A Power Grid Topology Resilience Enhancement Method Based on Graph Neural Network Pre-Disaster Prediction

Abstract

This invention discloses a method for enhancing power grid topology resilience based on graph neural network pre-disaster prediction. The steps are as follows: constructing a power grid graph structure with power grid equipment as nodes; predicting the probability of node damage using a graph attention network and defining a set of high-risk nodes; constructing and solving a mixed-integer programming model based on the high-risk nodes and their propagation paths to generate the optimal pre-disaster power grid structure adjustment strategy; combining this strategy with a deep reinforcement learning algorithm to generate a resource scheduling scheme; calculating core indicators of power grid resilience after simulation, and optimizing the structure adjustment strategy if the indicators are lower than a preset threshold; uploading the optimized strategy and scheduling scheme to the scheduling platform to complete the pre-disaster proactive defense deployment. This invention achieves synergistic linkage between pre-disaster risk prediction, topology reconstruction, and resource deployment, automatically generating differentiated pre-disaster defense schemes, transforming "post-disaster repair" into "pre-disaster deployment," and significantly improving the efficiency of pre-disaster prevention and control response.

Description

Technical Field

[0001] This invention relates to the field of power grid disaster prevention, mitigation, and intelligent operation and maintenance technology, specifically to a method for enhancing power grid topology resilience based on graph neural network pre-disaster prediction. Background Technology

[0002] Under the current technological background, power grid disaster risk assessment mainly relies on the "risk prediction + post-disaster response" model, which has several shortcomings:

[0003] Risk prediction remains a reactive, after-the-fact response.

[0004] Existing research mainly focuses on using meteorological, geological, and historical fault data to train predictive models and output fault probabilities or risk levels. However, these results can only serve as a reference for post-disaster emergency response and cannot form pre-disaster intervention measures, making it difficult to break the cascading chain of "typhoon → landslide → tower collapse → line tripping".

[0005] The power grid topology is static and cannot be adjusted.

[0006] Traditional power grid operation is based on static topology. Even if high-risk nodes and critical channels are identified before a disaster, there is a lack of structural reconfiguration capabilities to perform "disconnect-retain-switch" before the disaster. As a result, the impact of the disaster is directly transmitted to the backbone network, leading to large-scale power outages.

[0007] Emergency resource allocation is lagging behind;

[0008] The existing system mostly activates emergency teams and allocates resources after a disaster based on the extent of damage, often missing the "golden window for defense." This "post-disaster recovery" model makes it difficult for repair teams to arrive in time and for mobile energy storage devices to be deployed in advance, thus delaying the restoration of the power grid's resilience.

[0009] Graph neural networks have limited applications.

[0010] In recent years, graph neural networks (GNNs) have been applied in power grid risk prediction, capable of modeling topology and disaster propagation paths between nodes. However, most existing methods are only used for "disaster probability prediction" or "cascading failure simulation," lacking linkage design with power grid topology reconfiguration and resource deployment, and failing to answer the question of "how to proactively enhance resilience."

[0011] Lack of a disaster prevention paradigm;

[0012] Overall, current technologies remain within the "prediction-response" chain, lacking a new paradigm of "pre-disaster intervention-proactive deployment." There is an urgent need to utilize intelligent methods such as graph neural networks to achieve coordinated optimization of high-risk node identification, topology reconstruction, and resource pre-deployment before disasters, thereby improving power grid resilience in advance. Summary of the Invention

[0013] To address the shortcomings of existing technologies, this invention provides a method for enhancing the resilience of power grid topology based on pre-disaster prediction using graph neural networks, aiming to solve the problems in the background technology.

[0014] To achieve the above objectives, the present invention provides the following technical solution: a method for enhancing the topological resilience of power grids based on pre-disaster prediction using graph neural networks, comprising the following steps:

[0015] Step S1: Establish a power grid graph structure with power grid equipment as nodes; predict the damage probability of each node based on the graph attention network, and define a set of high-risk nodes based on the damage probability;

[0016] Step S2: Based on the set of high-risk nodes and their propagation paths obtained in Step S1, construct and solve a mixed integer programming model to generate the optimal power grid structure adjustment strategy before the disaster.

[0017] Step S3: Based on the pre-disaster optimal power grid structure adjustment strategy in Step S2, a resource scheduling scheme is generated using a deep reinforcement learning algorithm;

[0018] Step S4: Optimal pre-disaster power grid structure adjustment strategy based on the output of Step S2 The resource scheduling scheme output in step S3 is simulated and run. During the simulation cycle, the core indicators for measuring grid resilience are calculated and compared with the preset resilience threshold. If the core indicators are less than the resilience threshold, the optimal grid structure adjustment strategy is adjusted. Optimize it; otherwise, leave it unchanged.

[0019] Step S5: Apply the pre-disaster optimal power grid structure adjustment strategy output in Step S4. The resource allocation plan is uploaded to the allocation platform to complete proactive defense deployment before a disaster occurs.

[0020] Furthermore, the specific process of step S1 is as follows:

[0021] Step S1.1: Integrate multi-source heterogeneous data, including meteorological, geological, power grid equipment attributes, and historical fault data; treat power grid equipment as nodes in a graph structure; treat the relationships between power grid equipment as edges in a graph structure, and represent the graph structure using a ternary adjacency matrix.

[0022] Step S1.2: Construct the graph attention network and initialize the network parameters;

[0023] Step S1.3: Input the graph structure into the graph attention network to perform node-level disaster risk probability prediction; each node In the The feature vector of the layer is:

[0024] ;

[0025] In the formula, Indicates the first Layer nodes eigenvectors; Represents a node and nodes Attention weights between them; Represents the learnable parameter matrix; Represents a node The set of neighboring nodes; Indicates the first Layer nodes eigenvectors;

[0026] After multiple propagation stages, the probability of damage to each node is output. :

[0027] ;

[0028] In the formula, This represents the transpose of the learnable weight vector of the output layer; Indicates the first Layer nodes eigenvectors;

[0029] Based on the probability of damage to each node and preset threshold Define a set of high-risk nodes :

[0030] .

[0031] Furthermore, the specific process of step S2 is as follows:

[0032] Step S2.1: Define the decision variables for the mixed-integer programming model:

[0033] Represents a node The power supply status; Indicates the reserved node Power supply; Indicates a disconnected node Power supply;

[0034] Representing a path The operational status; Indicates the path between activated nodes ; Indicates the path between inactive nodes. ;

[0035] Step S2.2: Establish the objective function of the mixed-integer programming model:

[0036] ;

[0037] In the formula, Represents a node The size of the load it carries; Indicates the activation path Required operation time; A set representing paths; Indicates dynamic weighting factor;

[0038] Step S2.3: Set constraints, including grid connectivity constraints, radial operation constraints, power balance constraints, and equipment capacity constraints;

[0039] Step S2.4: Solve the mixed-integer programming model to obtain the optimal decision variables, i.e., the optimal pre-disaster power grid structure adjustment strategy. , Represents a node The optimal power supply state, Representing a path The optimal operational state.

[0040] Furthermore, resource scheduling is performed using deep reinforcement learning algorithms, as follows:

[0041] ;

[0042] In the formula, The state-action value function represents the state when the power grid is in a certain condition. At that time, the scheduling center selects and executes a resource scheduling action. The expected sum of all future rewards received; express The disaster map state at any given time is used to describe the dynamic evolution characteristics of disasters in the power grid topology. Indicates the state of the disaster map. express Solidified disaster map, Represents resource state variables. Indicates the state of damage, by Calculated; express Real-time resource scheduling decisions; Indicates the learning rate; Indicate execution Then, the feedback benefits obtained at the current moment; Indicates the discount factor; express The state of the disaster at any given moment; Indicates the candidate action for the next moment; Indicates in Disaster status at any given moment Next, candidate actions corresponding The maximum value in.

[0043] Furthermore, in step S4, the core indicators for measuring grid resilience are calculated. :

[0044] ;

[0045] In the formula, Indicates in At any given moment, the total amount of load that has not yet been restored in the power grid; This represents the total load of the power grid under normal conditions; Indicates the simulation period; This represents the integral.

[0046] Furthermore, in step S4, if the value is less than the resilience threshold, the optimal power grid structure adjustment strategy is then implemented. The specific process of optimization is as follows: If < toughness threshold Then, the core indicators are calculated through causal inference. Attention weights for graph attention networks gradient: , This indicates the sign of the partial derivative; the attention weights are then adjusted backwards based on the calculated gradient. .

[0047] Furthermore, the cross-entropy loss function is used. Train the graph attention network to minimize the prediction error;

[0048] ;

[0049] In the formula, Represents the actual label of the node; It represents the set of all nodes in a graph structure.

[0050] An electronic device includes a processor, a memory, and a bus, wherein the processor and the memory are connected via the bus, wherein the memory is used to store a set of program code, and the processor is used to call the program code stored in the memory to execute a power grid topology resilience enhancement method based on graph neural network pre-disaster prediction.

[0051] A non-volatile computer storage medium storing computer-executable instructions that execute a power grid topology resilience enhancement method based on graph neural network pre-disaster prediction.

[0052] Compared with existing technologies, the present invention has the following advantages:

[0053] (1) In a dynamic risk environment with multiple disasters superimposed, the present invention can accurately identify high-risk nodes and key links by modeling disaster coupling effects and cascading propagation paths in real time through graph neural networks, thereby realizing pre-disaster full-domain risk situation perception and equipment-level resilience assessment.

[0054] (2) This invention enables the coordinated linkage between pre-disaster risk prediction results and topology reconfiguration and resource deployment, and introduces a core indicator of power grid resilience as a unified evaluation indicator. This indicator is used both to quantitatively evaluate the simulation results in step S4 to determine whether the optimal pre-disaster power grid structure adjustment strategy needs further optimization, and to compare and screen among multiple candidate schemes, thereby automatically generating differentiated pre-disaster defense schemes. Through this mechanism, the traditional "post-disaster repair" can be transformed into "pre-disaster deployment," significantly improving the response efficiency of power grid pre-disaster prevention and control.

[0055] (3) While considering the scarcity of regional data and the difficulty of model transfer, this invention continuously verifies and optimizes the reconstruction strategy through the digital twin adversarial simulation mechanism. It can ensure rapid cross-regional adaptation of the strategy without large-scale target domain data, and ensure early deployment and safe operation in high-risk areas. Attached Figure Description

[0056] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0057] like Figure 1 As shown, the present invention provides a technical solution: a method for enhancing the topological resilience of power grids based on pre-disaster prediction using graph neural networks, comprising the following steps:

[0058] Step S1: Establish a power grid graph structure with power grid equipment as nodes; predict the damage probability of each node based on the graph attention network, and define a set of high-risk nodes based on the damage probability.

[0059] Step S1.1: Integrate multi-source heterogeneous data (meteorological, geological, power grid equipment attributes, historical fault data, etc.); abstract power grid equipment as nodes in a graph structure (node ​​attributes include equipment type, exposure, disaster resistance, etc.); abstract the relationships between power grid equipment as edges (such as physical connections, redundancy relationships, geographical proximity), and represent the graph structure using a ternary adjacency matrix.

[0060] Step S1.2: Construct a graph attention network (GAT) and initialize the network parameters (learnable weight matrix and attention mechanism parameters).

[0061] Step S1.3: Input the graph structure into a Graph Attention Network (GAT) to perform node-level disaster risk probability prediction; each node In the The feature vector of the layer is:

[0062] ;

[0063] In the formula, Indicates the first Layer nodes eigenvectors; Represents a node and nodes Attention weights between them; Represents the learnable parameter matrix; Represents a node The set of neighboring nodes; Indicates the first Layer nodes eigenvectors.

[0064] After multiple propagation stages, the probability of damage to each node is output. :

[0065] ;

[0066] In the formula, This represents the transpose of the learnable weight vector of the output layer; Indicates the first Nodes in the last layer eigenvectors.

[0067] Among them, the cross-entropy loss function is used. Train the graph attention network (GAT) to minimize prediction error;

[0068] ;

[0069] In the formula, Represents the actual label of the node; It represents the set of all nodes in a graph structure.

[0070] Based on the probability of damage to each node and preset threshold Define a set of high-risk nodes :

[0071] .

[0072] This step integrates heterogeneous data from multiple sources, including meteorology, geology, power grid equipment, and historical faults. It abstracts power grid equipment as nodes and spatial adjacency relationships as edges, constructing a unified spatiotemporal graph model. Nodes are then assigned multi-dimensional attributes such as equipment type, exposure level, and disaster resilience. Its purpose is to use Graph Attention Networks (GAT) to model the nonlinear cascading propagation of multiple disasters, accurately outputting the pre-disaster damage probability distribution of each node. This provides quantifiable risk information for subsequent assessments of "who to cut off, who to protect, and how to reconstruct," avoiding misjudgments caused by relying solely on experience or single indicators.

[0073] Step S2: Based on the set of high-risk nodes and their propagation paths obtained in Step S1, construct and solve a mixed integer programming model (MIP) to generate the optimal power grid structure adjustment strategy before the disaster.

[0074] In this context, the propagation path refers to the set of links in the power grid graph structure through which a high-risk node transmits its failure risk to adjacent nodes via edge sets (transmission lines). The propagation path depicts the cascading failure links that a disaster may form through the power grid topology, including not only the probability of damage to the high-risk node itself, but also reflecting its potential impact on adjacent nodes and downstream loads.

[0075] By introducing a set of high-risk nodes and their propagation paths, the mixed-integer programming model can simultaneously identify "risk source nodes" and "risk transmission channels." Before a disaster, it can proactively block the cascading propagation of disasters by disconnecting, switching, or retaining some nodes and lines, thereby generating the optimal power grid structure adjustment strategy before a disaster.

[0076] Step S2.1: Define the decision variables for the mixed-integer programming model:

[0077] Represents a node The power supply status; Indicates the reserved node Power supply; Indicates a disconnected node Power supply.

[0078] Representing a path The operational status; Indicates the path between activated nodes ; Indicates the path between inactive nodes. .

[0079] Step S2.2: Establish the objective function of the mixed integer programming model (MIP) to simultaneously minimize load loss and recovery time;

[0080] ;

[0081] In the formula, Represents a node The size of the load it carries; Indicates the activation path Required operation time (e.g., switch switching time); A set representing paths; Represents dynamic weighting factors. , Represents a node electrical load, This represents the preset basic weight constant.

[0082] Step S2.3: Set constraints:

[0083] Power grid connectivity constraints: ensure that the optimized network structure is connected and has no isolated islands.

[0084] Radial operation constraint: Distribution networks are usually required to operate radially, that is, there cannot be loops.

[0085] Power balance constraint: Ensure that the power output by the power source is equal to the power consumed by the load (including line loss).

[0086] Equipment capacity constraints: Ensure that equipment such as lines and transformers are not overloaded.

[0087] Step S2.4: Solve the mixed integer programming model (MIP) to obtain the optimal decision variables, i.e., the optimal pre-disaster power grid structure adjustment strategy. , Represents a node The optimal power supply state, Representing a path The optimal operational state.

[0088] Step S3: Based on the optimal power grid structure adjustment strategy before the disaster in Step S2, a resource scheduling scheme is generated using the Deep Reinforcement Learning (DRL) algorithm (to achieve efficient blocking of cascading fault propagation, transforming the traditional "post-disaster repair" into "pre-disaster deployment"), represented as:

[0089] ;

[0090] In the formula, The state-action value function represents the state when the power grid is in a certain condition. At that time, the scheduling center selects and executes a resource scheduling action. The expected sum of all future rewards that can be obtained (i.e., long-term returns). express The disaster map state at a given time (a joint representation of the power grid topology, node damage probability, failed power grid equipment, and resource distribution at a given time) is used to describe the dynamic evolution characteristics of disasters in the power grid topology. Indicates the state of the disaster map. express Solidified disaster map, Represents resource state variables (such as resource reserves and distribution). Indicates the state of damage, by Calculated; express Real-time resource scheduling decisions; Indicates the learning rate; Indicate execution Then, the feedback benefits obtained at the current moment; Indicates the discount factor; express The state of the disaster at any given moment; Indicates the candidate action for the next moment; Indicates in Disaster status at any given moment Next, candidate actions corresponding The maximum value in.

[0091] This step, based on the established pre-disaster topology, employs graph embedding-based deep reinforcement learning (as shown in the graph policy network GPN) to pre-deploy emergency resources such as repair teams, mobile energy storage vehicles, and temporary ring network cabinets to "risk blocking points." Its purpose is to leverage the trial-and-error and reward mechanisms of reinforcement learning to find the optimal spatial distribution and quantity configuration for blocking the propagation of cascading faults before a disaster occurs, maximizing resource utilization efficiency, shortening post-disaster recovery time, and transforming traditional "post-disaster repair" into "pre-disaster deployment."

[0092] Step S4: Optimal pre-disaster power grid structure adjustment strategy based on the output of Step S2 The resource scheduling scheme output in step S3 is simulated and run. During the simulation cycle, the core indicators for measuring grid resilience are calculated and compared with the preset resilience threshold. If the core indicators are less than the resilience threshold, the optimal grid structure adjustment strategy is adjusted. Optimize it, otherwise leave it unchanged.

[0093] Among them, the core indicators for measuring power grid resilience are calculated. :

[0094] ;

[0095] In the formula, Indicates in At any given moment, the total amount of load that has not yet been restored in the power grid; This represents the total load of the power grid under normal conditions; Indicates the simulation period; This represents the integral.

[0096] Among them, if it is less than the resilience threshold, then the optimal power grid structure adjustment strategy is adopted. The specific process of optimization is as follows:

[0097] like < toughness threshold Then, the core indicators are calculated through causal inference. Attention weights for Graph Attention Network (GAT) gradient: , This indicates the sign of the partial derivative; the attention weights are then adjusted backwards based on the calculated gradient. This forms a rolling optimization closed loop.

[0098] This step injects the topology reconfiguration and resource deployment schemes generated in steps S2-S3 into the power grid digital twin. Through adversarial simulation, it simulates the evolution of multiple disasters such as typhoons and thunderstorms, calculates the resilience index R in real time, and identifies key control variables using a causal inference model. Its purpose is to verify the effectiveness of the strategy in a low-cost, high-concurrency virtual environment, discover potential vulnerabilities, and inversely optimize the graph neural network parameters and reconfiguration strategy, forming a continuous self-evolutionary closed loop of "data-model-strategy," ensuring the transferability and robustness of the strategy across regions and disaster types.

[0099] Step S5: Apply the pre-disaster optimal power grid structure adjustment strategy output in Step S4. The resource scheduling plan is uploaded to the scheduling platform (Power Grid Dispatch System (EMS / SCADA)) to complete the proactive defense deployment before the disaster occurs.

[0100] An electronic device includes a processor, a memory, and a bus, wherein the processor and the memory are connected via the bus, wherein the memory is used to store a set of program code, and the processor is used to call the program code stored in the memory to execute a power grid topology resilience enhancement method based on graph neural network pre-disaster prediction.

[0101] A non-volatile computer storage medium storing computer-executable instructions that execute a power grid topology resilience enhancement method based on graph neural network pre-disaster prediction.

[0102] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for enhancing the topology resilience of a power grid based on pre-disaster prediction of a graph neural network, characterized in that, Comprising the following steps: Step S1: Establish a power grid graph structure with power grid equipment as nodes; predict the damage probability of each node based on the graph attention network, and define a high-risk node set based on the damage probability; Step S2: Based on the high-risk node set and its propagation path obtained in step S1, a mixed integer programming model is constructed and solved to generate a pre-disaster optimal power grid structure adjustment strategy; Step S3: On the basis of the pre-disaster optimal power grid structure adjustment strategy in step S2, a resource scheduling scheme is generated using a deep reinforcement learning algorithm; Step S4: Simulation running based on the pre-disaster optimal power grid structure adjustment strategy output in step S2 and the resource scheduling scheme output in step S3, calculating the core index measuring the power grid resilience in the simulation running period, comparing the core index with the preset resilience threshold, if less than the resilience threshold, optimizing the optimal power grid structure adjustment strategy, otherwise unchanged; Step S4: Simulation running based on the pre-disaster optimal power grid structure adjustment strategy output in step S2 and the resource scheduling scheme output in step S3, calculating the core index measuring the power grid resilience in the simulation running period, comparing the core index with the preset resilience threshold, if less than the resilience threshold, optimizing the optimal power grid structure adjustment strategy, otherwise unchanged; Step S4: Simulation running based on the pre-disaster optimal power grid structure adjustment strategy output in step S2 and the resource scheduling scheme output in step S3, calculating the core index measuring the power grid resilience in the simulation running period, comparing the core index with the preset resilience threshold, if less than Step S5: adjust the pre-disaster optimal power grid structure policy output by step S4 and resource scheduling scheme to the scheduling platform, complete the active defense deployment before the disaster occurs; The specific process of step S2 is: Step S2.1: Define the decision variables of the mixed integer programming model: representing a node whose power state is representing a node whose power is reserved representing a node whose power is disconnected representing a path of operational status; representing a path between activated nodes; representing a path between non-activated nodes; Step S2.2: Establish the objective function of the mixed integer programming model: ； wherein representing a node the size of the load carried; representing an activation path the required operation time; representing a set of paths; representing a dynamic weight factor; representing a set of all nodes in the graph structure; Step S2.3: Set the constraint conditions, including power grid connectivity constraint, radial operation constraint, power balance constraint and equipment capacity constraint; Step S2.4: solve the mixed integer programming model to obtain the optimal decision variables, i.e., the optimal pre-disaster grid structure adjustment strategy , denotes the optimal power supply state of node , denotes the optimal operation state of path ; denotes the set of all nodes in the graph structure; In step S4, the core indicator measuring the resilience of the power grid is calculated : ； wherein represents the total amount of loads in the power grid which have not been restored at the time instant represents the total amount of loads in the power grid which have not been restored at the time instant represents the total load of the power grid in the normal state; represents the simulation period; represents the integral.

2. The method of claim 1, wherein the method is based on a graph neural network. The specific process of step S1 is: Step S1.1: Fusion of multi-source heterogeneous data including meteorology, geology, power grid equipment attributes, historical fault data; the power grid equipment is taken as the node in the graph structure; the relationship between the power grid equipment is taken as the edge in the graph structure, and the graph structure is represented in the form of three-value adjacency matrix; Step S1.2: Construct the graph attention network and initialize the network parameters; Step S1.3: input the graph structure into the graph attention network for node-level disaster risk probability prediction; each node In the first The feature vector of the layer is: ； wherein, denotes the feature vector of node in layer ; denotes the attention weight between node and node ; denotes the learnable parameter matrix; denotes the set of neighbor nodes of node ; denotes the feature vector of node in layer ; The probability of compromise of each node after multiple layers of propagation : ； wherein denotes the transpose of the learnable weight vector of the output layer; denotes the feature vector of the node in the layer ; According to the probability of being compromised of each node and a preset threshold , define a high-risk node set : 。 3. The method of claim 2, wherein the method is based on a graph neural network. Resource scheduling using a deep reinforcement learning algorithm is represented as: ； wherein, denotes the state-action value function, representing the expected sum of all future rewards obtained when the power grid is in state at time t, the dispatch center selects and executes a certain resource scheduling action and then obtains the expected sum of all future rewards; denotes the disaster map state at time t, used to describe the dynamic evolution characteristics of the disaster in the power grid topology, denotes the disaster map state, denotes the disaster map after solidification, denotes the resource state variable, denotes the damage state, calculated by ; denotes the resource scheduling decision at time t; denotes the learning rate; denotes the feedback income obtained at the current time t after executing ; denotes the discount factor; denotes the disaster map state at time t; denotes the candidate action at the next time t+1; denotes the disaster map state at time t+1 after executing the candidate action ; the candidate action corresponds to the maximum value in .

4. The method of claim 3, wherein the method is characterized by: In step S4, if less than the resilience threshold, adjust the optimal power grid structure strategy The specific process of optimization is: if <resilience threshold , calculate the core indicators through causal inference The gradient of the attention weight of the graph attention network: , represents the partial derivative symbol; according to the obtained gradient after calculation, the attention weight is revised in reverse .​ 5. The method of claim 4, wherein the method is based on a graph neural network. Using a cross-entropy loss function training the graph attention network to minimize the prediction error; ； In the formula, represents the real label of the node; represents the set of all nodes in the graph structure.

6. An electronic device, comprising: A computer including a processor, a memory and a bus, wherein the processor and the memory are connected through the bus, wherein the memory is used to store a set of program codes, and the processor is used to call the program codes stored in the memory to execute the power grid topology resilience enhancement method based on the graph neural network pre-disaster prediction according to any one of claims 1-5.

7. A non-transitory computer storage medium storing computer-executable instructions, the computer-executable instructions comprising instructions for: The computer executable instructions execute the power grid topology resilience enhancement method based on the graph neural network pre-disaster prediction according to any one of claims 1-5.

Images