factor output from step S2 as initial features, construct a power grid topology graph structure; design a neural network with both spatial and temporal convolution modules, where spatial convolution processes device connection relationships and temporal convolution analyzes feature temporal changes; extract high-dimensional spatiotemporal features layer by layer through a three-layer network to generate the dynamic state vector of each device node, specifically:
[0078] Step S31: Construct a graph node feature matrix based on the node anomaly degree and line vulnerability factor extracted in step S2. As network input;
[0079] Step S32: Design cascaded spatiotemporal convolution kernels to capture the dynamic evolution of node states: , ,
[0080] In the formula, For the first Layer network output, It is a linear rectified activation function. For convolution operations, For spatial convolution kernels, For temporal convolution kernels, For trainable parameter vectors, For diagonal matrix operators, Given an adjacency matrix with self-loops, for The degree matrix;
[0081] Step S33: Set up a three-layer ST-GCN, with each layer having an output dimension of 64.
[0082] Step S4, Fault Propagation Factor Calculation: Using the dynamic state vector obtained in Step S3, calculate the fault penetration rate between any two connected nodes; construct a fault gradient field model based on the fault penetration rate to quantify the propagation direction and intensity of the fault in the power grid; finally, synthesize a dynamic fault propagation factor, which integrates three elements: fault penetration rate, gradient field suppression effect, and node topology weight, specifically:
[0083] Step S41: Based on the node feature vectors output by the third layer of the spatiotemporal graph neural network, calculate the fault penetration rate between nodes, representing the probability strength of a fault propagating from node i to node j. ,
[0084] In the formula, Let i be the fault penetration rate from node i to j. For the Sigmoid function, For trainable weight matrix, This is a vector concatenation operation. For node i, the third layer ST-GCN output, Let $\frac{i}{j}$ be the time delay for the fault to propagate from $i$ to $j$.
[0085] Step S42: Define the fault gradient field to reflect electromagnetic transient characteristics: ,
[0086] In the formula, Let be the fault gradient field vector. For the second-order mixed partial derivative of fault penetration rate with respect to node coordinates, This represents the topological connection direction vector;
[0087] Step S43: Generate dynamic propagation factors to quantify fault propagation strength, path stability, and node criticality. ,
[0088] In the formula, As a dynamic fault propagation factor, Let j be the degree of node j. It is the sum of the degrees of all nodes in the network.
[0089] Step S5, Entropy Quantification of Fault Propagation Paths: Based on the dynamic fault propagation factor generated in Step S4, construct a network-wide fault propagation probability matrix; calculate the entropy value for each node's propagation path selection, where the entropy value reflects the uncertainty of the fault propagation direction; combine the entropy value with the maximum value of the propagation factor to generate a risk propagation heatmap at the node level, specifically:
[0090] Step S51: Construct the fault propagation probability matrix using the dynamic fault propagation factor: ,
[0091] In the formula, This is the fault propagation probability matrix;
[0092] Step S52: Calculate the path selection entropy value to quantify the uncertainty of the fault propagation path: ,
[0093] In the formula, Choose an entropy value for the path of node i. Let i be the set of neighbors of node i;
[0094] Step S53: Calculate node-level risk heat values to generate a global risk propagation heat map of the power grid. ,
[0095] In the formula, Let i be the risk heat value of node i. The maximum DFPF value pointing to node i. Let be the infinite norm of the entropy vector.
[0096] Step S6, Situation Awareness Output: Map the risk heatmap from Step S5 onto the power grid geographic information system, using color gradients to identify risk levels; dynamically draw the most probable fault propagation path based on the propagation probability matrix; update the global situation map at millisecond frequency and push it to the power grid dispatch console in real time, specifically:
[0097] Step S61: Create a heatmap of risk propagation Overlay onto a GIS geographic information system;
[0098] Step S62: Render the risk distribution across the entire network in real time using chroma gradient mapping rules;
[0099] Step S63: Based on the propagation probability matrix Dynamically draw the fault propagation path: ,
[0100] In the formula, The most likely path for fault propagation. For the set of all feasible paths, It is a product of path probabilities. This is an operation to find the path with the highest probability.
[0101] Step S64: Periodically update the global situational awareness interface that integrates the heat map and propagation path, and automatically push it to the dispatch center console.
[0102] Please see Figure 2 As shown, the present invention is a power grid security situation awareness platform, including a multi-source data acquisition module, a dynamic feature engine module, a spatiotemporal graph neural network module, a fault propagation inference module, and a situation visualization module. The multi-source data acquisition module, the dynamic feature engine module, the spatiotemporal graph neural network module, the fault propagation inference module, and the situation visualization module are sequentially connected in communication.
[0103] The multi-source data acquisition module is used to integrate power grid topology data, real-time measurement data from the SCADA system, and meteorological data, and to construct a spatiotemporally aligned feature matrix based on timestamp synchronization and matching with the geographical coordinates of power grid equipment.
[0104] The dynamic feature engine module is used to calculate the node power anomaly degree reflecting the sudden change in node electrical state and the line meteorological vulnerability factor combined with the influence of real-time meteorology based on the spatiotemporally aligned feature matrix, and generate a dynamic feature matrix.
[0105] The spatiotemporal graph neural network module is used to extract spatiotemporal correlation features between power grid nodes by using the dynamic feature matrix as input, constructing a spatiotemporal graph neural network through the design of cascaded spatiotemporal convolution kernels, and extracting spatiotemporal correlation features between power grid nodes in a hierarchical manner.
[0106] The fault propagation inference module is used to calculate the fault penetration rate based on the node association features, obtain the dynamic fault propagation factor by combining the gradient field defined by the fault penetration rate, execute the dynamic fault propagation factor algorithm, construct the fault propagation probability matrix through the dynamic fault propagation factor, and quantify the propagation path selection entropy value based on the matrix to generate a risk heat map.
[0107] The situation visualization module is used to overlay the risk heat map onto the GIS geographic information system, render the risk distribution of the entire network in real time through color gradient mapping rules, dynamically draw the most likely fault propagation path based on the fault propagation probability matrix, periodically update the global situation awareness interface that integrates the heat map and the propagation path, and push the real-time situation information to the dispatch center console.
[0108] A specific application of this embodiment:
[0109] Background: Fault simulation of a 500kV regional power grid in East China;
[0110] Power grid scale: 38 nodes (including 4 thermal power plants, 12 substations, and 22 load nodes);
[0111] Sampling frequency: 100Hz (data update period 10ms);
[0112] Initial fault: At t=0s, the voltage at node 15 drops sharply (502kV→485kV).
[0113] Meteorological conditions: wind speed 6.5 m / s, ambient temperature 32℃, humidity 70%;
[0114] Step S1: Multi-source data acquisition and alignment:
[0115] Obtain the topological adjacency matrix: ;
[0116] Synchronous measurement data:
[0117] Node 15: ;
[0118] Route 15-16: (Rated 1.2kA);
[0119] Meteorological data mapping: ;
[0120] Construct the feature matrix:
[0121] ;
[0122] Step S2, Dynamic Feature Extraction:
[0123] Calculate the anomaly degree of node 15: ;
[0124] Calculate the vulnerability factor of lines 15-16: ;
[0125] Step S3, Spatiotemporal graph convolution:
[0126] Spatiotemporal convolution process (taking the first layer output as an example): ,
[0127] Node 15, third layer feature output:
[0128] ;
[0129] Step S4: Calculation of dynamic fault propagation factor:
[0130] Calculate the fault penetration rate from 15 to 16: ,
[0131] Calculate the fault gradient field: ,
[0132] Generate DFPF: ,
[0133] Step S5: Quantization of Fault Propagation Path Entropy:
[0134] The propagation probability of node 15 is constructed as shown in the table below:
[0135] Calculate the entropy value of node 15: ,
[0136] Generate a risk heatmap:
[0137] (Normalized value 0.81);
[0138] Step S6, Situation visualization output:
[0139] GIS map display:
[0140] Node 15: Orange (entropy value 0.81);
[0141] Node 16: Yellow (entropy value 0.43);
[0142] Other nodes: blue;
[0143] Predict the main propagation path: ,
[0144] Actual fault development:
[0145] t=3.2s: Line 15-16 overload tripped;
[0146] t=8.5s: Node 8 voltage collapse (consistent with the predicted path).
[0147] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.
[0148] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.
Claims
1. A power grid security situation awareness method, characterized in that, The sensing method includes the following steps: Step S1, Data Acquisition and Spatiotemporal Alignment: Acquire power grid topology, real-time measurement and meteorological data, and construct a spatiotemporally aligned feature matrix; Step S2, Fault Feature Extraction: Calculate the node power anomaly degree and line meteorological vulnerability factor as dynamic features based on the spatiotemporally aligned feature matrix; Step S3: Construct a spatiotemporal graph convolutional network: Utilize dynamic features to construct a spatiotemporal graph neural network by designing spatiotemporal convolutional kernels, and extract node association features hierarchically through the spatiotemporal graph neural network; Step S4, Fault Propagation Factor Calculation: Calculate the fault penetration rate through node association features, and calculate the dynamic fault propagation factor based on the gradient field defined by the fault penetration rate. Step S5, Fault Propagation Path Entropy Quantification: Construct a fault propagation probability matrix through dynamic fault propagation factors, calculate the path selection entropy value based on the fault propagation probability matrix, and generate a risk heat map; Step S6, Situational Awareness Output: Overlay the risk heat map onto the GIS system to dynamically render the fault propagation path.
2. The power grid security situation awareness method according to claim 1, characterized in that, Step S1 includes: S11: Collect power grid topology data ,in, For a set of nodes, Let be a set of edges, each edge containing electrical parameters. ; S12: Synchronously acquire SCADA system measurement data ,in, For timestamps, Number the nodes. For nodes exist Voltage amplitude at time 10:00 For the line arrive exist Current amplitude at time , The lines are respectively arrive exist Active or reactive power at any given moment; S13: Access meteorological data And it is matched with the geographical coordinates of the power grid equipment, where, For device geolocation tags; Step S14: Construct the spatiotemporal alignment matrix: ; In the formula, For the spatiotemporal alignment matrix, The total dimension of the features. For topological characteristic submatrices, This is a geospatial feature submatrix.
3. The power grid security situation awareness method according to claim 1, characterized in that, Step S2 includes: S21: Based on spatiotemporally aligned power grid data, generate a node anomaly degree that comprehensively reflects abrupt changes in electrical state, used to capture dynamic distortions in node operating states. ， In the formula, For nodes At any moment Anomaly at any given moment For nodes Power change gradient, The standard deviation of power across the entire network. The node's rated voltage, These are the upper and lower limits of the voltage safety limit, respectively. S22: For transmission lines, vulnerability factors are calculated by combining real-time meteorological impacts to assess the intensity of line fault risk. ， In the formula, Let i be the vulnerability factor of line i to j at time t. The rated current of the line. This is the short-circuit current threshold of the line. This is the meteorological impact coefficient. Let be the Euclidean norm of the meteorological vector at the location of the line.
4. The power grid security situation awareness method according to claim 1, characterized in that, Step S3 includes: S31: Construct a graph node feature matrix based on the node anomaly degree and line vulnerability factor extracted in step S2. As network input; S32: Design cascaded spatiotemporal convolution kernels to capture the dynamic evolution of node states: ， ， In the formula, For the first Layer network output, It is a linear rectified activation function. For convolution operations, For spatial convolution kernels, For temporal convolution kernels, For trainable parameter vectors, For diagonal matrix operators, Given an adjacency matrix with self-loops, for The degree matrix; S33: Set up a three-layer ST-GCN, with each layer having an output dimension of 64.
5. The power grid security situation awareness method according to claim 1, characterized in that, Step S4 includes: S41: Based on the node feature vectors output by the third layer of the spatiotemporal graph neural network, calculate the fault penetration rate between nodes, representing the probability strength of a fault propagating from node i to node j. ， In the formula, Let i be the fault penetration rate from node i to j. For the Sigmoid function, For trainable weight matrix, This is a vector concatenation operation. For node i, the third layer ST-GCN output, Let $\frac{i}{j}$ be the time delay for the fault to propagate from $i$ to $j$. S42: Defines the fault gradient field, reflecting electromagnetic transient characteristics. ， In the formula, Let be the fault gradient field vector. For the second-order mixed partial derivative of fault penetration rate with respect to node coordinates, This represents the topological connection direction vector; S43: Generate dynamic propagation factors to quantify fault propagation strength, path stability, and node criticality. ， In the formula, As a dynamic fault propagation factor, Let j be the degree of node j. It is the sum of the degrees of all nodes in the network.
6. The power grid security situation awareness method according to claim 1, characterized in that, Step S5 includes: S51: Constructing a fault propagation probability matrix using dynamic fault propagation factors: ， In the formula, This is the fault propagation probability matrix; S52: Calculate the path selection entropy value to quantify the uncertainty of the fault propagation path: ， In the formula, Choose an entropy value for the path of node i. Let i be the set of neighbors of node i; S53: Calculate node-level risk heat values to generate a global risk propagation heat map of the power grid. ， In the formula, Let i be the risk heat value of node i. The maximum DFPF value pointing to node i. It is the infinite norm of the entropy vector.
7. The power grid security situation awareness method according to claim 1, characterized in that, Step S6 includes: S61: Using a heatmap to depict risk propagation Overlay onto a GIS geographic information system; S62: Real-time rendering of the entire network risk distribution through chroma gradient mapping rules; S63: Based on the propagation probability matrix Dynamically draw the fault propagation path: ， In the formula, The most likely path for fault propagation. For the set of all feasible paths, It is a product of path probabilities. This is an operation to find the path with the highest probability. S64: Periodically update the global situational awareness interface that integrates heatmaps and propagation paths, and automatically push it to the dispatch center console.
8. A power grid security situation awareness platform, characterized in that, include: The system comprises a multi-source data acquisition module, a dynamic feature engine module, a spatiotemporal graph neural network module, a fault propagation and deduction module, and a situation visualization module, which are connected sequentially. The multi-source data acquisition module is used to integrate power grid topology data, real-time measurement data from the SCADA system, and meteorological data, and to construct a spatiotemporally aligned feature matrix based on timestamp synchronization and matching with the geographical coordinates of power grid equipment. The dynamic feature engine module is used to calculate the node power anomaly degree reflecting the sudden change in node electrical state and the line meteorological vulnerability factor combined with the influence of real-time meteorology based on the spatiotemporally aligned feature matrix, and generate a dynamic feature matrix. The spatiotemporal graph neural network module is used to extract spatiotemporal correlation features between power grid nodes by using the dynamic feature matrix as input, constructing a spatiotemporal graph neural network through the design of cascaded spatiotemporal convolution kernels, and extracting spatiotemporal correlation features between power grid nodes in a hierarchical manner. The fault propagation inference module is used to calculate the fault penetration rate based on the node association features, obtain the dynamic fault propagation factor by combining the gradient field defined by the fault penetration rate, execute the dynamic fault propagation factor algorithm, construct the fault propagation probability matrix through the dynamic fault propagation factor, and quantify the propagation path selection entropy value based on the matrix to generate a risk heat map. The situation visualization module is used to overlay the risk heat map onto the GIS geographic information system, render the risk distribution of the entire network in real time through color gradient mapping rules, dynamically draw the most likely fault propagation path based on the fault propagation probability matrix, periodically update the global situation awareness interface that integrates the heat map and the propagation path, and push the real-time situation information to the dispatch center console.