Power grid disaster resistance resilience evaluation method, device and equipment for natural disaster

Abstract

The application relates to a power grid disaster resistance resilience evaluation method, device and equipment for coping with natural disasters, in particular to the technical field of power systems, and more particularly to a method for evaluating the resilience of power grid components in a power grid based on multidimensional heterogeneous data. The method comprises the following steps: acquiring multidimensional heterogeneous data corresponding to a power grid; predicting a network using a preset component prediction network to evaluate a component resilience index of a power grid component in the power grid based on the multidimensional heterogeneous data; the preset component prediction network is used to fuse the multidimensional heterogeneous input data to obtain a fault conduction time sequence feature of a disaster type influence in a time dimension and a power grid component correlation feature in a space dimension, and evaluate the component resilience index according to the fault conduction time sequence feature and the power grid component correlation feature; performing fault component classification processing according to the component resilience index to determine an enhanced component set corresponding to the disaster type; determining a target component set from the enhanced component set and performing component enhancement processing on the target component set.

Description

Technical Field

[0001] This application relates to the field of power system technology, and in particular to a method, apparatus and equipment for assessing the disaster resilience of power grids in response to natural disasters. Background Technology

[0002] Power grids are highly vulnerable to natural disasters that can physically damage grid components, leading to cascading failures and repair challenges. Strengthening critical components in the grid is an effective way to enhance grid resilience. In this regard, resilience assessment and critical component identification are crucial and can be achieved by considering data from both the power grid and natural disasters.

[0003] In the context of power grid disaster resilience assessment and key component identification, the Monte Carlo method is typically used to simulate the random impact of natural disasters on the power grid and the power grid's response process through a large number of repeated experiments.

[0004] However, due to the inherent algorithmic characteristics of the Monte Carlo method, as well as the complexity of the power grid system and the special nature of disaster data, this method suffers from problems such as excessive computation time and low evaluation accuracy in practical applications, making it difficult to meet the actual needs of efficient and accurate evaluation of key power grid components. Summary of the Invention

[0005] In view of this, this application provides a method, apparatus and equipment for assessing the disaster resilience of power grids in response to natural disasters. The main purpose is to improve the technical problems of excessive calculation time and low assessment accuracy in the practical application of related technologies, which make it difficult to meet the actual needs of efficient and accurate assessment of key power grid components.

[0006] Firstly, this application provides a method for assessing the disaster resilience of power grids in response to natural disasters, the method comprising: Acquire multidimensional heterogeneous data corresponding to the power grid. The multidimensional heterogeneous data includes power grid component data, disaster data, and power grid environmental factors. Power grid component data includes power grid resilience indicators, fault types in fault groups, fault probabilities corresponding to fault types, and power losses corresponding to fault types. Disaster data includes disaster types. Using a pre-defined component prediction network, the component resilience index of the power grid components in the power grid is evaluated based on multi-dimensional heterogeneous data. The pre-defined component prediction network is used to fuse multi-dimensional heterogeneous input data to obtain the fault propagation time-series characteristics of disaster type impact in the time dimension and the power grid component correlation characteristics in the spatial dimension. Based on the fault propagation time-series characteristics and the power grid component correlation characteristics, the component resilience index is evaluated. Based on the component resilience index, the failed components are classified and the set of reinforcement components corresponding to the disaster type is determined. Determine the target component set from the enhanced component set, and perform component enhancement processing on the target component set.

[0007] Secondly, this application provides a power grid disaster resilience assessment device for responding to natural disasters, the device comprising: The acquisition module is configured to acquire multidimensional heterogeneous data corresponding to the power grid. The multidimensional heterogeneous data includes power grid component data, disaster data, and power grid environmental factors. The power grid component data includes power grid resilience indicators, fault types in fault groups, fault probabilities corresponding to fault types, and power losses corresponding to fault types. The disaster data includes disaster types. The evaluation module is configured to use a preset component prediction network to evaluate the component resilience index of the power grid components based on multidimensional heterogeneous data. The preset component prediction network is used to fuse multidimensional heterogeneous input data to obtain the fault propagation time-series characteristics of the disaster type impact in the time dimension and the power grid component correlation characteristics in the spatial dimension. Based on the fault propagation time-series characteristics and the power grid component correlation characteristics, the component resilience index is evaluated. The determination module is configured to classify faulty components based on component resilience indicators and determine the set of enhancement components corresponding to the disaster type. The enhancement module is configured to determine the target component set from the enhancement component set and perform component enhancement processing on the target component set.

[0008] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method of the first aspect.

[0009] Fourthly, this application provides an electronic device, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the computer program to implement the method of the first aspect.

[0010] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the method of the first aspect.

[0011] By employing the above technical solution, this application provides a method, apparatus, and equipment for assessing the resilience of power grids in response to natural disasters. Compared with related technologies, this application first acquires multidimensional heterogeneous data corresponding to the power grid. This multidimensional heterogeneous data includes power grid component data, disaster data, and power grid environmental factors. The power grid component data includes power grid resilience indicators, fault types in fault groups, fault probabilities corresponding to fault types, and power losses corresponding to fault types. The disaster data includes disaster types. Using a pre-set component prediction network, the component resilience indicators corresponding to power grid components in the power grid are assessed based on the multidimensional heterogeneous data. The pre-set component prediction network is used to fuse multidimensional heterogeneous input data to obtain the fault propagation time-series characteristics of the disaster type's influence in the time dimension and the power grid component correlation characteristics in the spatial dimension. Based on the fault propagation time-series characteristics and the power grid component correlation characteristics, the component resilience indicators are assessed. Based on the component resilience indicators, faulty components are classified to determine the set of enhanced components corresponding to the disaster type. From the enhanced component set, a target component set is determined, and the target component set is subjected to component enhancement processing. By predicting the temporal patterns of disaster evolution and fault propagation in the time dimension and the distribution and cross-regional correlation characteristics of power grid components in the spatial dimension through preset components, the system identifies the component resilience indicators of each power grid component under different disaster types. By classifying and eliminating conventional components that do not require major modifications, the system provides multiple sets of components that can be enhanced. Target components can then be selected from these sets for further enhancement. This effectively addresses the core shortcomings of related technologies, such as excessive computation time, insufficient accuracy, and poor adaptability of simple neural networks. Even in complex and random scenarios, the system can still stably output high-precision power grid component identification and evaluation results, meeting the practical needs for efficient and accurate evaluation of key power grid components.

[0012] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0013] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 A flowchart illustrating the power grid disaster resilience assessment method for responding to natural disasters provided in an embodiment of this application is shown. Figure 2 This illustration shows a schematic diagram of an example Transformer network hierarchy provided in an embodiment of this application; Figure 3 This illustration shows a schematic diagram of the structure of an example Transformer network encoder and decoder provided in an embodiment of this application; Figure 4 A schematic diagram of an example quantum neural network provided in an embodiment of this application is shown; Figure 5 This illustration shows an overall framework diagram of an example provided by an embodiment of this application; Figure 6 A schematic diagram of the structure of the power grid disaster resilience assessment device for responding to natural disasters provided in an embodiment of this application is shown. Detailed Implementation

[0016] To better understand the above-mentioned objectives, features, and advantages of this application, the solution of this application will be further described below. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0017] The core logic of the Monte Carlo method in related technologies relies on the statistical regularities of massive random sampling samples to approximate the true results. The sample size directly determines the applicability of the method, and the complexity of power grid disaster resilience assessment scenarios further amplifies its time-consuming drawbacks. On the one hand, the power grid system contains a massive number of components, with complex electrical coupling relationships between them. Furthermore, the probability and extent of damage caused by natural disasters to different components exhibit significant randomness. Simultaneously, the spatial propagation characteristics of disasters and the chain reactions following component failures must also be incorporated into the simulation process. To cover these complex scenarios, the Monte Carlo method requires constructing a stochastic model containing multidimensional variables. Each sample must completely simulate the entire process of "disaster occurrence - component damage - fault propagation - power grid state response," with a single sample simulation involving a large number of electrical parameter calculations and logical judgments. On the other hand, to minimize sampling errors, the Monte Carlo method typically requires tens of thousands or even hundreds of thousands of repeated sampling experiments, with the sample size and computation time increasing approximately linearly. As the scale of the power grid expands or the assessment dimensions become more refined, the number of variables in the model increases exponentially. The computational load of a single sample simulation rises dramatically, leading to a sharp increase in overall computation time, often requiring hours or even days to complete an assessment. Power grid disaster resilience assessments frequently serve emergency decision-making and component reinforcement planning, placing clear demands on assessment efficiency. Excessive computation time prevents the Monte Carlo method from outputting assessment results in a timely manner, making it difficult to adapt to real-time decision-making needs and unable to support dynamic resilience assessment and emergency dispatch after a power grid disaster.

[0018] Although the Monte Carlo method can theoretically approximate the true value by increasing the sample size, in the scenario of power grid disaster resilience assessment, its actual calculation accuracy consistently falls short of the stringent requirements for identifying key components due to multiple limitations, and may even produce misleading results. First, the inherent characteristics of sampling error determine its upper limit of accuracy. The accuracy of the Monte Carlo method is inversely proportional to the square root of the sample size; to reduce the error by an order of magnitude, the sample size needs to be increased by two orders of magnitude. If the sample size is reduced to control time consumption, the sampling error will increase significantly, causing the assessment results to deviate from the true situation; if the sample size is increased to improve accuracy, it will fall into the dilemma of "infinitely extended time consumption," making it difficult to find a balance.

[0019] In addition, the Monte Carlo method has several derivative drawbacks that further limit its application in power grid disaster resilience assessment. First, it has poor adaptability to scenarios with multiple superimposed disasters. When faced with complex scenarios such as typhoons and rainstorms, or earthquakes and secondary disasters, the dimensionality and uncertainty of variables increase significantly. The sample size requirements and computational complexity of the Monte Carlo method rise exponentially, further exacerbating the contradiction between time consumption and accuracy, and may even prevent the assessment from being completed within a reasonable timeframe. Second, it consumes excessive hardware resources. The simulation and calculation of massive sampling samples require high-performance computing equipment. In large-scale power grid assessment scenarios, the demand for CPU and memory is extremely high, increasing the hardware cost of the assessment and making it difficult to promote its application in small and medium-sized power companies or grassroots power grid operation and maintenance departments.

[0020] In data-driven assessment methods, neural networks are widely used in various prediction and identification tasks due to their strong nonlinear fitting and adaptive learning capabilities. Some studies have attempted to use simple neural networks to predict and assess key power grid components. However, the coupled system of power grids and natural disasters is extremely complex, and simple neural networks are significantly insufficient in characterizing the input-output relationship of this system, making it difficult to meet the accuracy requirements of key component assessment and limiting their practical application in assessing the disaster resilience of power grids.

[0021] Although simple neural networks can make preliminary predictions, the input-output relationship between the power grid and natural disaster coupled systems is highly complex, dynamic, and coupled. Due to structural limitations, simple neural networks cannot effectively characterize these complex relationships, resulting in insufficient prediction accuracy and poor generalization ability. Specifically, this manifests in the following aspects: First, the complex coupling relationships between input features are difficult to capture. The input features of the power grid and disaster data are not independent but rather have multiple nonlinear coupling relationships. The shallow structure and limited number of neurons in simple neural networks cannot construct sufficiently complex feature mapping relationships. They can only capture simple linear or low-order nonlinear relationships between features, making it difficult to uncover these multi-dimensional, strongly coupled deep-seated correlation patterns. This leads to low information utilization of input features and biased prediction results. Second, dynamic spatiotemporal correlations and uncertainties are difficult to characterize. The occurrence of natural disasters has significant spatiotemporal dynamic characteristics, such as the randomness of typhoon paths and the uneven spatiotemporal distribution of rainstorm intensity. Furthermore, power grid component failures also have spatiotemporal cascading effects. In addition, the input data itself contains a large amount of uncertainty, such as disaster intensity prediction errors, measurement deviations of power grid component parameters, and incomplete historical fault data. Simple neural networks lack the ability to model dynamic spatiotemporal information, failing to effectively integrate feature changes across time and space dimensions. Furthermore, they exhibit poor robustness to data uncertainty, struggling to distinguish between data noise and valid information. This leads to a significant decrease in the stability of prediction results when facing dynamically changing disaster scenarios and uncertain data, making it impossible to accurately pinpoint the core deficiencies of critical components. Secondly, the multi-dimensional relationships between output targets cannot be covered. The assessment of critical power grid components is not a single-objective prediction but requires simultaneous consideration of the impact of component failures on power grid reliability, repair difficulty, economic losses, and social impacts, among other indicators. Complex trade-offs also exist between the various output targets. Simple neural networks typically employ single-output or independent multi-output designs, failing to characterize the inherent relationships and trade-offs between output targets. They can only predict single indicators, resulting in one-sided assessment results that cannot comprehensively reflect the deficiencies of critical components and their overall impact on power grid resilience.

[0022] To address the technical problems of excessive computation time and low evaluation accuracy in practical applications of related technologies, which fail to meet the actual needs for efficient and accurate evaluation of key power grid components, this embodiment provides a method for assessing the disaster resilience of power grids in response to natural disasters. Figure 1 As shown, the method includes: Step 101: Obtain multidimensional heterogeneous data corresponding to the power grid.

[0023] The multidimensional heterogeneous data includes power grid component data, disaster data, and power grid environmental factors. Power grid component data includes power grid resilience indicators, fault types in fault groups, fault probabilities corresponding to fault types, and power losses corresponding to fault types. Disaster data includes disaster types.

[0024] Correspondingly, multidimensional heterogeneous data can be various types of power grid-related data with different sources, formats, attributes, and dimensions; power grid component data can be data related to the operation, failure, and performance of various core components and equipment within the power grid, which belongs to the core attribute data within the power grid and directly determines the disaster resistance capability and system resilience of the components; disaster data can be data related to natural disasters with time and spatial location attributes corresponding to the power grid to be monitored, such as dynamic information on the time, location, intensity, and scope of natural disasters occurring in the power grid, which characterizes the impact pattern of external disasters on the power grid; disaster types can be the types of natural disasters that the power grid is about to respond to, which are monitored through meteorological monitoring systems; power grid environmental factors can be external environmental conditions that affect the operation and disaster resistance capability of the power grid, other than disasters, which belong to static or slowly changing external data and affect the probability and degree of component failure.

[0025] Specifically, grid resilience indicators can be core indicators for quantifying the grid's disaster resistance, fault tolerance, and rapid recovery capabilities, assessing the resilience of the grid and its components. Fault families can be sets of faulty components categorized according to the location, cause, scope of impact, and equipment type of grid faults, grouping similar or related faults together for easier classification analysis and modeling, such as transmission line fault families and substation equipment fault families. Fault types can be the overall fault manifestation corresponding to a fault family, representing the actual fault forms that grid components may experience. Fault probabilities can be the likelihood of different fault types occurring under different disasters. Power loss can be the numerical value of the grid's power supply capacity loss caused by a certain fault type, quantifying the power supply impact of faults, such as the size of the interrupted load, the number of users experiencing power outages, and the power shortage.

[0026] In some embodiments, the power grid can be continuously monitored, and various data corresponding to the power grid can be collected synchronously and integrated to form multidimensional heterogeneous data. For example, power grid component data may include line load rate, equipment operating years, and frequency of daily operation and maintenance; disaster data may include typhoon landfall time, typhoon duration, and typhoon wind force change periods in the time dimension, and typhoon path latitude and longitude, affected power grid line coordinates, disaster area grid division, and spatial distribution of wind intensity in the spatial dimension.

[0027] Step 102: Using a preset component prediction network, evaluate the component resilience index of the power grid components in the power grid based on multidimensional heterogeneous data.

[0028] Among them, the preset component prediction network is used to integrate multi-dimensional heterogeneous input data to obtain the fault propagation time series characteristics of disaster type impact in the time dimension and the grid component correlation characteristics in the spatial dimension, and evaluate the component resilience index based on the fault propagation time series characteristics and grid component correlation characteristics.

[0029] Correspondingly, power grid components can refer to various core basic equipment and components that constitute the power grid. They are the smallest functional units of power grid operation and the core objects of fault occurrence and resilience assessment, covering the entire process of power transmission, transformation and distribution, such as transmission lines, distribution transformers, insulators, switchgear, cables, circuit breakers, towers, disconnect switches, etc.

[0030] In some embodiments, grid resilience indicators and component resilience indicators under different disaster conditions can be collected. A preset component prediction network is constructed based on Transformer networks and quantum neural networks. Training samples are built based on the grid resilience indicators and component resilience indicators. The preset component prediction network is then trained to predict the component resilience indicators corresponding to grid components in the current disaster type. Accordingly, the component resilience indicator can be used as an assessment indicator to quantify the impact of a single grid component on the overall grid's disaster resistance resilience. The higher the value, the more critical and vulnerable the component is, and the greater the impact on grid resilience after a failure. It serves as a direct basis for prioritizing component reinforcement and identifying critical components.

[0031] Specifically, fault propagation timing characteristics can include dynamic features of power grid faults extracted from the time dimension, which depict the temporal changes of power grid faults from occurrence to spread and then to gradual recovery throughout the entire process of natural disaster occurrence, development, and dissipation, reflecting the transmission, spread, and evolution trend of faults over time, and belong to dynamic timing characteristics; power grid component correlation characteristics can be component topology correlation characteristics extracted from the spatial dimension based on the component correlation relationships between components, which, based on the actual physical topology of the power grid, explores the connection relationships, coupling relationships, and upstream and downstream dependencies between different components, reflecting the spatial linkage between components, where a fault in one component will affect related components, and belong to spatial topology characteristics that combine static and dynamic aspects.

[0032] Step 103: Classify the faulty components according to the component resilience index and determine the set of reinforcement components corresponding to the disaster type.

[0033] The disaster type can be any disaster scenario in which the power grid needs component reinforcement during a current or impending natural disaster, such as blizzards or torrential rain. The reinforcement component set can be a collection of components selected after fault classification that require priority for disaster hardening, performance improvement, and redundancy modification.

[0034] In some embodiments, by comprehensively evaluating and classifying multiple factors such as component resilience indicators, failure probability, power outage loss after failure, and component correlation, weak links in the power grid's disaster resistance can be identified, and strengthening them can significantly improve the overall power grid's disaster resistance resilience.

[0035] Step 104: Determine the target component set from the enhanced component set, and perform component enhancement processing on the target component set.

[0036] In some embodiments, all power grid components are sorted according to component resilience indicators, and then prioritized or ranked by vulnerability level based on the fault type of each component. This identifies critical vulnerable components under different disaster types, allowing for the selection of a target component set for priority reinforcement. This enables rapid identification of target reinforcement components in natural disaster scenarios. In this way, weak components that must be reinforced are precisely located, ultimately forming a reinforcement list that can be directly implemented, avoiding resource waste, achieving power grid disaster resistance enhancement, reducing the identification time of target components, and thus improving component reinforcement efficiency. For example, component reinforcement measures for the target component set may include: installing wind-resistant reinforcement devices on main lines and replacing insulators with high-strength ones; installing waterproof and shockproof protective shells on transformers in the core area and configuring backup units to improve the overall resilience of the power grid under typhoon scenarios.

[0037] Compared with related technologies, this embodiment first acquires multidimensional heterogeneous data corresponding to the power grid. This multidimensional heterogeneous data includes power grid component data, disaster data, and power grid environmental factors. Power grid component data includes power grid resilience indicators, fault types in fault groups, fault probabilities corresponding to fault types, and power losses corresponding to fault types. Disaster data includes disaster types. Using a pre-set component prediction network, the component resilience indicators corresponding to power grid components are evaluated based on the multidimensional heterogeneous data. The pre-set component prediction network is used to fuse multidimensional heterogeneous input data to obtain the fault propagation time-series characteristics of the disaster type's impact in the time dimension, and the power grid component correlation characteristics in the spatial dimension. Based on the fault propagation time-series characteristics and the power grid component correlation characteristics, the component resilience indicators are evaluated. Faulty components are classified according to the component resilience indicators to determine the set of enhanced components corresponding to the disaster type. A target component set is determined from the enhanced component set, and component enhancement processing is performed on the target component set. By predicting the temporal patterns of disaster evolution and fault propagation in the time dimension and the distribution and cross-regional correlation characteristics of power grid components in the spatial dimension through preset components, the system identifies the component resilience indicators of each power grid component under different disaster types. By classifying and eliminating conventional components that do not require major modifications, the system provides multiple sets of components that can be enhanced. Target components can then be selected from these sets for further enhancement. This effectively addresses the core shortcomings of related technologies, such as excessive computation time, insufficient accuracy, and poor adaptability of simple neural networks. Even in complex and random scenarios, the system can still stably output high-precision power grid component identification and evaluation results, meeting the practical needs for efficient and accurate evaluation of key power grid components.

[0038] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, in order to fully illustrate the implementation of this embodiment, step 101 may optionally include: obtaining the power loss after the cascading fault repair corresponding to the fault type caused by the disaster type; and determining the power grid resilience index corresponding to the fault type based on the power loss and the fault probability corresponding to the fault type.

[0039] In some embodiments, due to various natural disasters such as earthquakes, floods, and blizzards, the power grid is extremely vulnerable. Power loss can refer to the cumulative loss of power supply / load deficit caused by a cascading failure resulting from a possible fault type within the disaster type, from its occurrence to the completion of fault investigation, repair, and power restoration. It represents the total actual power supply loss caused to the power grid by the cascading failure, rather than the short-term power outage gap at the moment of the fault. Fault probability can be the likelihood that a particular fault type will be independently triggered under different disaster types, thereby inducing a cascading failure. The power grid resilience index can be a comprehensive quantitative score of the probability of occurrence of a fault type and the resulting power loss under different disaster types, to assess the resilience of the power grid in the face of severe natural disasters.

[0040] For example, grid resilience indicators can be used to describe the grid's ability to withstand natural disasters, and can be quantitatively measured by the expected power loss (power deficit) caused by natural disasters. To assess the grid's resilience in the face of severe natural disasters, grid resilience indicators (grid elasticity measures) can be defined as follows: ; in, It is a power grid resilience indicator. s It is a set of independent, physically damaged components that are destroyed by a disaster. A set of faulty components corresponds to a fault type (such as a major fault). It is the probability of the occurrence of the set of faulty components. This represents a group of primary faults that may occur in the power grid. In the resilient lifecycle of the power grid, the aftermath of a fault involves rapid functional degradation, rapid recovery, and slow recovery phases. The power loss associated with the set of faulted components after rapid repair can be quantified as follows: At this point, all the cascading faults caused by the relay protection action have been repaired.

[0041] Optionally, based on power loss and the fault probability corresponding to the fault type, the power grid resilience index corresponding to the fault type can be determined. Specifically, this may include: obtaining the disaster occurrence probability corresponding to the disaster type and the set of faulty components corresponding to the disaster type; and determining the power grid resilience index of the set of faulty components under the disaster type based on the component failure probability and disaster occurrence probability of each faulty component in the set of faulty components, as well as the power loss.

[0042] In some embodiments, grid resilience measures the expected loss of power supply to the grid due to natural disasters. According to this definition, a smaller value... This indicates a higher level of resilience in the power grid. Furthermore, the power grid resilience index can be expressed as: ; in, Indicates the type of disaster The probability of disaster occurrence, Indicates the type of disaster The following fault component set The probability of failure occurrence, Indicates the type of disaster Fault components in the following fault component set The probability of component failure. It can represent the set of all components in the power grid. This can represent the set of all normal components that have not been damaged by natural disasters. Optionally, step 101 may further include: obtaining a power grid component topology diagram of the power grid; determining the power grid components and the component relationships between them based on the power grid component topology diagram; determining the primary fault groups corresponding to the power grid based on the component relationships; and determining the power grid component data corresponding to the power grid components in the fault component set based on the fault component set corresponding to the fault type in the primary fault group.

[0043] The power grid component topology diagram is a structural schematic diagram that intuitively presents the physical and electrical connections of the power grid. It clearly marks the location, wiring method, and upstream and downstream links of all core components in the power grid. It can be divided into transmission topology and distribution topology, serving as the basis for fault analysis and component correlation determination. Correspondingly, component correlation is based on the power grid component topology diagram, showing the electrical coupling, upstream and downstream dependencies, and cascading effects between various power grid components. It reflects whether a fault in a single component will affect other components and is the core basis for classifying fault groups. Component correlation can include: upstream and downstream series relationships: such as main line → transformer → switchgear, where an upstream fault directly affects the downstream; parallel relationships at the same level: parallel lines and backup components, mutually redundant, and faults easily affect each other; regional correlation relationships: components in the same region or within the same disaster impact range have similar fault risks. Correspondingly, the power topology network can be represented as... ,in It can represent a set of nodes. It can represent a set of edges.

[0044] Furthermore, based on the strength of component relationships and fault propagation paths, groups of components with strong correlations, where a failure of one component can easily trigger a chain reaction of failures, can be classified into the same fault component set. Based on the influence range of the fault component set, multiple fault component sets can be identified as primary fault groups, such as trunk line fault groups or urban substation equipment fault groups, to facilitate subsequent fault analysis and probability calculations based on the set.

[0045] Optionally, step 102 may specifically include: inputting multidimensional heterogeneous data into a preset component prediction network; using the Transformer network in the preset component prediction network to extract multidimensional heterogeneous features corresponding to the multidimensional heterogeneous data and disaster spatiotemporal features corresponding to the disaster type, generating fused features corresponding to the power grid; using the quantum neural network in the preset component prediction network to perform angle transformation and quantum entanglement processing on the fused features, generating component resilience indicators corresponding to the power grid components, the component resilience indicators including resilience-criticality measures corresponding to each power grid component, used to assess the sensitivity of the power grid components to the power grid resilience indicators.

[0046] Optionally, a training set corresponding to a preset component prediction network can be obtained first. The preset component prediction network can then be trained based on this training set. The trained preset component prediction network can then be used to evaluate the grid resilience index, obtaining the resilience-criticality measure of each part of the grid during a disaster. The training set may include the historical disaster types corresponding to the grid, the power losses corresponding to the historical disaster types, and the component resilience indexes corresponding to the historical disaster types.

[0047] In some embodiments, different power grid components have varying degrees of impact on power grid resilience. The impact of power grid components on power grid resilience can be quantified by applying sensitivity analysis methods, i.e., by using component resilience indices for quantitative assessment. This can be obtained by calculating the partial derivative of the power grid resilience index with respect to the component failure probability under the current disaster type. ; Among them, components It is except components Other damaged components Indicators of grid resilience for grid components The sensitivity coefficient of the component failure probability. The higher the value, the better the grid resilience of the components. The more sensitive it is to changes in the probability of damage.

[0048] For example, the grid resilience index can be used as the input to the Transformer network in the preset component prediction network, and the component resilience index can be used as the output of the quantum neural network in the preset component prediction network for model training, and the following steps can be used for encoding: Positional encoding: In addition to word embeddings, Transformers also need to use positional embeddings to represent the position of words in a sentence. Because Transformers do not use an RNN structure but instead rely on global information, they cannot utilize word order information, which is crucial for NLP. Therefore, Transformers use positional embeddings to store the relative or absolute positions of words in the sequence. Location Embedding express, The dimension is the same as that of the word Embedding. It can be obtained through training or calculated using a certain formula. The Transformer uses the latter, and the calculation formula is as follows: ; in, Indicates the position of a word in a sentence. express The dimension (same as word embedding), The dimension representing an even number. This indicates odd-numbered dimensions. This formula is used for calculation. Can make It can adapt to sentences longer than any sentence in the training set. For example, if the longest sentence in the training set is 20 words, and a sentence of length 21 suddenly appears, the 21st embedding can be calculated using a formula. It also allows the model to easily calculate relative positions for fixed-length spacing. , It can be used It was calculated that, because , Add the word embedding and the position embedding of the word to obtain the word representation vector. , This is the final input to the Transformer; then, c is input into a pre-defined component prediction network for training, where the Transformer's hierarchical structure is as follows: Figure 2 As shown, the multidimensional heterogeneous input data is read word by word and sentence by sentence through multiple encoder layers. The semantics of the multidimensional heterogeneous data text are compressed into a deep encoded information. Then, multiple decoders are used to parse the encoded information and generate fused features that conform to the reading format of the quantum neural network. Specifically, the structure diagram of the encoder and decoder in the Transformer network is shown below. Figure 3As shown, the encoder / decoder layers are repeatedly stacked N layers, extracting high-order features layer by layer. Inputs can be multidimensional heterogeneous data, outputs can be encoded information, and probabilities can be fused features, specifically including InputEmbedding, Positional Encoding, Multi-Head Attention, Add&Norm (residual connections + layer normalization), FeedForward, OutputEmbedding, Masked Multi-Head Attention, Linear, Softmax activation function, etc.

[0049] Optionally, the fused features are transformed by angle conversion and quantum entanglement processing using a quantum neural network in a preset component prediction network to generate a component resilience index corresponding to the power grid component. Specifically, this may include: using a quantum neural network to convert the fused features into quantum state representations through angle encoding; and using a quantum entanglement layer in the quantum neural network to perform feature interaction based on the quantum state representations to generate a component resilience index corresponding to the power grid component.

[0050] For example, a schematic diagram of a quantum neural network structure is shown below. Figure 4 As shown, fused features can be transformed through angular encoding, such as through... The gate encodes the fused features into quantum state representations, which are then passed through two strongly entangled layers. Indicator prediction is performed. In another circuit with a strongly entangled layer, the circuit consists of angle encoding, followed by a first rotation sequence and a CNOT gate.

[0051] Optionally, step 103 may specifically include: sorting the component resilience indices corresponding to the power grid components, and using a clustering algorithm to determine the set of enhanced components corresponding to the disaster type based on the sorting results and the component correlation between the power grid components.

[0052] For example, components can be sorted according to their resilience indicators and classified using clustering algorithms (such as K-Means or DBSCAN) based on the component relationships in the power grid component topology diagram. This results in different clusters, such as a high-indicator set with high indicator values ​​and strong component correlation, a medium-indicator set with moderate indicator values ​​and strong component correlation, and a low-indicator set with low indicator values ​​and weak component correlation. Optionally, target component sets can be determined based on the classification labels of the enhanced component sets, such as high-indicator set, medium-indicator set, and low-indicator set. For example, the high-indicator set can be used as the target component set for timely disaster resistance reinforcement. By calculating and sorting the key resilience indicators of all components in the power grid, critical components requiring special attention and priority reinforcement can be identified.

[0053] As one possible implementation method, such as Figure 5 As shown, the initial total power supply capacity of the power grid before the fault can be calculated based on the power grid component topology before and after the fault recovery caused by the current disaster type. ), and the remaining power supply capacity of the restored power grid. ), using formula Calculate the power supply loss corresponding to the disaster type. (and combined with disaster type) The probability of natural disasters ), disaster types The following set of failure components (Probability of failure set) s The probability of failure Disaster type iThe calculation can be iteratively performed from 1 to m. Combining fault probability and power supply loss, the Power Grid Resilience Metric is obtained using the formula for calculating the Power Grid Resilience Metric. Then, using the formula for calculating the Component Criticality Metric, the corresponding Component Criticality Metric is calculated. This metric serves as a training sample for the model. It is input into a pre-defined component prediction network composed of a Transformer and a quantum neural network, outputting a ranking result of the Component Criticality Metric. This ranking table includes data such as component number, fault count ranking, and criticality value, quantifying the vulnerability and priority of each component. Finally, the K-means clustering algorithm is used to group the components to be hardened in the table based on their criticality metrics and topological relationships, resulting in... k Several clusters, as a set of enhancement components (Allocation of hardening budgets), such as Cluster 1. Cluster 2. Cluster 3. Cluster 4, and through , , They represent the 1st, 2nd, 3rd, and 4th respectively. k Vulnerability increments for each cluster are used to optimize the overall assessment process, identify target enhancement sets from the set of enhancement components, improve the efficiency and accuracy of critical component identification, and thus enhance the disaster resilience of the power grid.

[0054] Among them, the Transformer large-scale model, based on self-attention mechanism and multi-head attention architecture, breaks through the structural limitations of traditional neural networks. It possesses powerful capabilities in complex relationship modeling, multi-dimensional information fusion, and dynamic feature capture, especially in representing complex input-output relationships in power grid-natural disaster coupling systems. It can effectively compensate for the technical deficiencies of Monte Carlo methods and simple neural networks in power grid disaster resilience assessment and key component identification. The Transformer large-scale model, relying on self-attention mechanism, multi-head attention architecture, and encoder-decoder structure, has a powerful ability to represent complex input-output relationships. It can effectively solve the core problems of strong coupling of multi-dimensional features, significant spatiotemporal dynamic correlation, and interlocking of multiple output targets in power grid-natural disaster coupling systems, thus overcoming the technical limitations of Monte Carlo methods and simple neural networks. By breaking local constraints through a self-attention mechanism, it performs global correlation analysis on multi-dimensional heterogeneous input data, including power grid component parameters, spatiotemporal disaster characteristics, and environmental factors. This quantifies the strength of direct and indirect correlations between features, accurately uncovering deep nonlinear coupling patterns such as disaster intensity and component failure probability, and cross-regional component failure cascading effects. Simultaneously, it leverages multi-head attention to learn feature correlation patterns at different scales in parallel, improving the utilization rate of input feature information and avoiding assessment biases caused by insufficient feature correlation mining. Combining a location encoding mechanism and an encoder-decoder collaborative architecture, Transformer efficiently fuses the temporal patterns of disaster evolution and fault propagation with the spatial distribution of power grid components and cross-regional correlation features, achieving accurate modeling of dynamic spatiotemporal information. Furthermore, through global feature fusion and multi-head attention redundancy verification mechanisms, it enhances its resistance to data uncertainties such as disaster prediction errors and parameter measurement deviations, ensuring the stability of assessment results in dynamic scenarios. At the output end, its flexible decoder and cross-attention linkage mechanism can achieve collaborative modeling of multi-dimensional assessment objectives such as power supply reliability, repair difficulty, economic loss, and social impact. It accurately depicts the inherent trade-offs between various output indicators, abandons the one-sidedness of simple neural networks with single output or independent multiple outputs, and constructs a full-link representation system of "input global correlation - spatiotemporal dynamic modeling - multi-output collaborative optimization". At the same time, with its generalization ability and sparse optimization architecture, it can adapt to multi-disaster superposition scenarios and the needs of power grids of different scales. While taking into account the assessment efficiency and accuracy, it provides accurate and comprehensive technical support for the assessment of key power grid components, and fully meets the high-precision representation requirements of input-output relationships in complex coupled systems.

[0055] Correspondingly, quantum neural networks (QNNs), as a fusion of quantum computing and traditional neural networks, possess a core advantage in that they rely on the inherent randomness of quantum mechanics. This allows them to precisely adapt to and cancel out the multiple uncontrollable randomnesses inherent in power grid disaster scenarios, while simultaneously balancing computational efficiency and assessment accuracy. This provides breakthrough support for power grid disaster resilience assessment and key component identification. The superposition state of qubits brings intrinsic randomness to quantum gate operations due to quantum fluctuations and quantum entanglement. This randomness differs from the passively generated sampling randomness of Monte Carlo methods, possessing the characteristics of being controllable, superimposed, and interferable. It is perfectly suited to the complex randomness in power grid disaster scenarios, including fluctuations in natural disaster intensity, uncertain propagation paths, random component failure probabilities, unpredictable chain reactions of failures, and intertwined data measurement biases and noise interference. Through the "hedging and cancellation" effect, QNNs play a core role. On the one hand, they actively simulate various random variables using quantum superposition states, simultaneously covering multiple random scenarios without massive sampling. On the other hand, they leverage quantum entanglement... By encoding the stochastic correlation between disasters, component failures, and grid state responses, the model training and inference processes are naturally integrated, eliminating the need for constructing additional complex stochastic models. Furthermore, quantum interference suppresses invalid random noise such as data measurement biases and disaster-irrelevant fluctuations, while simultaneously enhancing effective random signals such as the core impact patterns of disasters on key components. This achieves an integrated approach to stochastic adaptation, interference cancellation, and effective information extraction, completely overcoming the dilemma of traditional methods in balancing error reduction and control time consumption. This characteristic not only eliminates the need for passive sampling in quantum neural networks but also transforms the uncontrollable randomness of disaster scenarios into effective inputs. Combined with its inherent parallel computing capabilities, strong ability to characterize complex correlations, and high robustness, it effectively addresses the core shortcomings of traditional Monte Carlo methods, such as excessive computation time, insufficient accuracy, and poor adaptability of simple neural networks. Even in complex stochastic scenarios, it can stably output high-precision evaluation results, providing efficient, accurate, and stable technical support for grid disaster resilience assessment and key component identification, significantly enhancing the grid's disaster resilience capabilities.

[0056] Compared with related technologies, this embodiment can determine the component association relationship between power grid components based on the power grid component topology diagram, and determine the power grid resilience index corresponding to the fault type based on the power loss after the cascading fault repair corresponding to the fault type caused by the disaster type and the fault probability corresponding to the fault type. By using the Transformer network in the preset component prediction network, multidimensional heterogeneous features corresponding to multidimensional heterogeneous data and disaster spatiotemporal features corresponding to the disaster type are extracted to generate the fused features corresponding to the power grid. By using the quantum neural network in the preset component prediction network, the fused features are subjected to angle transformation and quantum entanglement processing to generate the component resilience index corresponding to the power grid component. Based on the index, multiple enhanced component sets are obtained, and the target component set is selected for component enhancement processing. The temporal laws of disaster evolution and fault propagation in the time dimension are fused with the distribution and cross-regional association features of power grid components in the spatial dimension to achieve accurate modeling of dynamic spatiotemporal information. At the same time, through global feature fusion and multi-head attention redundancy verification mechanism, the anti-interference ability of data uncertainty such as disaster prediction error and parameter measurement deviation is enhanced to ensure the stability of the evaluation results in dynamic scenarios.

[0057] Furthermore, embodiments of this application provide a power grid disaster resilience assessment device for responding to natural disasters, such as... Figure 6 As shown, the device includes: an acquisition module 31, an evaluation module 32, a determination module 33, and an enhancement module 34.

[0058] The acquisition module 31 is configured to acquire multidimensional heterogeneous data corresponding to the power grid. The multidimensional heterogeneous data includes power grid component data, disaster data, and power grid environmental factors. The power grid component data includes power grid resilience indicators, fault types in fault groups, fault probabilities corresponding to fault types, and power losses corresponding to fault types. The disaster data includes disaster types. The evaluation module 32 is configured to use a preset component prediction network to evaluate the component resilience index of the power grid components based on multidimensional heterogeneous data. The preset component prediction network is used to fuse multidimensional heterogeneous input data to obtain the fault propagation time-series characteristics of the disaster type impact in the time dimension and the power grid component correlation characteristics in the spatial dimension. Based on the fault propagation time-series characteristics and the power grid component correlation characteristics, the component resilience index is evaluated. Module 33 is configured to classify faulty components based on component resilience indicators and determine the set of enhanced components corresponding to the disaster type. Enhancement module 34 is configured to determine a target component set from the enhanced component set and perform component enhancement processing on the target component set.

[0059] In some embodiments, the acquisition module 31 is specifically configured to acquire the power loss after repair of the cascading faults corresponding to the fault type caused by the disaster type; and determine the power grid resilience index corresponding to the fault type based on the power loss and the fault probability corresponding to the fault type.

[0060] In some embodiments, the acquisition module 31 is further configured to acquire the probability of occurrence of a disaster corresponding to a disaster type, and the set of faulty components corresponding to a disaster type; and to determine the power grid resilience index of the set of faulty components under a disaster type based on the component failure probability and the probability of occurrence of a disaster of each faulty component in the set of faulty components, as well as the power loss.

[0061] In some embodiments, the evaluation module 32 is specifically configured to input multidimensional heterogeneous data into a preset component prediction network; use the Transformer network in the preset component prediction network to extract the multidimensional heterogeneous features corresponding to the multidimensional heterogeneous data and the spatiotemporal features of the disaster type to generate the fused features corresponding to the power grid; use the quantum neural network in the preset component prediction network to perform angle transformation and quantum entanglement processing on the fused features to generate the component resilience index corresponding to the power grid component. The component resilience index includes the resilience-criticality metric value corresponding to each power grid component, which is used to evaluate the sensitivity of the power grid component to the power grid resilience index.

[0062] In some embodiments, the evaluation module 32 is specifically configured to use a quantum neural network to convert the fused features into quantum state representations through angle encoding; and to use the quantum entanglement layer in the quantum neural network to perform feature interactions based on the quantum state representations to generate a component resilience index corresponding to the power grid component.

[0063] In some embodiments, the acquisition module 31 is further configured to acquire a power grid component topology diagram of the power grid; determine power grid components and component association relationships between power grid components based on the power grid component topology diagram; determine the primary fault group family corresponding to the power grid based on the component association relationships; and determine the power grid component data corresponding to the power grid component in the fault component set based on the fault component set corresponding to the fault type in the primary fault group family.

[0064] In some embodiments, the determining module 33 is further configured to sort the component resilience indices corresponding to the power grid components, and determine the set of enhanced components corresponding to the disaster type using a clustering algorithm based on the sorting results and the component association relationships between the power grid components.

[0065] It should be noted that other corresponding descriptions of the functional units involved in the power grid disaster resilience assessment device for responding to natural disasters provided in this application embodiment can be found in the following references. Figure 1 The corresponding description in [the document] will not be repeated here.

[0066] Based on the above, Figure 1 As illustrated in the example, correspondingly, embodiments of this application also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described... Figure 1 The example method shown.

[0067] Based on the above, Figure 1 As illustrated, correspondingly, embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements the above-described... Figure 1 The example method shown.

[0068] Based on this understanding, the technical solutions of the embodiments of this application can be embodied in the form of a software product. The software product can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0069] Based on the above, Figure 1 The method shown, and Figure 6 To achieve the above objectives, the present application also provides an electronic device, comprising a storage medium and a processor; the storage medium for storing a computer program; and the processor for executing the computer program to implement the above-described virtual device embodiments. Figure 1 The method shown.

[0070] Optionally, the aforementioned electronic device may also include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, an input unit, etc.

[0071] Those skilled in the art will understand that the physical device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0072] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0073] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms, or it can be implemented by hardware. This application can determine the component association relationship between power grid components based on the power grid component topology diagram, and determine the power grid resilience index corresponding to the fault type based on the power loss after the cascading fault repair corresponding to the fault type caused by the disaster type and the fault probability corresponding to the fault type. Using the Transformer network in the preset component prediction network, multidimensional heterogeneous features corresponding to multidimensional heterogeneous data and disaster spatiotemporal features corresponding to the disaster type are extracted to generate the fusion features corresponding to the power grid. Using the quantum neural network in the preset component prediction network, the fusion features are subjected to angle transformation and quantum entanglement processing to generate the component resilience index corresponding to the power grid component. Based on the index, multiple enhanced component sets are obtained, and the target component set is selected from them for component enhancement processing. The temporal laws of disaster evolution and fault propagation in the time dimension and the distribution and cross-regional association features of power grid components in the spatial dimension are fused to achieve accurate modeling of dynamic spatiotemporal information. At the same time, through global feature fusion and multi-head attention redundancy verification mechanism, the anti-interference ability of data uncertainty such as disaster prediction error and parameter measurement deviation is enhanced, ensuring the stability of the evaluation results in dynamic scenarios.

[0074] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the term "comprising" or any other variations thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0075] The above are merely specific embodiments of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to these embodiments, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for assessing the disaster resilience of power grids in response to natural disasters, characterized in that, include: Acquire multidimensional heterogeneous data corresponding to the power grid. The multidimensional heterogeneous data includes power grid component data, disaster data, and power grid environmental factors. The power grid component data includes power grid resilience indicators, fault types in fault groups, fault probabilities corresponding to fault types, and power losses corresponding to fault types. The disaster data includes disaster types. Using a preset component prediction network, the component resilience index corresponding to the power grid component in the power grid is evaluated based on the multidimensional heterogeneous data. The preset component prediction network is used to fuse the multidimensional heterogeneous input data to obtain the fault propagation time series characteristics of the disaster type impact in the time dimension and the power grid component correlation characteristics in the spatial dimension. The component resilience index is evaluated based on the fault propagation time series characteristics and the power grid component correlation characteristics. Based on the component resilience index, the faulty components are classified and the set of reinforcement components corresponding to the disaster type is determined. A target component set is determined from the enhanced component set, and the target component set is subjected to component enhancement processing.

2. The method according to claim 1, characterized in that, The acquisition of multidimensional heterogeneous data corresponding to the power grid includes: Obtain the power loss after repairing the cascading faults corresponding to the fault types caused by the disaster type; Based on the power loss and the fault probability corresponding to the fault type, the power grid resilience index corresponding to the fault type is determined.

3. The method according to claim 2, characterized in that, The step of determining the power grid resilience index corresponding to the fault type based on the power loss and the fault probability corresponding to the fault type includes: Obtain the probability of occurrence of the disaster corresponding to the disaster type, and the set of faulty components corresponding to the disaster type; Based on the component failure probability of each faulty component in the faulty component set, the probability of the disaster occurrence, and the power loss, the power grid resilience index of the faulty component set under the disaster type is determined.

4. The method according to claim 1, characterized in that, The method of using a preset component prediction network to evaluate the component resilience index corresponding to the power grid components in the power grid based on the multidimensional heterogeneous data and the disaster type includes: The multidimensional heterogeneous data is input into the preset component prediction network; Using the Transformer network in the preset component prediction network, multidimensional heterogeneous features corresponding to the multidimensional heterogeneous data and disaster spatiotemporal features corresponding to the disaster type are extracted to generate fused features corresponding to the power grid. Using the quantum neural network in the preset component prediction network, the fusion features are subjected to angle transformation and quantum entanglement processing to generate the component resilience index corresponding to the power grid component. The component resilience index includes the resilience-criticality metric value corresponding to each power grid component, which is used to evaluate the sensitivity of the power grid component to the power grid resilience index.

5. The method according to claim 4, characterized in that, The step of using the quantum neural network in the preset component prediction network to perform angle transformation and quantum entanglement processing on the fused features to generate the component resilience index corresponding to the power grid component includes: Using the quantum neural network, the fused features are converted into quantum state representations through angle encoding; By utilizing the quantum entanglement layer in the quantum neural network, feature interactions are performed based on the quantum state representation to generate the component resilience index corresponding to the power grid component.

6. The method according to claim 1, characterized in that, The acquisition of multidimensional heterogeneous data corresponding to the power grid also includes: Obtain the power grid component topology diagram of the power grid; The power grid components and their associated relationships are determined based on the power grid component topology diagram. Based on the component associations, the primary fault group family corresponding to the power grid is determined; Based on the set of fault components corresponding to the fault types in the first-level fault group, determine the power grid component data corresponding to the power grid components in the set of fault components.

7. The method according to claim 6, characterized in that, The step of classifying faulty components based on the component resilience index to determine the set of enhanced components corresponding to the disaster type includes: The component resilience indices corresponding to the power grid components are sorted, and based on the sorting results and the component correlations between the power grid components, a clustering algorithm is used to determine the set of enhanced components corresponding to the disaster type.

8. A power grid disaster resilience assessment device for responding to natural disasters, characterized in that, include: The acquisition module is configured to acquire multidimensional heterogeneous data corresponding to the power grid. The multidimensional heterogeneous data includes power grid component data, disaster data, and power grid environmental factors. The power grid component data includes power grid resilience indicators, fault types in fault groups, fault probabilities corresponding to fault types, and power losses corresponding to fault types. The disaster data includes disaster types. The evaluation module is configured to use a preset component prediction network to evaluate the component resilience index of the power grid components in the power grid based on the multidimensional heterogeneous data. The preset component prediction network is used to fuse the multidimensional heterogeneous input data to obtain the fault propagation time-series characteristics of the disaster type impact in the time dimension and the power grid component correlation characteristics in the spatial dimension, and evaluate the component resilience index based on the fault propagation time-series characteristics and the power grid component correlation characteristics. The determination module is configured to classify faulty components based on the component resilience index and determine the set of reinforcement components corresponding to the disaster type. An enhancement module is configured to determine a target component set from the enhanced component set and perform component enhancement processing on the target component set.

9. 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 method of any one of claims 1 to 7.

10. An electronic device comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

Images