A power grid operation state risk-oriented situation awareness method, system, device and medium

Abstract

This invention discloses a situational awareness method, system, device, and medium for power grid operation status risks. The method includes: acquiring multi-source data of power grid nodes; preprocessing the multi-source data to obtain structured node attribute data; constructing a spatiotemporal fusion graph sequence of the power grid based on the structured node attribute data using graph serialization; predicting node power using the spatiotemporal fusion graph sequence to obtain a basic predicted value of node power, and correcting the basic predicted value of node power to obtain a prediction result; calculating line risk indicators based on the prediction result and constructing corresponding time series for risk trend prediction; identifying target lines based on the risk trend prediction result, tracing risk paths to obtain risk propagation chains, and overlaying the risk propagation chains onto an information layer to achieve situational awareness of power grid operation status risks, thereby improving the perceptibility and operability of operation management.

Description

Technical Field

[0001] This invention relates to the field of power system security analysis technology, and in particular to a situational awareness method, system, device and medium for power grid operation status risks. Background Technology

[0002] Currently, the continuous rise in the penetration rate of new energy sources and the increasing complexity of power system structures have significantly enhanced the uncertainty and dynamism of power grid operation. Traditional operation monitoring methods are no longer sufficient to meet the demands for accurate identification of power grid status and risk prevention in complex environments. Existing technologies generally employ physical modeling-based methods for power load forecasting and status assessment, primarily relying on power flow calculations and state estimation to analyze power grid operating parameters, while also incorporating static thresholds or empirical rules for risk assessment. In addition, some existing research attempts to introduce data-driven models, such as time series analysis and neural networks, to predict and model power grid data, or to model power grid topology features using graph structures to assist in operational status analysis.

[0003] However, existing technologies generally have significant shortcomings, including a lack of effective modeling of the spatiotemporal correlation characteristics of the power grid, making it difficult to fully utilize the dynamic changes in node attributes and the coupling characteristics between the network structure, resulting in low accuracy of operational status identification results; in addition, most prediction models are of a single type and fail to fully integrate physical laws and data residual characteristics, resulting in insufficient response capabilities to sudden risks. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention provides a situational awareness method, system, device and medium for power grid operation status risks.

[0005] This invention provides a situational awareness method, system, device, and medium for power grid operation status risks, addressing the problems of low accuracy in power grid operation risk prediction and difficulty in identifying risk evolution paths in existing technologies.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a situational awareness method for power grid operation status risks, comprising: Obtain multi-source data of power grid nodes, preprocess the multi-source data to obtain structured node attribute data; Based on the structured node attribute data, a spatiotemporal fusion graph sequence of the power grid is constructed using graph serialization; The node power is predicted using the spatiotemporal fusion graph sequence of the power grid to obtain the basic predicted value of the node power, and the basic predicted value of the node power is corrected to obtain the prediction result. Based on the prediction results, the route risk index is calculated, and the corresponding time series is constructed to predict the risk trend. Based on the risk trend prediction results, the target line is identified, and the risk path is traced to obtain the risk propagation chain. The risk propagation chain is then superimposed on the information layer to achieve situational awareness of power grid operation status risks.

[0007] As a preferred embodiment of the situational awareness method for power grid operation status risks described in this invention, the construction of the power grid spatiotemporal fusion map sequence includes: Construct a graph structure based on the actual topology of the power grid; For each node in the graph structure, construct a node attribute vector and append the node attribute vector to the corresponding node to obtain the static graph structure; The observation period is divided into multiple consecutive time slices according to a preset time interval; The static graph structure is copied at each time slice, and the attribute vectors of all nodes are updated to generate the corresponding time sequence graph. Arrange all the time series diagrams to obtain the spatiotemporal fusion diagram sequence of the power grid.

[0008] The beneficial effect of this preferred technical solution is that it can integrate topological structure and dynamic attributes to accurately characterize the spatiotemporal evolution of the power grid.

[0009] As a preferred embodiment of the situational awareness method for power grid operation status risks described in this invention, the prediction of node power includes: Based on the power grid topology and node status information corresponding to each moment in the power grid spatiotemporal fusion graph sequence; The power flow calculation method is used to make basic predictions of node power and generate basic predicted values ​​of node power.

[0010] As a preferred embodiment of the situational awareness method for power grid operation status risks described in this invention, the correction of the node power baseline prediction value includes: To obtain the actual node power, a graph neural network is used to perform residual learning on the prediction error between the basic predicted value of node power and the actual node power, so as to obtain the residual predicted value of node power. The prediction results are obtained by adding the basic predicted values ​​of node power to the predicted values ​​of node power residuals node by node.

[0011] As a preferred embodiment of the situational awareness method for power grid operation status risks described in this invention, risk trend prediction includes: Based on the prediction results, the load rate of each transmission line is calculated; Based on the load rate, the over-limit risk index for each line is calculated using predefined risk mapping rules; Record the risk indicators of exceeding limits in chronological order to construct a time series of risk indicators at the line level; By using time-series forecasting models, risk indicators within a specified future time range are predicted based on historical risk indicator time series, thus constructing future risk trends.

[0012] The beneficial effect of this preferred technical solution is that it can dynamically reflect the trend of risk evolution and support early intervention and proactive prevention and control.

[0013] As a preferred embodiment of the situational awareness method for power grid operation status risks described in this invention, the identification of target lines includes: Based on the risk trend prediction results, the transmission lines are judged to obtain the target lines; Locate the position of the target line in the power grid topology to obtain the starting node for path tracing; Starting from the initial node, use a graph search algorithm to find all reachable paths leading to the target path; By combining nodes and edges on reachable paths, a risk propagation chain can be formed.

[0014] The beneficial effect of this preferred technical solution is that it can accurately track the risk transmission path and provide key evidence for precise prevention and control.

[0015] As a preferred embodiment of the situational awareness method for power grid operation status risks described in this invention, the method of overlaying the risk propagation chain onto the information layer includes: Map all nodes and edges in the static graph structure to the geographic information layer according to their corresponding geographic coordinates; Each node and edge is visually encoded based on the corresponding risk index of exceeding the limit. On the geographic information layer, the path indicated by the risk propagation chain is drawn, a risk situation map is generated and displayed in real time.

[0016] Secondly, the present invention provides a situational awareness system for power grid operation status risks, comprising: The preprocessing module is used to acquire multi-source data of power grid nodes, preprocess the multi-source data, and obtain structured node attribute data. The graph sequence construction module is used to construct a spatiotemporal fusion graph sequence of the power grid based on the structured node attribute data and using graph serialization. The prediction module is used to predict the node power using the spatiotemporal fusion graph sequence of the power grid, obtain the basic predicted value of the node power, and correct the basic predicted value of the node power to obtain the prediction result. The risk assessment module is used to calculate the line risk index based on the prediction results and construct the corresponding time series for risk trend prediction. The path awareness module is used to identify target lines based on risk trend prediction results, track risk paths to obtain risk propagation chains, and overlay the risk propagation chains onto the information layer to achieve situational awareness of power grid operation status risks.

[0017] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the situational awareness method for power grid operation status risks.

[0018] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the situational awareness method for power grid operation state risks.

[0019] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention combines fundamental physical model prediction with residual learning modeling to jointly model node power. This joint modeling effectively eliminates the problem of single models failing to adequately capture complex dynamic changes, and also improves the invention's responsiveness to sudden abnormal states and the foresight of risk prediction. By constructing a time series of risk indicators from the prediction results and performing trend analysis, this invention can dynamically reflect the changing trends of risk indicators and identify the direction of risk evolution in advance, supporting early intervention and proactive control. This invention achieves source tracing of high-risk lines based on a graph structure. By constructing a risk propagation chain, this invention clarifies the propagation path of risk from the starting node to high-risk lines, providing key path basis for subsequent precise prevention and control. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. 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.

[0021] Figure 1 This is a schematic diagram of the overall process logic of a situational awareness method for power grid operation status risks provided in one embodiment of the present invention. Detailed Implementation

[0022] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0023] Example 1, referring to Figure 1 As an embodiment of the present invention, a situational awareness method for power grid operation status risks is provided, comprising: S100: Acquire multi-source data of power grid nodes, preprocess the multi-source data, and obtain structured node attribute data; In this embodiment of the invention, the multi-source data includes real-time load data, new energy power output, meteorological information, topology, and historical disturbance events.

[0024] Real-time load data is collected from all load nodes in the power grid, with a 15-minute time interval. The data is recorded as nodes. In time The active power load value is expressed in MW. The renewable energy power output covers renewable energy nodes such as photovoltaic and wind power. The sampling period is 15 minutes, and the data record is time. node The actual grid-connected power, in MW. The topology is the obtained current power grid topology, containing a set of nodes. With edge set ,node This includes load nodes, generation nodes, substations, etc.; edges This represents a power transmission line, including electrical properties such as impedance and capacitance. The topology should be a directed graph, recording the power flow direction between nodes. Meteorological information is collected according to the geographical location of the nodes, including wind speed (m / s), temperature (°C), and solar irradiance (W / m²). All information corresponds one-to-one with nodes and is organized by time. Maintain synchronization. The scope of historical disturbance event records includes, but is not limited to, tripping, frequency disturbances, and voltage fluctuations; each record contains information such as the event occurrence point, timestamp, event type, and duration.

[0025] Preprocessing includes consistency verification of power data, employing a dual mechanism: relying on power flow consistency analysis based on grid physical constraints and combining machine learning models to construct prediction baselines and perform residual discrimination.

[0026] Specifically, select the power-related data that needs to be verified, including real-time load data of nodes and power output of new energy sources; For each node Using the LSTM prediction model, based on past The predicted power value is obtained by using historical data from each moment to predict the power value at the current time point. The residual is obtained by calculating the difference between the actual observed value and the predicted power value. , is represented as , ; in, For time The observed values, For time The predicted value; Based on the historical error distribution of the nodes, the upper bound of the 95% confidence interval of their residuals is calculated as the threshold. Node data whose residual exceeds the confidence threshold is considered to be in time... There are outlier data points; After removing outlier data points, structured node attribute data is generated.

[0027] Data points identified as anomalous are removed based on the actual application scenario, either by being removed directly from the sequence or by linear interpolation correction using adjacent normal points. After anomaly handling, a new valid data sequence, i.e., structured node attribute data, is generated for subsequent calculations and modeling.

[0028] It should be noted that after the present invention completes the verification, it can obtain data with stronger consistency and higher robustness, providing a data foundation for the subsequent construction of the spatiotemporal fusion map of the power grid.

[0029] S200: Based on structured node attribute data, a spatiotemporal fusion graph sequence of the power grid is constructed using graph serialization; S300: Node power is predicted using the spatiotemporal fusion graph sequence of the power grid to obtain the basic predicted value of node power, and the basic predicted value of node power is corrected to obtain the prediction result; S400: Calculate the line risk index based on the prediction results and construct the corresponding time series to predict the risk trend; S500: Based on the risk trend prediction results, the target line is identified and the risk path is traced to obtain the risk propagation chain. The risk propagation chain is then overlaid on the information layer to achieve situational awareness of power grid operation status risks.

[0030] It should be noted that this invention ensures data quality through multi-source data preprocessing and utilizes graph serialization to construct a spatiotemporal fusion graph sequence to integrate topological and dynamic attributes. This invention employs a combination of physical models and deep learning for power prediction to improve prediction accuracy, and uses time series data of risk indicators to achieve trend prediction to support early prevention and control. This invention achieves precise source tracing through risk propagation chain tracking and visualization, improving the accuracy, foresight, and operability of power grid operation risk situation awareness.

[0031] In this embodiment of the invention, step S200 includes the following sub-steps A1-A5; In A1: Construct a graph structure based on the actual topology of the power grid; In A2: Construct a node attribute vector for each node in the graph structure, and append the node attribute vector to the corresponding node to obtain the static graph structure; In A3: The observation period is divided into multiple consecutive time slices according to a preset time interval; In A4: the static graph structure is copied at each time slice, and the attribute vectors of all nodes are updated to generate the corresponding time sequence graph; In A5: Arrange all time series diagrams to obtain the power grid spatiotemporal fusion diagram sequence.

[0032] In this embodiment of the invention, a graph structure is constructed based on the actual topology of the power grid as follows: in, It is a set of nodes, including all load nodes, generation nodes, substations, etc. Let be the edge set, representing all transmission lines and their connections; After constructing the graph structure, state and structural attributes are assigned to each node, and a node attribute vector is constructed. , is represented as: in, For nodes In time Active load; For nodes In time The output of new energy sources; For nodes In time Meteorological information, including wind speed, temperature, and irradiance; For nodes Electrical centrality; Electrical centrality is calculated using the reciprocal shortest path method, expressed as: in, Indicates the total number of nodes; Represents a node With nodes The shortest path length between the edges in the graph is calculated in units of the number of edges connected in the graph.

[0033] In the node attribute vector It reflects the degree of connectivity between a node and other nodes in the entire network. The larger the value, the shorter the average distance between the node and other nodes, and the more critical it is in the topology.

[0034] The attribute information from the node attribute vector is appended to each node in the node set to obtain the static graph structure. ; The observation period is divided into multiple consecutive time slices by setting the time span and sampling interval, with each time slice corresponding to a time point. At each time point, the initial static graph structure is copied, keeping the node set and edge set unchanged, and the attribute vectors of all nodes are updated so that the node attributes reflect the real-time state at the corresponding moment.

[0035] The node attribute vector update operation uses structured time series data, reassigning values ​​to all nodes at each time point, including the current active load, renewable energy output, meteorological information, and electrical centrality. Active load, renewable energy output, and meteorological information are collected in real-time using historical data, while electrical centrality is recalculated based on the current network structure and node status to reflect the dynamic changes in the importance of nodes to grid stability at different times.

[0036] The attribute-bearing graph generated at each time point is used as a time series graph. All time series graphs are arranged in chronological order to form a sequence of spatiotemporal fusion graphs of the power grid.

[0037] For example, if the time span is 30 minutes and the sampling interval is 5 minutes, the spatiotemporal fusion graph sequence of the power grid consists of 6 attribute graphs in sequence, which correspond to the network evolution process from the start time to the sixth time point.

[0038] It should be noted that the spatiotemporal fusion diagram sequence of the power grid fully preserves the structural characteristics and operating status information of the power grid, and has the ability to simultaneously capture the spatial structure and temporal dynamics of the power grid, providing continuous input for subsequent node power prediction and operational status risk assessment.

[0039] In this embodiment of the invention, step S300 includes the following sub-steps B1-B2; In B1: Based on the power grid topology and node status information corresponding to each moment in the power grid spatiotemporal fusion graph sequence; In B2: The power flow calculation method is used to make a basic prediction of the node power and generate the basic predicted value of the node power.

[0040] In one optional embodiment, the power flow calculation method can be an AC power flow calculation model. Based on the power grid topology, a set of nonlinear equations for node voltage amplitude and phase angle is constructed. Considering multi-dimensional electrical quantities such as active power, reactive power, and voltage amplitude, the Newton-Raphson method or fast decoupling method is used for iterative calculation to solve for the voltage amplitude and phase angle of each node. Based on the solution results, the active power and reactive power injected into each node are calculated to generate the basic predicted value of node power. In another optional embodiment, the power flow calculation method can also be an optimal power flow calculation model, which aims to minimize power generation cost, minimize network loss, or optimize voltage quality. It establishes a multi-objective optimization function, sets constraints, and uses optimization algorithms such as interior point method or genetic algorithm to solve the problem, thereby obtaining the optimal power allocation scheme for each node. The optimized active power injection of the node is used as the basic predicted value of the node power. In this embodiment of the invention, the power flow calculation method includes a simplified DC power flow model; Specifically, based on the power grid topology and node status information corresponding to each moment in the power grid spatiotemporal fusion diagram sequence, a simplified DC power flow model is used to make basic predictions of node active power injection, generating basic predicted values ​​of node power. Read the voltage phase angle information and node connection relationships of all nodes in the graph at the current time, and determine the voltage phase angle difference between the starting and ending nodes connected by each edge. Based on the susceptance value and voltage phase angle difference of the connecting lines between nodes, obtain the active power injection of each node in turn, and complete the basic prediction.

[0041] During the prediction process, the active power injection value is independently calculated for each node, and the predicted value is used as the initial node power prediction result for the current time. After all nodes are predicted, the basic node power prediction vector for the current time point is formed.

[0042] It should be noted that the node power basic prediction vector retains the linear characteristics of the power grid structure and physical relationship, and has global consistency and interpretability.

[0043] In this embodiment of the invention, after completing steps B1-B2, step S300 further includes steps B3-B4. In B3: Obtain the true node power, and use a graph neural network to learn the residual between the prediction error of the basic predicted value of node power and the true node power to obtain the node power residual prediction value. In B4: The node power baseline prediction value and the node power residual prediction value are added node by node to obtain the prediction result.

[0044] In one optional embodiment, the graph neural network can be a graph attention network. For each node, the attention weight between it and its neighboring nodes is calculated to reflect the degree of influence of different neighbors on the current node. The features of the neighboring nodes are weighted and summed according to the attention coefficient to generate a higher-order feature representation of the node. By stacking multi-head attention layers, deeper residual features are extracted layer by layer, and the output features are mapped to the power residual space to generate the node power residual prediction value. In another alternative embodiment, the graph neural network can also be a graph sampling and aggregation network. For each node, a fixed number of neighbor nodes are randomly sampled from the neighbor set. The computational complexity is controlled, and the neighbor features are fused with the node's own features by means aggregation, LSTM aggregation or pooling aggregation. Through multi-layer aggregation operations, a node embedding representation containing local topological information is generated. The embedding vector is input into a fully connected layer to map and obtain the node power residual prediction value. In this embodiment of the invention, the graph neural network includes a graph convolutional network; Specifically, the node attribute vector and topological connectivity at the current time point in the spatiotemporal fusion graph sequence of the power grid are used as input. Following the node adjacency structure of the graph convolutional network, the attribute vectors of the current node and its neighboring nodes are jointly input into the graph convolution operation. In each graph convolution operation, the state information of each node under structural adjacency is weighted and aggregated sequentially to generate a high-dimensional feature representation of the node, used to characterize the influence of local topology and attributes on power error. The output features of the graph convolutional network are mapped to the power residual space to generate the node power residual prediction value, representing the correction magnitude between the current node's basic prediction value and the true value.

[0045] During the training of the graph convolutional network, the actual power of the nodes measured in history is used as the supervision signal. The loss is calculated by calculating the squared error between the final predicted value of the node and the actual power. The parameters of the graph convolutional network are updated by backpropagation, so that the graph convolutional network has the ability to effectively fit the residual of the node power.

[0046] In this embodiment of the invention, after obtaining the prediction results of the basic physical model and the prediction results of the graph convolutional network residuals, the two are summed node by node to form the final node power prediction result. The final prediction result and the actual node power at the corresponding time are used as training targets to construct the fusion model training process.

[0047] Specifically, historical node attribute data and corresponding actual power data are collected over a certain period to construct a spatiotemporal fusion graph sequence consistent with the static graph topology. For each time slice, the basic physical model is executed to obtain the basic prediction results of the nodes, and the current graph structure and node attributes are used as input to the graph convolutional network to generate residual prediction results. The basic prediction results and residual prediction results are added node by node to generate the final prediction value, which is compared with the actual observed power in that time slice node by node to obtain the squared error value. The prediction errors of all time slices are averaged to construct the overall loss value as the training objective function to quantify the overall prediction performance of the fusion model across all time periods. Based on the current loss function, gradient backpropagation is performed on the graph convolutional network parameters to update its ability to fit the residual features.

[0048] It should be noted that by integrating physical laws and data-driven advantages, and correcting basic prediction errors through residual learning, the power prediction accuracy and emergency response capability under complex dynamic scenarios are improved.

[0049] In this embodiment of the invention, step S400 includes the following sub-steps C1-C4; In C1: Based on the prediction results, calculate the load rate of each transmission line; In C2: Based on the load rate, the over-limit risk index of each line is calculated using predefined risk mapping rules; In C3: Record the over-limit risk indicators in chronological order to construct a time series of risk indicators at the line level; In C4: Using a time-series forecasting model, risk indicators within a specified future time range are predicted based on historical risk indicator time series, thus constructing future risk trends.

[0050] In this embodiment of the invention, calculating the load rate of each transmission line includes reading the start and end nodes of each transmission line in the topology, combining the node power prediction results with the line parameters, calculating the power flow of the transmission line, and taking its absolute value as the actual load of the line at the current moment. Using the maximum allowable transmission power of each line as a reference value, its load rate, i.e., the ratio between the actual load of the line and the maximum allowable value, is calculated to form a load rate sequence.

[0051] The predefined risk mapping rules include classifying risk levels based on load rate sequences and calculating the over-limit risk index for each line.

[0052] If the load factor does not exceed 80%, the corresponding risk indicator is 0; if the load factor is between 80% and 100%, it is linearly mapped to a risk value between 0 and 1; if the load factor exceeds 100%, the over-limit risk value is calculated in a linear growth manner, and the risk level increases with the increase of the over-limit range. To enhance the risk sensitivity of over-limit lines, a risk penalty factor is introduced to adjust the weight of the over-limit portion, ensuring that high-risk lines have significant identification capabilities in subsequent trend analysis and alarm determination.

[0053] For example, if the maximum allowable power of a transmission line is 100 MW and the predicted power is 85 MW, then the load factor is 0.85, corresponding to an over-limit risk index of 0.25. If the predicted power is 110 MW, then the load factor is 1.10, corresponding to an over-limit risk index of 1.20. After calculating the over-limit risk indexes for all lines, a complete risk index vector will be formed, providing an input basis for subsequent risk trend prediction.

[0054] By continuously recording the over-limit risk index values ​​at fixed time intervals, a time series of risk indicators at the line level is constructed.

[0055] In one optional embodiment, the time-series prediction model can be a gated recurrent unit. Based on the current input and the hidden state of the previous time step, it calculates a reset gate to control the degree of forgetting of historical information; calculates an update gate to balance the fusion ratio of historical information and the current candidate state; combines the reset gate result to generate the candidate hidden state of the current time step; integrates the historical state and the candidate state using the update gate to output the predicted value of the risk indicator at the current time step; and recursively predicts the risk trend within a specified time range in the future based on the historical risk indicator sequence. In another optional embodiment, the time series prediction model can also be a Transformer time series model, adding positional encoding to the input risk indicator sequence to retain temporal positional information; capturing the long-range dependencies between risk indicators at different times in the sequence through a self-attention mechanism; performing a nonlinear transformation on the attention output through a feedforward network to extract high-order temporal features; using the encoder output, generating future risk indicator prediction values ​​time-by-time through a decoder; arranging the prediction results in chronological order to construct a risk trend within a specified future time range; In this embodiment of the invention, the time-series prediction model includes a long short-term memory network; Specifically, we selected Long Short-Term Memory (LSTM) networks as the risk evolution modeling method and trained the LTM networks using risk indicator vectors from multiple consecutive time points as input. The acquisition of historical risk indicator time series involves employing a sliding window strategy. This strategy uses a fixed window length to extract historical risk indicator sequences in chronological order and use them as input samples. The corresponding prediction target is the risk indicator value at one or more time points after the window ends. This process covers multiple time periods during the training phase, updating parameters by minimizing the error between the predicted values ​​and the actual risk indicators.

[0056] Based on a trained Long Short-Term Memory (LSTM) network, the system uses newly generated risk indicators from multiple consecutive time points as input to generate real-time predictions of risk indicators within a specified future timeframe, thus constructing future risk trends. The prediction timeframe can be set according to requirements; for example, it can be set to predict risk indicator trends at different time granularities, such as 5 minutes, 10 minutes, or 15 minutes, ensuring that risk warnings are forward-looking.

[0057] It should be noted that by constructing a time series of risk indicators and using a long short-term memory network for trend prediction, the evolution of risks can be dynamically reflected, the direction of risk development can be identified in advance, and forward-looking decision support can be provided for proactive prevention and control of the power grid.

[0058] In this embodiment of the invention, step S500 includes the following sub-steps D1-D4; In D1: Based on the risk trend prediction results, the transmission lines are judged to obtain the target lines; In D2: Locate the position of the target line in the power grid topology to obtain the starting node for path tracing; In D3: Starting from the initial node, use a graph search algorithm to find all reachable paths leading to the target path; In D4: Combine nodes and edges on reachable paths to form a risk propagation chain.

[0059] In one optional embodiment, the graph search algorithm can be a depth-first search algorithm. The end node of the target line obtained by backtracking is used as the starting node of the path tracking and marked as visited. Starting from the starting node, the algorithm explores downstream nodes along a connecting edge until a high-risk line is reached or the traversal cannot continue. If the current path cannot continue or the target has been reached, the algorithm backtracks to the previous node and explores other unvisited branch paths. Each time a reachable path from the starting node to a high-risk line is found, the nodes and edges on the path are recorded as a risk propagation chain in traversal order. The above process is repeated until all branches are traversed, forming a complete set of risk propagation chains. In another optional embodiment, the graph search algorithm can also be Dijkstra's shortest path algorithm. Based on a static graph structure, a weighted graph is constructed by using line risk indicators or electrical distances as edge weights. The distance of the starting node is set to 0, and the distances of all other nodes are set to infinity, establishing a priority queue. The unvisited node with the smallest distance from the priority queue is selected as the current node. The cumulative distance from the current node to each neighboring node is calculated. If it is less than the known distance, the distance table is updated and the predecessor node is recorded. After reaching a high-risk line, the shortest risk propagation path from the starting node to the target line is reconstructed in reverse based on the predecessor node record. This process is repeated for all high-risk lines to generate multiple optimal risk propagation chains. In this embodiment of the invention, the graph search algorithm includes a breadth-first search algorithm; Specifically, based on the predicted risk index values, a threshold judgment is made for the predicted risk index value of each transmission line. When the predicted risk index of any line exceeds the set risk threshold, it is determined that a risk event has been triggered. The risk threshold is uniformly set as a fixed constant, for example, 0.85. If the predicted risk index value of any line is greater than 0.85, it is determined that there is a potential risk event in the power grid at the current moment.

[0060] To identify the target route in a risk event where the predicted risk index value exceeds the risk threshold, the end node connected to the target route is traced back in reverse as the starting node for path tracking. Using the starting node as the root node, a breadth-first search algorithm is used to search for paths in the current static graph structure, traversing the connecting edges and downstream nodes in turn to find all reachable paths that propagate to the route that has been identified as high-risk. All nodes and edges on the path are then combined to form a complete risk propagation chain.

[0061] During the search process, each identified complete path is recorded as a sequential sequence representing the directed relationships between nodes and edges, until all high-risk paths are covered. All valid risk paths are then aggregated in a structured manner to form a risk propagation chain, which serves as the input for subsequent situational awareness visualization.

[0062] For example, if the predicted risk index of a certain transmission line is 0.92, which exceeds the threshold of 0.85, and the line connects nodes A and B, then a breadth-first search is performed from node A to the downstream nodes to form a path such as node A → line 1 → node C → line 2 → node D → line 3 → node B, and this path is recorded as a complete risk propagation chain.

[0063] It should be noted that this invention can accurately identify high-risk routes and trace their origins, constructing a complete risk transmission chain and providing key path evidence for precise prevention and control.

[0064] In this embodiment of the invention, after completing steps D1-D4, step S500 further includes steps D5-D7. In D5: Map all nodes and edges in the static graph structure to the geographic information layer according to their corresponding geographic coordinates; In D6: Each node and edge is visually encoded according to the corresponding risk index of exceeding the limit; In D7: On the geographic information layer, draw the path indicated by the risk propagation chain, generate a risk situation map and display it in real time.

[0065] In this embodiment of the invention, visual encoding processing includes color mapping processing; For risk paths and limit-crossing risk indicators, based on the static graph structure, all nodes and edges are mapped to the geographic information layer according to their corresponding geographic coordinates, and each node and edge is color-mapped according to its corresponding risk indicator value. Specifically, all identified risk paths and exceedance risk indicators undergo spatial visualization. Based on the constructed static power grid diagram structure, all nodes and edges are mapped to two-dimensional or three-dimensional geographic information layers according to their corresponding geographic coordinates. The method of overlaying onto the real geographic information layer is chosen to ensure accurate positioning of the power grid structure in geographic space. Each node and edge is color-mapped according to its corresponding risk indicator value, using a fixed color gradient scheme. For example, green to red represents low risk to high risk, with the higher the risk indicator, the more red the color.

[0066] Based on the risk propagation chain, each path from the starting node to the high-risk line is drawn, and the path direction is marked with highlights and directed arrows to ensure that the trend of situation changes is intuitively identifiable. All risk paths retain their structural order to ensure a complete reflection of the risk diffusion chain. After the layers are drawn, the situation map is rendered in real-time using WebGL technology and displayed on the interactive front-end page, supporting user operations such as zooming, rotating, and querying. This ensures that the dynamic display of the power grid's operational risk situation is intuitive, clear, and interactive.

[0067] It should be noted that by using geographic information visualization and interactive display, the distribution and spread of risks can be presented intuitively, thereby improving the situational awareness and decision-making efficiency of operators.

[0068] The above is an illustrative scheme of a situational awareness method for power grid operation status risk according to this embodiment. It should be noted that the technical solution of this situational awareness system for power grid operation status risk and the technical solution of the aforementioned situational awareness method for power grid operation status risk belong to the same concept. Details not described in detail in the technical solution of the situational awareness system for power grid operation status risk in this embodiment can be found in the description of the technical solution of the aforementioned situational awareness method for power grid operation status risk.

[0069] The situational awareness system for power grid operation status risks in this embodiment includes: The preprocessing module is used to acquire multi-source data of power grid nodes, preprocess the multi-source data, and obtain structured node attribute data. The graph sequence construction module is used to construct a spatiotemporal fusion graph sequence of the power grid based on the structured node attribute data and using graph serialization. The prediction module is used to predict the node power using the spatiotemporal fusion graph sequence of the power grid, obtain the basic predicted value of the node power, and correct the basic predicted value of the node power to obtain the prediction result. The risk assessment module is used to calculate the line risk index based on the prediction results and construct the corresponding time series for risk trend prediction. The path awareness module is used to identify target lines based on risk trend prediction results, track risk paths to obtain risk propagation chains, and overlay the risk propagation chains onto the information layer to achieve situational awareness of power grid operation status risks.

[0070] This embodiment also provides a computer device suitable for situational awareness of power grid operational status risks, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement a situational awareness method for power grid operation status risks as proposed in the above embodiments.

[0071] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements a situational awareness method for power grid operation status risks as proposed in the above embodiments.

[0072] The storage medium proposed in this embodiment and the situational awareness method for power grid operation status risk proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0073] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computing device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0074] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A situational awareness method for power grid operating state risk, characterized in that, include: Obtain multi-source data of power grid nodes, preprocess the multi-source data to obtain structured node attribute data; Based on the structured node attribute data, a spatiotemporal fusion graph sequence of the power grid is constructed using graph serialization; The node power is predicted using the spatiotemporal fusion graph sequence of the power grid to obtain the basic predicted value of the node power, and the basic predicted value of the node power is corrected to obtain the prediction result. Based on the prediction results, the route risk index is calculated, and the corresponding time series is constructed to predict the risk trend. Based on the risk trend prediction results, the target line is identified, and the risk path is traced to obtain the risk propagation chain. The risk propagation chain is then superimposed on the information layer to achieve situational awareness of power grid operation status risks.

2. The situational awareness method for power grid operation state risk according to claim 1, characterized in that, The sequence for constructing the spatiotemporal fusion map of the power grid includes: Construct a graph structure based on the actual topology of the power grid; For each node in the graph structure, construct a node attribute vector and append the node attribute vector to the corresponding node to obtain the static graph structure; The observation period is divided into multiple consecutive time slices according to a preset time interval; The static graph structure is copied at each time slice, and the attribute vectors of all nodes are updated to generate the corresponding time sequence graph. Arrange all the time series diagrams to obtain the spatiotemporal fusion diagram sequence of the power grid.

3. The situational awareness method for grid operation state risk according to claim 2, characterized in that, Predicting node power includes: Based on the power grid topology and node status information corresponding to each moment in the power grid spatiotemporal fusion graph sequence; The power flow calculation method is used to make basic predictions of node power and generate basic predicted values ​​of node power.

4. The situational awareness method for grid operation state risk according to claim 3, characterized in that, The correction to the baseline prediction of the node power includes: To obtain the actual node power, a graph neural network is used to perform residual learning on the prediction error between the basic predicted value of node power and the actual node power, so as to obtain the residual predicted value of node power. The prediction results are obtained by adding the basic predicted values ​​of node power to the predicted values ​​of node power residuals node by node.

5. The situational awareness method for grid operation state risk according to claim 4, characterized in that, Risk trend forecasting includes: Based on the prediction results, the load rate of each transmission line is calculated; Based on the load rate, the over-limit risk index for each line is calculated using predefined risk mapping rules; Record the risk indicators of exceeding limits in chronological order to construct a time series of risk indicators at the line level; By using time-series forecasting models, risk indicators within a specified future time range are predicted based on historical risk indicator time series, thus constructing future risk trends.

6. The situational awareness method for power grid operation status risks as described in claim 5, characterized in that, Identifying the target route includes: Based on the risk trend prediction results, the transmission lines are judged to obtain the target lines; Locate the position of the target line in the power grid topology to obtain the starting node for path tracing; Starting from the initial node, use a graph search algorithm to find all reachable paths leading to the target path; By combining nodes and edges on reachable paths, a risk propagation chain can be formed.

7. A situational awareness method for power grid operation status risks as described in claim 5 or 6, characterized in that, Overlaying the risk transmission chain onto the information layer includes: Map all nodes and edges in the static graph structure to the geographic information layer according to their corresponding geographic coordinates; Each node and edge is visually encoded based on the corresponding risk index of exceeding the limit. On the geographic information layer, the path indicated by the risk propagation chain is drawn, a risk situation map is generated and displayed in real time.

8. A situational awareness system for power grid operation status risks, employing a situational awareness method for power grid operation status risks as described in any one of claims 1-7, characterized in that, include: The preprocessing module is used to acquire multi-source data of power grid nodes, preprocess the multi-source data, and obtain structured node attribute data. The graph sequence construction module is used to construct a spatiotemporal fusion graph sequence of the power grid based on the structured node attribute data and using graph serialization. The prediction module is used to predict the node power using the spatiotemporal fusion graph sequence of the power grid, obtain the basic predicted value of the node power, and correct the basic predicted value of the node power to obtain the prediction result. The risk assessment module is used to calculate the line risk index based on the prediction results and construct the corresponding time series for risk trend prediction. The path awareness module is used to identify target lines based on risk trend prediction results, track risk paths to obtain risk propagation chains, and overlay the risk propagation chains onto the information layer to achieve situational awareness of power grid operation status risks.

9. A computer device, characterized in that, include: A memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of a situational awareness method for power grid operating state risks as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed by a processor, implements the steps of a situational awareness method for power grid operation status risks as described in any one of claims 1 to 7.

Images