Electric power system vulnerability assessment method and system considering extreme weather influence

Abstract

The invention relates to the technical field of power system assessment, and discloses a power system vulnerability assessment method and system considering extreme weather influence, and the method comprises the steps: building a line fault probability model of a power system in extreme weather to determine the line time-varying fault probability under dynamic weather influence; generating a typical scene through a time sequence relativity matching generative adversarial network model with gradient penalty and Monte Carlo simulation; taking the generated typical scene as input, simulating cascading failure evolution through an alternating current cascading failure model, and outputting a cascading failure evolution path; and based on a fault evolution result, through a spectrogram theory and a Bayesian network, in combination with severity analysis, structural vulnerability evaluation and causal-probability pre-judgment, multi-level vulnerability quantification of the power system is realized, and the problems that single-level evaluation is insufficient and causal association is ignored in the prior art are solved. And the risk assessment precision and the disaster prevention capability of the power system in extreme weather are improved.

Description

Technical Field

[0001] This invention relates to the field of power system assessment technology, and in particular to a method and system for assessing the vulnerability of power systems that takes into account the impact of extreme weather. Background Technology

[0002] The impact of extreme weather on the power system is characterized by its wide-ranging and multi-layered effects: from the perspective of power source-load balance, extreme weather reduces wind and solar power output while triggering a surge in loads during extreme heat / cold weather; from the perspective of power grid structure, it increases the failure rate of power poles and lines, thereby triggering cascading failures. Therefore, conducting vulnerability assessments of the power system in response to the impacts of extreme weather is of great significance.

[0003] Currently, power system vulnerability is divided into structural vulnerability (network topology importance) and state vulnerability (deviation of state variables). However, existing assessment methods mostly adopt single-level assessment (only analyzing equipment failures or topology changes), which easily misses local weak points, leading to underestimation of risk and misjudgment. At the same time, the scarcity of renewable energy and load samples under extreme weather conditions makes it difficult to accurately construct source-load output scenarios, which also affects the accuracy of the assessment.

[0004] Therefore, improving the accuracy of vulnerability assessment of power systems under extreme weather conditions has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] This invention provides a method and system for assessing the vulnerability of power systems considering the impact of extreme weather, addressing the problem of how to improve the accuracy of vulnerability assessment of power systems under extreme weather conditions.

[0006] To address the aforementioned technical problems, the first aspect of this invention provides a method for assessing the vulnerability of power systems considering the impact of extreme weather, comprising: The topology data of the power system and its meteorological data and source-load data under extreme weather conditions are obtained, and a line fault probability model of the power system is constructed based on the topology data and the meteorological data to calculate the time-varying fault probability of the line. Based on the source load data, the source load output sample set under extreme weather conditions is expanded by a time series relative matching generative adversarial network model with gradient penalty, and the expansion result is combined with the time-varying fault probability of the line to generate typical scenarios through Monte Carlo simulation. Using the aforementioned typical scenario as input, the cascading failure evolution process under line faults and source load fluctuations is simulated through an AC cascading failure model, and the cascading failure evolution path and system operating status at termination are output. Based on the cascading failure evolution path and the system operating state at the time of termination, a first-level surface vulnerability analysis is performed by constructing a severity model, a second-level intrinsic structural vulnerability analysis is performed using spectral theory, and a third-level cascading failure causal probability analysis is performed based on Bayesian networks to obtain the multi-level vulnerability assessment results of the power system.

[0007] As one preferred embodiment, the step of constructing a line fault probability model of the power system based on the topology data and the meteorological data to calculate the time-varying line fault probability includes: A meteorological impact model is constructed based on the meteorological data, and geographical data of each transmission line in the power system is obtained. The meteorological impact model is then corrected using the geographical data to obtain the time-series corrected wind speed for each transmission line. Based on the time-series corrected wind speed and the topology data, an overall failure probability model for each of the transmission lines is constructed, and the service life of each of the overall failure probability models is corrected to obtain the line fault probability model of the power system in order to quantify the time-varying fault probability of the lines.

[0008] As one preferred embodiment, the expansion of the source load output sample set under extreme weather conditions based on the source load data using a time-series relative matching generative adversarial network model with gradient penalty includes: Features are extracted from the source load data and the meteorological data respectively to obtain a feature set, which is then used to reduce the dimensionality of the feature set through an embedding function to generate low-dimensional features. The low-dimensional features are reconstructed using a recovery function to obtain high-dimensional features, and the reconstruction loss between the low-dimensional features and the high-dimensional features is calculated. A feature vector set is constructed by random sampling, and the feature vector set is input into a generator for processing to obtain sample features. The supervision loss between the sample features and the low-dimensional features is calculated. The sample features and the low-dimensional features are input together into the discriminator for processing to obtain the sample probability and the true probability, and the gradient-penalized relative matching unsupervised loss between the sample probability and the true probability is calculated. Based on the reconstruction loss, the supervision loss, and the relative matching unsupervised loss, a total loss function is constructed. With the goal of minimizing the total loss function, the generator and the discriminator are alternately optimized through the backpropagation algorithm until a preset number of training iterations are reached. The optimized generator is then used to expand the source load output sample set under extreme weather conditions, resulting in an expanded extreme weather source load output sample set as the expansion result.

[0009] As one preferred embodiment, the step of combining the augmented results with the time-varying fault probability of the line to generate typical scenarios through Monte Carlo simulation includes: Based on the time-varying fault probability of the line and the extended sample set of extreme weather source load output, several fault scenarios are obtained by constructing multi-time-period scenarios through sequential Monte Carlo. For each of the aforementioned fault scenarios, the power loss of the scenario is calculated as a clustering feature, and based on the clustering feature, the K-means clustering algorithm is used to cluster the fault scenarios to obtain typical scenarios.

[0010] As one preferred embodiment, the process of simulating the cascading failure evolution under line faults and source-load fluctuations using an AC cascading failure model, taking the typical scenario as input, and outputting the cascading failure evolution path and the system operating state at termination, includes: Based on the typical scenario and the topology data, a node power balance equation is constructed to perform optimal AC power flow calculation and obtain node voltage and line power flow data. The state of the power system is checked based on the node voltage and the line power flow data. When the check result exceeds the limit, a protection mechanism is triggered to update the typical scenario. Based on the updated typical scenario, iteratively execute the optimal AC power flow calculation steps and system state verification steps until the preset iteration termination condition is met, and output the cascading failure evolution path and the system running state at the time of termination.

[0011] As one preferred embodiment, the step of analyzing primary surface weaknesses by constructing a severity model based on the cascading failure evolution path and the system operating state at the time of termination, and analyzing secondary internal structural weaknesses using spectral theory, includes: The line's rated capacity percentage is determined based on the system's operating state at the time of termination. A line overload severity model is constructed based on the line's rated capacity percentage, and a node low-voltage severity model is constructed based on the system's operating state at the time of termination. Based on the line overload severity model and the node low voltage severity model, the expected risk corresponding to the surface weak link of the power system is quantified to obtain the first-level analysis results. The power flow weights of each transmission line are determined by the system operating state at the time of termination, in order to construct a weight matrix, and a Laplace matrix is ​​constructed based on the weight matrix. Based on the topology data and the cascading failure evolution path, a network adjacency matrix is ​​constructed and combined with the Laplace matrix to quantify the network structural vulnerability index corresponding to the inherent structural weaknesses of the power system, thereby obtaining secondary analysis results.

[0012] As one preferred embodiment, the three-level cascading failure causal probability analysis based on Bayesian networks includes: Based on the time-series corrected wind speed, the power flow data of the line, and the cascading fault evolution path, wind speed nodes, power flow mutation nodes, and fault nodes are determined respectively. Edges are set based on the dependency relationship of the corresponding fault mechanism between the wind speed nodes, the power flow mutation nodes, and the fault nodes to construct a Bayesian network. A training dataset is constructed based on the wind speed node, the tidal current change node, and the fault node. The conditional probability of each node in the Bayesian network is trained using the training dataset to obtain the trained Bayesian network. The real-time power flow data and real-time meteorological data of the power system are input into the trained Bayesian network for processing, so as to realize the causal probability prediction of the cascading failures of the power system and obtain the three-level analysis results.

[0013] A second aspect of the present invention provides a power system vulnerability assessment system that takes into account the impact of extreme weather, comprising: The probability calculation module is used to acquire the topology data of the power system and its meteorological data and source-load data under extreme weather conditions, and to construct the line fault probability model of the power system based on the topology data and the meteorological data in order to calculate the time-varying fault probability of the line. The scenario generation module is used to expand the source load output sample set under extreme weather conditions based on the source load data by using a time series relative matching generative adversarial network model with gradient penalty, and combine the expansion result with the time-varying fault probability of the line to generate typical scenarios through Monte Carlo simulation. The fault evolution module is used to simulate the cascading fault evolution process under line faults and source load fluctuations through an AC cascading fault model, taking the typical scenario as input, and outputting the cascading fault evolution path and the system operating status at the time of termination. The multi-level assessment module is used to perform first-level surface vulnerability analysis by constructing a severity model based on the cascading failure evolution path and the system operating state at the time of termination, second-level internal structural vulnerability analysis by using spectral theory, and third-level cascading failure causal probability analysis based on Bayesian networks, so as to obtain the multi-level vulnerability assessment results of the power system.

[0014] As one preferred embodiment, the step of constructing a line fault probability model of the power system based on the topology data and the meteorological data to calculate the time-varying line fault probability includes: A meteorological impact model is constructed based on the meteorological data, and geographical data of each transmission line in the power system is obtained. The meteorological impact model is then corrected using the geographical data to obtain the time-series corrected wind speed for each transmission line. Based on the time-series corrected wind speed and the topology data, an overall failure probability model for each of the transmission lines is constructed, and the service life of each of the overall failure probability models is corrected to obtain the line fault probability model of the power system in order to quantify the time-varying fault probability of the lines.

[0015] As one preferred embodiment, the expansion of the source load output sample set under extreme weather conditions based on the source load data using a time-series relative matching generative adversarial network model with gradient penalty includes: Features are extracted from the source load data and the meteorological data respectively to obtain a feature set, which is then used to reduce the dimensionality of the feature set through an embedding function to generate low-dimensional features. The low-dimensional features are reconstructed using a recovery function to obtain high-dimensional features, and the reconstruction loss between the low-dimensional features and the high-dimensional features is calculated. A feature vector set is constructed by random sampling, and the feature vector set is input into a generator for processing to obtain sample features. The supervision loss between the sample features and the low-dimensional features is calculated. The sample features and the low-dimensional features are input together into the discriminator for processing to obtain the sample probability and the true probability, and the gradient-penalized relative matching unsupervised loss between the sample probability and the true probability is calculated. Based on the reconstruction loss, the supervision loss, and the relative matching unsupervised loss, a total loss function is constructed. With the goal of minimizing the total loss function, the generator and the discriminator are alternately optimized through the backpropagation algorithm until a preset number of training iterations are reached. The optimized generator is then used to expand the source load output sample set under extreme weather conditions, resulting in an expanded extreme weather source load output sample set as the expansion result.

[0016] As one preferred embodiment, the step of combining the augmented results with the time-varying fault probability of the line to generate typical scenarios through Monte Carlo simulation includes: Based on the time-varying fault probability of the line and the extended sample set of extreme weather source load output, several fault scenarios are obtained by constructing multi-time-period scenarios through sequential Monte Carlo. For each of the aforementioned fault scenarios, the power loss of the scenario is calculated as a clustering feature, and based on the clustering feature, the K-means clustering algorithm is used to cluster the fault scenarios to obtain typical scenarios.

[0017] As one preferred embodiment, the fault evolution module includes: Based on the typical scenario and the topology data, a node power balance equation is constructed to perform optimal AC power flow calculation and obtain node voltage and line power flow data. The state of the power system is checked based on the node voltage and the line power flow data. When the check result exceeds the limit, a protection mechanism is triggered to update the typical scenario. Based on the updated typical scenario, iteratively execute the optimal AC power flow calculation steps and system state verification steps until the preset iteration termination condition is met, and output the cascading failure evolution path and the system running state at the time of termination.

[0018] As one preferred embodiment, the step of analyzing primary surface weaknesses by constructing a severity model based on the cascading failure evolution path and the system operating state at the time of termination, and analyzing secondary internal structural weaknesses using spectral theory, includes: The line's rated capacity percentage is determined based on the system's operating state at the time of termination. A line overload severity model is constructed based on the line's rated capacity percentage, and a node low-voltage severity model is constructed based on the system's operating state at the time of termination. Based on the line overload severity model and the node low voltage severity model, the expected risk corresponding to the surface weak link of the power system is quantified to obtain the first-level analysis results. The power flow weights of each transmission line are determined by the system operating state at the time of termination, in order to construct a weight matrix, and a Laplace matrix is ​​constructed based on the weight matrix. Based on the topology data and the cascading failure evolution path, a network adjacency matrix is ​​constructed and combined with the Laplace matrix to quantify the network structural vulnerability index corresponding to the inherent structural weaknesses of the power system, thereby obtaining secondary analysis results.

[0019] As one preferred embodiment, the three-level cascading failure causal probability analysis based on Bayesian networks includes: Based on the time-series corrected wind speed, the power flow data of the line, and the cascading fault evolution path, wind speed nodes, power flow mutation nodes, and fault nodes are determined respectively. Edges are set based on the dependency relationship of the corresponding fault mechanism between the wind speed nodes, the power flow mutation nodes, and the fault nodes to construct a Bayesian network. A training dataset is constructed based on the wind speed node, the tidal current change node, and the fault node. The conditional probability of each node in the Bayesian network is trained using the training dataset to obtain the trained Bayesian network. The real-time power flow data and real-time meteorological data of the power system are input into the trained Bayesian network for processing, so as to realize the causal probability prediction of the cascading failures of the power system and obtain the three-level analysis results.

[0020] A third aspect of the present invention provides an electronic device including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the power system vulnerability assessment method considering the impact of extreme weather as described above.

[0021] A fourth aspect of the present invention provides a computer-readable storage medium comprising a stored computer program, wherein when the device containing the computer-readable storage medium executes the computer program, it implements the power system vulnerability assessment method considering the impact of extreme weather as described above.

[0022] Compared with the prior art, the beneficial effects of the embodiments of the present invention are as follows: By constructing a line fault probability model, the probability of faults on different lines under meteorological conditions can be considered more accurately, greatly improving the accuracy of line fault probability prediction. A gradient-penalized time-series relative matching generative adversarial network model and Monte Carlo simulation generate typical scenarios covering various operating states of the power system under extreme weather conditions, helping to comprehensively assess the vulnerability of the power system under different circumstances. An AC cascading fault model simulates the evolution process of cascading faults, clarifying the scope of fault impact and the final operating state of the system, providing important reference for formulating targeted preventative measures. Based on the fault evolution results, through spectral theory and Bayesian networks, combined with severity analysis, structural vulnerability assessment, and causal-probability prediction, a multi-level vulnerability assessment and quantification of the power system is achieved, encompassing the identification of comprehensive surface phenomena, analysis of internal structure, and revelation of deep causal relationships. This solves the problems of insufficient single-level assessment and neglect of causal relationships in existing technologies, realizing a comprehensive and accurate assessment of the vulnerability of the power system under extreme weather conditions. Attached Figure Description

[0023] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart of a power system vulnerability assessment method considering the impact of extreme weather, provided in a certain embodiment of the present invention; Figure 2This is a structural diagram of the Time-RpGAN-GP model provided in a certain embodiment of the present invention; Figure 3 This is a diagram illustrating the process of generating a typical scene using Monte Carlo simulation, provided in a certain embodiment of the present invention. Figure 4 This is an improved IEEE 14-node system architecture diagram provided in a certain embodiment of the present invention; Figure 5 This is a time-varying fault probability diagram of a transmission line provided in a certain embodiment of the present invention; Figure 6 This is a sample result of the extended extreme scenario source load output of the Time-RpGAN-GP model provided in a certain embodiment of the present invention; Figure 7 This is a power system severity distribution map at the initial moment of a typhoon's impact, provided by a certain embodiment of the present invention; Figure 8 This is a diagram showing the identification results of structurally weak branches in a system based on spectral analysis, provided in a certain embodiment of the present invention. Figure 9 This is a diagram showing the identification results of structurally weak nodes in a system based on spectral analysis, provided in a certain embodiment of the present invention. Figure 10 This is a Bayesian network structure diagram of an IEEE 14-node system provided in a certain embodiment of the present invention; Figure 11 This is a structural diagram of a power system vulnerability assessment system that takes into account the impact of extreme weather, provided in a certain embodiment of the present invention; Figure 12 This is a structural diagram of an electronic device provided in a certain embodiment of the present invention; Figure label: Among them, 10 is the probability calculation module; 20 is the scene generation module; 30 is the fault evolution module; 40 is the multi-level evaluation module; 5000 is the electronic device; 5001 is the processor; 5002 is the bus; 5003 is the memory; and 5004 is the transceiver. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings and examples. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0026] In this invention description, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In this invention description, unless otherwise stated, "a plurality of" means two or more. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items. Those skilled in the art will understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0027] In the description of this invention, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is merely for describing specific embodiments and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0028] In one embodiment, such as Figure 1 As shown, the first aspect of the present invention provides a method for assessing the vulnerability of power systems considering the impact of extreme weather, comprising: S1. Obtain the topology data of the power system and its meteorological data and source-load data under extreme weather conditions, and construct the line fault probability model of the power system based on the topology data and the meteorological data to calculate the time-varying fault probability of the line. Specifically, this invention addresses the impact of extreme weather (taking typhoons as an example) on power systems, employing a four-stage process: data collection, fault scenario generation, fault evolution analysis, and multi-dimensional vulnerability assessment. First, three types of core power system data are collected to provide a foundation for subsequent modeling and assessment: power system topology data includes transmission line parameters (conductor outer diameter, cross-sectional area, mass per unit length, mean and standard deviation of tensile strength of steel-cored aluminum stranded wire / steel single wire / aluminum single wire, etc.), tower parameters (top diameter, base diameter, mean and standard deviation of concrete bending strength, number of connecting conductors, average span, etc.), and node and branch connection relationships (power system busbars, line topology, etc.); meteorological data under extreme weather conditions includes typhoon landfall coordinates, movement speed, initial central pressure, maximum wind speed radius, hourly wind speed / direction data, etc.; and source-load data under extreme weather conditions includes wind power under typhoon influence (cut-in wind speed, cut-out wind speed, rated wind speed), photovoltaic output time-series data, and power load fluctuation data, etc. Then, the collected data undergoes preprocessing such as cleaning and standardization to improve data accuracy and facilitate subsequent use.

[0029] In one embodiment, the step of constructing a line fault probability model of the power system based on the topology data and the meteorological data to calculate the time-varying line fault probability includes: A meteorological impact model is constructed based on the meteorological data, and geographical data of each transmission line in the power system is obtained. The meteorological impact model is then corrected using the geographical data to obtain the time-series corrected wind speed for each transmission line. Based on the time-series corrected wind speed and the topology data, an overall failure probability model for each of the transmission lines is constructed, and the service life of each of the overall failure probability models is corrected to obtain the line fault probability model of the power system in order to quantify the time-varying fault probability of the lines.

[0030] Specifically, this invention takes typhoons as an example for detailed analysis, establishing a power system line failure probability model that combines the spatial propagation characteristics of typhoon weather with the power network topology. Typhoon disasters can lead to damage to equipment exposed to the external space in the power system, and can also cause drastic fluctuations in the output of new energy sources and loads, exacerbating the risk of power system supply-demand imbalance. Therefore, it is necessary to accurately simulate the wind speed and direction within the geographically defined area of ​​the power system during typhoon landfall, construct a power system transmission line failure probability model, and generate source-load output scenarios involving extreme meteorological factors.

[0031] This invention constructs a meteorological impact model based on meteorological data: After a typhoon makes landfall, it is assumed that the typhoon's landfall path is a straight line along the initial landfall angle, and the pressure difference at the typhoon's center gradually decreases over time, satisfying the following equation: In the formula, To determine the pressure difference at the center of the back wind; The central pressure difference at the initial moment of landfall; t represents the angle between the typhoon's direction of movement and the coordinate axis of the reference position; t represents the time after landfall.

[0032] The impact range of a typhoon after landfall can be represented by the radius of its maximum wind speed, which is the distance between the point of maximum wind speed and the center of the typhoon in the horizontal dimension, satisfying the following formula: In the formula, Let be the radius of the maximum wind speed at time t.

[0033] The gradient wind speed of a typhoon can be calculated from the pressure difference at the typhoon's center, specifically the gradient wind speed at the radius of the maximum wind speed at time t. for: In the formula, K is an empirical coefficient for wind speed calculation, which satisfies f is the Coriolis force parameter of Earth's rotation; This is the Earth's rotational angular velocity; This refers to the latitude of the typhoon's landfall location.

[0034] The maximum radius wind speed can be obtained based on the typhoon gradient wind speed. : In the formula, This represents the speed at which the typhoon moves.

[0035] By typhoon landfall point Construct a reference coordinate system with the origin. Assuming the typhoon moves along a straight line along the initial landfall angle after making landfall, then the position of the typhoon's center at time t is... satisfy: The wind speeds at other points within the typhoon's wind field, i.e., the meteorological impact model, can be obtained by calculating the distance between that point and the typhoon's center. In the formula, Let r be the wind speed at time t; Let t be the distance from that point to the center of the typhoon. This represents the maximum distance from this point to the center of the typhoon.

[0036] To improve the modeling accuracy of the wind field model, the Batts typhoon model is modified by considering topographic factors. Therefore, this invention acquires the geographical data of each transmission line in the power system, such as the terrain type (peaks, slopes, basins, valley mouths, etc.), key terrain parameters (total height of peaks, tangent of slope inclination angle, building height, etc.), and geographical location of the line (latitude and longitude coordinates of each line tower). Based on the acquired geographical dataset, a wind speed correction factor is proposed to modify the meteorological impact model for different terrain conditions. By multiplying the wind speed correction factor with the meteorological impact model, the time-series corrected wind speed of each transmission line can be obtained.

[0037] Among them, for mountain peaks and slopes, the correction factor It can be represented as: In the formula, is the tangent of the angle of inclination of the mountain peak or slope towards the windward side; when the tangent exceeds 0.3, it is taken as 0.3; k is a constant coefficient, taken as 2.2 for mountain peaks and 1.4 for slopes; H is the total height of the mountain peak or slope; z is the building height, taken as when the building height is greater than 2.5H. .

[0038] For enclosed terrain such as mountain basins and valleys Available The value is taken from the middle; for terrain such as valley mouths and mountain passes that are in the same direction as the strong winds. Available Take the value from the middle.

[0039] During typhoon landfall, the impact of typhoons on power transmission lines mainly consists of two parts. First, typhoon wind speeds exert tensile stress on the conductors; excessive tensile stress can cause conductor breakage. Second, excessive wind loads on the towers generate excessive bending moments, leading to tower fracture. Analyzing the failure states of power transmission lines under typhoon influence requires first analyzing the wind loads borne by the transmission lines. The wind load per unit length of the conductor is affected by horizontal wind loads and the conductor's own weight, which can be determined based on time-corrected wind speeds and power system topology data, expressed as: In the formula, F is the wind load per unit length of the conductor; The horizontal wind load on the conductor. V is the gravitational load on the conductor; V is the wind speed on the conductor, which is the time-corrected wind speed of the transmission line; D is the outer diameter of the conductor. If LGJ240 / 30 steel-cored aluminum stranded wire is selected, the outer diameter of the conductor is 21.60mm. The conductor non-uniformity coefficient is preferably 0.80; The conductor shape factor is preferably 1.1; The wind pressure variation coefficient is preferably 1.17; denoted as θ, where θ is the angle between the conductor and the wind pressure; M is the mass per unit length of the conductor; and g is the acceleration due to gravity.

[0040] Based on the wind load per unit length of the conductor and its actual shape, the stress on the conductor can be calculated. The stress between any two points on the conductor is related to the perpendicular distance between the two points. Generally, the highest suspension point of the conductor represents the maximum stress on the entire conductor; analyzing the line condition at the highest suspension point can help determine if the conductor has failed. (Stress at the highest suspension point of the conductor) It can be calculated as: In the formula, The tension at the highest suspension point; Where is the cross-sectional area of ​​the conductor; T is the tension at the lowest point of the conductor's sag; It is the horizontal distance between the highest point of the conductor and the lowest point of the sag.

[0041] Similar to conductors, the wind load on power poles during typhoons mainly comes from two parts. One part is the bending moment generated on the pole by the conductor under wind load, and the other part is the bending moment generated on the pole itself by the horizontal wind load of the typhoon. Generally speaking, the maximum bending moment of the pole is located at the base. Analyzing whether the base of the pole is fractured can determine the failure state of the pole. The bending moment at the base of the pole during a typhoon can be calculated as follows: In the formula, This generates a bending moment for the conductor at the base of the tower. The horizontal wind load of the typhoon generates a bending moment at the base of the tower; n is the number of conductors connected to the tower, which is determined by the grid structure. For the shape factor, a 12m concrete tower with a strength grade of G is selected, preferably 1.3; The average span between conductors; The lever arm of the conductor is located at the point of force application on the tower, extending to the base of the tower. The diameter of the tower tip; The diameter of the pole root is the pole diameter.

[0042] Therefore, the combined bending moment borne by the base of the tower is... and The vector sum.

[0043] When conducting reliability analysis on transmission line structures, it is necessary to compare the magnitude of the stress on the conductors, the magnitude of the bending moment on the towers, and the actual strength of the transmission line to determine its failure state. Therefore, the actual strength of the transmission line should also be analyzed. Existing conductors typically use steel-cored aluminum stranded wire, composed of steel and aluminum single wires twisted together. According to tensile testing, the tensile strength of steel-cored aluminum stranded wire follows a normal distribution with specific parameters, and its corresponding probability density function... This can be expressed as: In the formula, , , The values ​​are the average tensile strengths of steel-cored aluminum stranded wire, steel single wire, and aluminum single wire, respectively. The average tensile strengths of steel single wire and aluminum single wire are 1250MPa and 180MPa, respectively. , , These are the standard deviations of tensile strength for steel-cored aluminum stranded wire, steel single wire, and aluminum single wire, respectively. The standard deviations of tensile strength for steel single wire and aluminum single wire are 52.44 MPa and 12.91 MPa, respectively. , These are coefficients related to the physical properties of steel single-wire and aluminum single-wire, respectively. The wear coefficient is preferably 0.7; The coefficient for strength loss of aluminum single wire is preferably 0.95; This refers to the number of strands in a single aluminum wire. The cross-sectional area of ​​the aluminum single wire; The coefficient for single-line strength loss of steel is preferably 0.85; This refers to the number of strands in a single steel wire. It represents the cross-sectional area of ​​a single steel wire.

[0044] Most existing power poles are constructed from cast concrete in a cylindrical shape. Extensive failure tests on concrete power poles show that their bending strength also follows a normal distribution, with a probability density function... It can be represented as: In the formula, , The mean and standard deviation of the bending strength of the tower; , These are the amplification factor and the coefficient of variation, typically taken as 1.2 and 0.2 respectively. This represents the actual bending moment borne by the tower.

[0045] In summary, after calculating the conductor stress, inductive bending moment, and the strength of the components themselves, the failure rate of the conductor and tower is determined based on the probability density distribution function of the conductor and tower component strength. , They can be represented as: In the formula, , This represents the average strength of the conductor and tower components. , This represents the standard deviation of the strength of conductors and tower components.

[0046] A transmission line consists of multiple conductors and towers. While failures in each conductor and tower are independent events, the failure of any one component can cause the entire line to shut down. Therefore, the failure probability of a transmission line can be modeled as a series connection of multiple conductors and towers, i.e., the overall failure probability model of the entire transmission line. Represented as: In the formula, and They represent transmission lines respectively The number of conductors and towers.

[0047] Under normal weather conditions, the failure probability of transmission lines can be divided into three stages. In the initial stage, the transmission line is in a break-in period with other power equipment, and the failure probability is relatively high at this time. As the service time increases, the failure probability of the line decreases. In the stable stage, the transmission line operates stably, and the failure probability remains basically unchanged. In the wear and tear period, due to irreversible factors such as component aging, the failure probability of the transmission line continues to increase until the line is decommissioned.

[0048] Based on the analysis of transmission line failure probability under normal conditions, in the initial stage, the transmission line is running in with other equipment, but the strength of the transmission line components remains unchanged. During the stable operation period, the strength of the transmission line components remains basically unchanged. During the wear and tear period, the strength of the transmission line components gradually decreases, reducing its ability to withstand wind loads. Therefore, in the initial and stable operation stages, the depreciation due to service life does not need to be considered. However, during the wear and tear period, the transmission line failure probability needs to be adjusted for line damage.

[0049] By analyzing historical data on the failure probabilities of transmission lines at different service lifespans, the least squares method can be used to fit correction parameters for the transmission line failure probabilities. , is represented as: In the formula, The service life of the pole / tower; This marks the inflection point between the period of stable operation and the period of wear and tear. and The least squares fitting coefficients are used during the loss period. .

[0050] Therefore, by multiplying the transmission line failure probability correction parameter with each overall failure probability model, the service life of the model can be corrected, and the line failure probability model of the power system can be obtained. The corrected model can then be used to quantify the time-varying failure probability of each transmission line.

[0051] This invention considers the alteration of meteorological elements by geographical factors to construct a meteorological impact model, thereby improving the accuracy of meteorological impact assessment. The time-series corrected wind speed for each transmission line not only considers the static influence of geographical factors on wind speed but also reflects the changing characteristics of wind speed over time, enhancing the ability to reflect time-series characteristics. The more comprehensive failure probability model, which comprehensively considers the influence of meteorological factors and the topology of the power system on line failure, can more accurately assess the failure probability of the line. Adjusting the model for service life makes the line failure probability model more consistent with reality.

[0052] S2. Based on the source load data, the source load output sample set under extreme weather conditions is expanded by a time series relative matching generative adversarial network model with gradient penalty, and the expansion result is combined with the time-varying fault probability of the line to generate typical scenarios through Monte Carlo simulation. In one embodiment, the step of expanding the source load output sample set under extreme weather conditions based on the source load data using a time-series relative matching generative adversarial network model with gradient penalty includes: Features are extracted from the source load data and the meteorological data respectively to obtain a feature set, which is then used to reduce the dimensionality of the feature set through an embedding function to generate low-dimensional features. The low-dimensional features are reconstructed using a recovery function to obtain high-dimensional features, and the reconstruction loss between the low-dimensional features and the high-dimensional features is calculated. A feature vector set is constructed by random sampling, and the feature vector set is input into a generator for processing to obtain sample features. The supervision loss between the sample features and the low-dimensional features is calculated. The sample features and the low-dimensional features are input together into the discriminator for processing to obtain the sample probability and the true probability, and the gradient-penalized relative matching unsupervised loss between the sample probability and the true probability is calculated. Based on the reconstruction loss, the supervision loss, and the relative matching unsupervised loss, a total loss function is constructed. With the goal of minimizing the total loss function, the generator and the discriminator are alternately optimized through the backpropagation algorithm until a preset number of training iterations are reached. The optimized generator is then used to expand the source load output sample set under extreme weather conditions, resulting in an expanded extreme weather source load output sample set as the expansion result.

[0053] Specifically, besides impacting power lines exposed to the external environment, the accompanying high wind speeds, precipitation, and temperature drops of typhoons have a complex effect on the output of renewable energy sources such as wind and solar power. This can also lead to a surge in electricity load, severely impacting the power balance of the integrated energy system. In addition to wind power output, solar and load output also exhibit new data characteristics under extreme conditions. However, there is a lack of clear physical correlation models between these components and meteorological factors during typhoon events. Because typhoons and other extreme weather events are low-probability events with limited sample sizes, it is impossible to extract solar-load output characteristics under extreme weather conditions from batch statistical data, making it difficult to directly construct power prediction models using neural networks. Therefore, considering the temporal correlation between power and meteorological data during typhoon weather, this invention proposes a Time-series Relativistic Pairing Generative Adversarial Network with Gradient Penalty (Time-RpGAN-GP) source-load power scenario generation model to describe the uncertainty of renewable energy and load power under extreme scenarios.

[0054] Traditional GAN ​​game theory models only capture the overall probability distribution of the time series, failing to utilize the autoregressive prior information between different time points, thus the accuracy of the generated results needs improvement. Furthermore, the training paradigm of the generator-discriminator adversarial game in GAN models suffers from convergence difficulties and pattern collapse. Convergence difficulties indicate that the model does not fit the training data well, making generated samples easily distinguishable from original samples. Pattern collapse indicates that the generative model can only capture a small amount of data features from the training samples, generating samples that reflect only a small portion of the original data features, resulting in low recall.

[0055] To address the above challenges, this invention proposes the Time-RpGAN-GP model, an improvement upon the traditional GAN ​​model. This model improves upon the traditional GAN ​​model in two aspects. First, to address the difficulty in capturing temporal features, it incorporates raw data for supervised training, referencing autoregressive models. Through adversarial joint training using supervised and unsupervised losses, the model learns progressively time-series features. Second, to address the problem of model convergence difficulties and proneness to collapse, this invention modifies the traditional GAN ​​loss function into a relative matching loss function with gradient penalty. The relative matching loss not only requires the discriminator to distinguish between real and fake samples but also aims to maximize the difference between them. The gradient penalty, after regularization, is applied to the loss function, making the training process more stable.

[0056] The Time-RpGAN-GP model proposed in this invention comprises four parts: an embedding function (mapping static and dynamic features to a low-dimensional latent coding space, reducing the dimensionality of the original space), a recovery function (reconstructing the static and dynamic features of the low-dimensional latent coding space into high-dimensional features), a generator, and a discriminator. Its model structure is as follows: Figure 2 As shown, firstly, static features are extracted from the source-load data, such as wind turbine rated power, photovoltaic installed capacity, and load baseline value. Secondly, dynamic features are extracted from meteorological data, such as hourly wind speed and solar radiation intensity (time-aligned with the source-load data). The two are combined into a feature set and input into the embedding function (Encoder) of the Time-RpGAN-GP model for dimensionality reduction. The embedding function uses a fully connected network to reduce the dimensionality of the static features to obtain low-dimensional static features. Then, an LSTM network is used to reduce the dimensionality of the dynamic features to obtain low-dimensional dynamic features.

[0057] The dimensionality-reduced features are input into the recovery function (Decoder) for reconstruction. The recovery function uses a multi-layer fully connected network to upscale the low-dimensional static features, obtaining high-dimensional static features, and then uses a bidirectional LSTM network to upscale the low-dimensional dynamic features, obtaining high-dimensional dynamic features to verify the effectiveness of dimensionality reduction (ensuring no information loss). Next, the reconstruction loss between the reconstructed high-dimensional features and the low-dimensional features is calculated, which measures the difference between the reconstructed data and the original data, and can be expressed as: In the formula, The reconstruction loss of the Time-RpGAN-GP model; Represents the original static features Reconstructing static features Euclidean distance; Representing the features of the original time series With reconstructed time series features Euclidean distance.

[0058] A generator is used to generate "fake" low-dimensional static and dynamic features from random noise. This involves randomly sampling static and dynamic noise vectors from a simple random distribution (such as a normal distribution) to construct a feature vector set, which is then input into the generator for processing. The low-dimensional features generated by the generator are mapped and converted into source load output samples, i.e., static and dynamic sample features. This process can be represented as: In the formula, , These are the static and dynamic sample features generated by the generator, respectively. , These are static and dynamic generative networks, respectively. , These are randomly selected static and dynamic noise vectors, respectively.

[0059] Calculate the supervised loss of the Time-RpGAN-GP model based on the sample features generated by the generator. Its ability to evaluate the generator's ability to capture temporal features can be expressed as: In the formula, The distance between low-dimensional dynamic features and dynamic sample features is the Euclidean distance.

[0060] The discriminator's role is to distinguish between generated samples and real samples. The real features (i.e., the low-dimensional features corresponding to the source payload data) and the generated sample features are concatenated into an input vector and then fed into the discriminator. The hidden states are calculated using a bidirectional LSTM network, and the probabilities corresponding to these two types of features are obtained through a Sigmoid function. This process can be represented as: In the formula, , These represent the probabilities that static and dynamic samples are true samples, respectively, with a range of [value missing]. ; , These are the discrimination networks for static and dynamic samples, respectively. , These are the forward and backward hidden state sequences, respectively; , It is a forward and backward loop function.

[0061] Then, the unsupervised loss of the Time-RpGAN-GP model is calculated based on the discriminator's output. It is used to measure the distance between the samples generated by the generator and the real samples, providing feedback to the generator, and can be expressed as: In the formula, , The results of the discrimination of static and dynamic features of the original data; , To generate discrimination results for static and dynamic features of the data.

[0062] However, traditional unsupervised loss functions only require the discriminator to distinguish between real and fake samples. The discriminator only needs to perform a binary classification between real and generated samples, and all data points are divided into real or fake classes by a single discrimination boundary, without requiring the distance between real and fake samples to be as large as possible. Under this setting, the generator only needs to make the generated data slightly exceed the discrimination boundary to achieve convergence. However, this simplified loss objective function has multiple degenerate local minima in the solution space, which can easily lead to mode collapse. The relative matching loss function proposed in this invention provides an effective solution to this problem. By coupling real data and generated data, it evaluates the realism of generated samples relative to real samples, thereby effectively maintaining the stability of the decision boundary in the neighborhood of each real sample, and thus avoiding mode collapse. It can be represented as: Furthermore, to improve the training stability of GANs with small training sets and avoid the mode collapse problem caused by zero training gradient when the sample distributions do not overlap, this invention introduces gradient penalty to replace the weight pruning method in traditional GANs. The 1-Lipschitz constraint is applied to the loss function as a gradient penalty, and after regularization, the gradient penalty term is obtained. for: Static and dynamic features are constructed using interpolation distributions, satisfying the following: In the formula, These are random numbers between 0 and 1. Compared to global parameters, gradient penalty only applies constraints in the critical region between real and generated samples, avoiding over-regularization. Furthermore, its continuity extends to the entire space, ensuring comprehensive coverage of the sample space and computational feasibility. Combining the gradient penalty term yields the relative matching unsupervised loss function with gradient penalty. It can be represented as: In the formula, The weight coefficients for gradient penalty.

[0063] The reconstruction loss, supervised loss, and unsupervised relative matching loss with gradient penalty are weighted and summed to obtain the total loss function. The optimal weights for the three are 0.4, 0.2, and 0.4, respectively. Finally, with the goal of minimizing the total loss function, the generator and discriminator are alternately optimized through backpropagation until a preset number of training iterations or loss fluctuations of no more than 5% are achieved. The optimized generator is then used to expand the source-load output sample set under extreme weather conditions, resulting in the expanded extreme weather source-load output sample set (i.e., wind power, photovoltaic, and load time-series data under extreme typhoon weather, with a time step of 1 hour and a sample size of ≥5000 groups).

[0064] The Time-RpGAN-GP model constructed in this invention effectively extracts and reduces dimensionality through embedding functions, reducing computational load. It then uses a recovery function for feature reconstruction to verify the effectiveness of dimensionality reduction (ensuring no information loss). A generator extracts vectors from a random distribution to generate source-load samples consistent with the real data distribution. A discriminator distinguishes between real and generated samples, outputting the probability of sample authenticity. Furthermore, a three-loss collaborative optimization model is employed. The reconstruction loss measures the difference between the original and reconstructed features, ensuring dimensionality reduction / reconstruction accuracy. The supervised loss constrains the temporal consistency between generated dynamic features and real dynamic features. The gradient-penalized relative matching unsupervised loss addresses the convergence difficulties and mode collapse issues of traditional GANs, effectively improving the overall model performance. The expanded sample set increases data diversity, more comprehensively reflecting various source-load output conditions under extreme weather, providing richer and more accurate data support for subsequent source-load prediction, power system planning and operation, and improving the scientific rigor and reliability of related decisions.

[0065] In one embodiment, combining the augmentation results with the time-varying fault probability of the line to generate typical scenarios through Monte Carlo simulation includes: Based on the time-varying fault probability of the line and the extended sample set of extreme weather source load output, several fault scenarios are obtained by constructing multi-time-period scenarios through sequential Monte Carlo. For each of the aforementioned fault scenarios, the power loss of the scenario is calculated as a clustering feature, and based on the clustering feature, the K-means clustering algorithm is used to cluster the fault scenarios to obtain typical scenarios.

[0066] Specifically, this invention divides the typhoon impact period into T time periods (e.g., 24 hours, with one time period per hour), and performs time-by-time sampling based on the calculated time-varying line fault probability and the extended sample set of extreme weather source load output. For each time period t, Bernoulli sampling is used to determine whether the line is faulty (1 = fault, 0 = normal) based on the line fault probability. Then, a set of source load data matching time period t is randomly selected from the extended sample set. Subsequently, scenario simulation is constructed through sequential Monte Carlo simulation, so that each scenario is represented by the line fault state + source load output, thus obtaining several fault scenarios.

[0067] For each fault scenario, the total load loss during the typhoon impact period is calculated (i.e., the difference between the time-series load data of the transmission line under extreme weather conditions and the power supply of the line). Using the calculated total load loss as a clustering feature, multiple fault scenarios are clustered into several classes through K-means clustering. The central scenario of each class is selected as a typical scenario, which includes information such as the line fault status and source load output timing, to ensure a balance between scenario representativeness and computational efficiency. In one embodiment, the Monte Carlo simulation process for generating typical scenarios is as follows: Figure 3 As shown.

[0068] This invention constructs multi-time-period scenarios based on sequential Monte Carlo simulation, which can comprehensively reflect the dynamic characteristics of the system, increase the diversity of scenarios, and thus improve the accuracy of simulation. For each fault scenario, the power loss of the scenario is calculated as a clustering feature and then reduced, which can simplify the scenario set, improve the computational efficiency, focus on key scenarios, enhance the pertinence of decision-making, and thus improve the reliability and disaster resistance of the system under extreme weather conditions.

[0069] S3. Using the typical scenario as input, simulate the cascading failure evolution process under line faults and source load fluctuations through the AC cascading failure model, and output the cascading failure evolution path and the system operating status at the time of termination. In one embodiment, step S3 includes: Based on the typical scenario and the topology data, a node power balance equation is constructed to perform optimal AC power flow calculation and obtain node voltage and line power flow data. The state of the power system is checked based on the node voltage and the line power flow data. When the check result exceeds the limit, a protection mechanism is triggered to update the typical scenario. Based on the updated typical scenario, iteratively execute the optimal AC power flow calculation steps and system state verification steps until the preset iteration termination condition is met, and output the cascading failure evolution path and the system running state at the time of termination.

[0070] Specifically, extreme weather directly impacts the power system, causing issues such as load fluctuations and line faults. However, these changes in supply-demand balance and grid structure often lead to cascading failures. Cascading failures are related to both inherent system characteristics and initial faults. To address the impact of extreme weather, based on typical scenarios and topology data, a time-series initial fault set for extreme weather, including branches, nodes, and generators, is established. As shown below: In the formula, The initial set of disconnected branches at time t; The load change at the corresponding load node at time t; The corresponding change in power output of the new energy node at time t; The total simulation period together constitutes the initial fault set for extreme weather time series.

[0071] After an initial fault occurs, the operating state of the power system changes. When the severity of the power system reaches a certain threshold, the system protection mechanism will be triggered, leading to new operating conditions. This invention uses AC-CFM (Alternating Current Chain Fault Model) to simulate the evolution process of a chain fault triggered by an initial fault (line fault + source-load fluctuation), thereby solving the problem of dynamically depicting the propagation of faults from a single component to the system level.

[0072] First, AC power flow calculation is performed: based on the set of disconnected branches in the initial fault set, load changes, and changes in renewable energy output, the nodal power balance equation is constructed, which is expressed by the following formula: In the formula, P i Q i Let V be the active power and reactive power of node i, respectively; N be the total number of nodes in the system; V i V j Y represents the voltage magnitudes at nodes i and j, respectively; ij δ is the admittance between nodes i and j; i δ j θ represents the voltage phase angles at nodes i and j, respectively; ij To guide Y ij The angle of the argument.

[0073] The power flow equations are then solved using the Newton-Raphson method. If convergence is achieved (residual ≤ 1e-6), the node voltage and line power flow current data are output. If the AC power flow does not converge, the line capacity constraint (relaxed line capacity is 1.1 times the original maximum) and node voltage constraint (relaxed voltage range is between 0.8pu and 1.2pu) are relaxed. The optimal power flow model is constructed with the minimum network loss as the objective function. The model is solved using the interior point method to obtain the relaxed node voltage and line power flow current data.

[0074] The node voltage obtained through power flow calculation is compared with the preset rated voltage. If the node voltage is less than the preset rated voltage, it indicates that the node is undervoltage and the power flow calculation result exceeds the limit. Alternatively, the line power flow current data obtained through power flow calculation is compared with the preset rated current. If the line power flow current data is greater than the preset rated current, it indicates that the line is overloaded and the power flow calculation result exceeds the limit. In this case, the protection mechanism is triggered: for overloaded lines, the line is disconnected and added as a fault line to the typical scenario to update the typical scenario; for low-voltage nodes, 30% of the load of the node is cut off, and the load change and renewable energy output of the node are updated. The updated data is then added to the typical scenario to update the typical scenario.

[0075] The optimal AC power flow calculation steps and system state verification steps are iteratively executed according to the updated typical scenario until any of the following conditions are met, at which point the cascading fault terminates: the optimal power flow does not converge (the system is on the verge of collapse); the system state does not exceed the limit (no new faults are triggered); the number of iterations is ≥10 (to avoid infinite loops); the cascading fault evolution path (the set of disconnected branches of the transmission line at each moment) and the system operating state at the time of termination (node ​​voltage distribution, line power flow current, line power flow current power, total load loss, etc.) are output.

[0076] This invention, through simulation and analysis of the cascading failure evolution process, provides a deeper understanding of the occurrence, development, and propagation mechanisms of cascading failures in power systems under complex operating conditions. It can assess the system's resilience and recovery capabilities under different failure scenarios and identify system weaknesses. The output cascading failure evolution path and system operating status at termination provide power system operators and managers with intuitive and detailed information, enabling them to formulate targeted emergency plans and fault handling strategies. When cascading failures occur, they can take rapid and accurate measures to reduce the impact of the failures on the system and ensure the safe and stable operation of the power system.

[0077] S4. Based on the cascading failure evolution path and the system operating state at the time of termination, a first-level surface vulnerability analysis is performed by constructing a severity model, a second-level internal structural vulnerability analysis is performed by using spectral theory, and a third-level cascading failure causal probability analysis is performed based on Bayesian networks to obtain the multi-level vulnerability assessment results of the power system. In one embodiment, the step of performing a first-level surface vulnerability analysis by constructing a severity model based on the cascading failure evolution path and the system operating state at the time of termination, and performing a second-level internal structural vulnerability analysis using spectral theory, includes: The line's rated capacity percentage is determined based on the system's operating state at the time of termination. A line overload severity model is constructed based on the line's rated capacity percentage, and a node low-voltage severity model is constructed based on the system's operating state at the time of termination. Based on the line overload severity model and the node low voltage severity model, the expected risk corresponding to the surface weak link of the power system is quantified to obtain the first-level analysis results. The power flow weights of each transmission line are determined by the system operating state at the time of termination, in order to construct a weight matrix, and a Laplace matrix is ​​constructed based on the weight matrix. Based on the topology data and the cascading failure evolution path, a network adjacency matrix is ​​constructed and combined with the Laplace matrix to quantify the network structural vulnerability index corresponding to the inherent structural weaknesses of the power system, thereby obtaining secondary analysis results.

[0078] Specifically, the source-load output and line fault probability model under extreme weather conditions provides a probabilistic operating boundary for the power system. Based on this probabilistic operating boundary, system operating risks can be further analyzed, and the initial impact of extreme weather on the power system can be quantified. The operating status of lines and nodes is used to observe the system's response. For lines, source-load imbalance and line damage will exacerbate power flow concentration and line overload. This invention introduces the percentage of rated capacity (PR) to characterize the operating status of a line. It is determined based on the system operating status at termination and is the ratio of the actual power flow current to the rated power flow current of the line. Then, a line overload severity model is constructed based on the percentage of rated capacity. As shown below: In the formula, d and c are parameters of the line overload severity function; typically, d is taken as 10 and c as -9; PR k PR is the percentage of the rated capacity of the k-th line; k, min This represents the maximum safe capacity percentage for the k-th line, preferably 20%.

[0079] For nodes, the voltage amplitude in the system's operating state at the time of termination can be directly used as the node's operating state. In extreme weather conditions, insufficient system power supply can lead to voltage drops at some nodes. Therefore, the node low-voltage severity model... As shown below: In the formula, The parameter V is the node low-voltage severity function. ref and V lim These are the reference voltage and the extreme low voltage, respectively, with values ​​of 1.0 pu and 0.9 pu.

[0080] Based on the calculation and ranking of line overload severity models and node low-voltage severity models, the weak links in the power system under the influence of extreme weather can be identified. These weak links are determined by both their severity and probability of occurrence, reflecting the expected risk of each link in the power system under uncertain events. The power system experiences a fault C under extreme weather Ψ. i Risk Expectation As shown below: The expected risk of the power system under extreme weather Ψ, which corresponds to the primary analysis result of the expected risk of the weakest link in the power system, is as follows: In the formula, This is the result of the first-level analysis.

[0081] This invention analyzes the severity of power system operation to identify probabilistic weak links in the power system. The weak links identified based on scenarios reflect the initial impact of extreme weather on the power system.

[0082] The power flow weights of transmission lines, reflecting differences in system function, are determined based on line power flow current data and are calculated using the following formula: In the formula, w ij P represents the power flow weight of transmission line ij; ij R represents the active power from node i to node j; ij X ij These are the resistance and reactance of transmission line ij, respectively.

[0083] A weight matrix is ​​constructed based on the calculated power flow weights. The Laplace matrix is ​​defined in spectral theory as... Where D and W are the node degree diagonal matrix and the network weight matrix, respectively, then the Laplacian matrix is... As shown below: In the formula, and These are the set of nodes and the set of branches in the network, respectively.

[0084] The Laplace matrix reflects the cumulative benefit of each link in a network to other links when subjected to a disturbance. Its mathematical properties can characterize the importance of each link in the system's function, thus identifying the system's weak points. Subsequently, a network adjacency matrix is ​​constructed based on topological data and the presence of connections between nodes in the cascading failure evolution path. Finally, a vulnerability index, spectral radius, is constructed based on the Laplace matrix and the network adjacency matrix. It is the largest eigenvalue of the adjacency matrix, and its spectral radius is related to the weights of the closest connections in the network, affecting the system's maximum power transfer capability; algebraic connectivity. It is the second smallest eigenvalue of the Laplace matrix. Algebraic connectivity measures the structural compactness of the power grid and the system's immunity to disturbances; natural connectivity... It is calculated by the exponent of the eigenvalues ​​of the adjacency matrix, and characterizes the number of redundant paths and fault tolerance of the power network. It is expressed by the following formula: In the formula, v i represents the eigenvalues ​​of the elements in the network adjacency matrix; N represents the total number of eigenvalues ​​in the network adjacency matrix.

[0085] Effective impedance λ R It is the sum of the equivalent resistances between all pairs of nodes in the network, and can also be represented by the eigenvalues ​​of the Laplace matrix. That is, the overall impedance of the system is proportional to the sum of the reciprocals of the non-zero eigenvalues, and is used to reflect the overall transmission efficiency and energy loss of the power network. It is expressed by the following formula: In the formula, represents the eigenvalues ​​of the elements in the Laplace matrix; N is the total number of eigenvalues ​​in the Laplace matrix.

[0086] The above structural vulnerability indicators can quantify the network performance of a system from different dimensions. To achieve structural vulnerability assessment and identification of weak points in different aspects, an N-1 fault scan method for lines and nodes is adopted to analyze network performance under different component outage or failure conditions, thereby identifying their role in system function. This invention first calculates the structural vulnerability indicators under fault-free conditions, and then recalculates the above indicators for each component (line / node) after simulating its failure, thus obtaining the indicator change value. The larger the change value, the more important the corresponding component. To synthesize the evaluation conclusions of various indicators, the network structural vulnerability indicators corresponding to the inherent structural weaknesses of the power system are obtained by summing the indicators. The assessment results based on the structural vulnerability of the power system are linearly superimposed on the scenario vulnerability assessment results to form the identification results of the weak points of the power system.

[0087] This invention uses the percentage of the line's rated capacity to intuitively reflect the load status of the line when a fault terminates, thereby constructing a severity model that can accurately locate surface-level risk points, prevent surface-level faults in advance, and improve system operational safety. By constructing a key matrix to quantify network structure vulnerability indicators, it can further explore the inherent vulnerabilities of the system and improve the system's ability to cope with cascading failures. Through a comprehensive assessment of the system's weak links, it provides a scientific and quantitative basis for decisions on power system planning, design, operation, and maintenance.

[0088] In one embodiment, the three-level cascading failure causal probability analysis based on Bayesian networks includes: Based on the time-series corrected wind speed, the power flow data of the line, and the cascading fault evolution path, wind speed nodes, power flow mutation nodes, and fault nodes are determined respectively. Edges are set based on the dependency relationship of the corresponding fault mechanism between the wind speed nodes, the power flow mutation nodes, and the fault nodes to construct a Bayesian network. A training dataset is constructed based on the wind speed node, the tidal current change node, and the fault node. The conditional probability of each node in the Bayesian network is trained using the training dataset to obtain the trained Bayesian network. The real-time power flow data and real-time meteorological data of the power system are input into the trained Bayesian network for processing, so as to realize the causal probability prediction of the cascading failures of the power system and obtain the three-level analysis results.

[0089] Specifically, this invention also introduces a causal-probabilistic dual-drive prediction method that combines Bayesian networks and Monte Carlo simulation to deeply analyze the causal relationship of fault propagation, thereby providing more scientific theoretical guidance for effectively curbing the occurrence of cascading faults and improving the resilience of power systems.

[0090] This invention sets fault nodes based on the operating status of each line at time t in the cascading fault evolution path (discrete variable: 0 for fault outage, 1 for normal operation), wind speed nodes based on the wind speed of each line at time t in the time-series corrected wind speed data (continuous variable, discretized into 3 levels: low wind <8m / s, medium wind 8~12m / s, high wind >12m / s), and power flow mutation nodes based on the power flow mutation amount of each line at time t in the power flow data (continuous variable, discretized into 3 levels: small mutation <10%, medium mutation 10%~20%, large mutation >20%). Based on the dependencies of the corresponding fault mechanisms between these nodes, such as high wind speed corresponding to a wind speed node causing the line corresponding to the fault node to shut down, power flow mutation corresponding to a power flow mutation node causing the line corresponding to the fault node to trip, outage of the upstream line corresponding to a fault node causing power flow transfer of the downstream line corresponding to the power flow mutation node, and another fault node triggering an outage, edges are set, and a Bayesian network is constructed based on the nodes and edges.

[0091] A Bayesian network can generally be modeled as a system S containing multiple random events, where any random event in the system can be uniquely mapped to a continuous or discrete random variable X. Assume that system S contains N random variables. Then a sheet containing a set of nodes can be used. Sum of edges The picture To characterize system S. The set of nodes... ={ , ,…, } represents N random variables, and the set of edges. ={ , ,…, } represents the dependency relationship between random variables. Let's consider the graph... Middle Exists, and the dependency is determined by the node. Pointing to node Then it is called a node. As the parent node, the node For child nodes. Record the node. The set of parent nodes is Then the node Conditional probability function of occurrence It can be represented as: Bayesian network parameters That is, the diagram The set of conditional probabilities of all nodes in the set is represented as: This invention aligns wind speed nodes, tidal current change nodes, and fault nodes according to time step t to form a training dataset. This includes the typhoon wind speed at the location of the transmission line corresponding to the wind speed node. The power flow change caused by net load fluctuations on the transmission lines corresponding to the power flow change node. and the operating status of the transmission lines corresponding to the fault nodes. , can be represented as: In the formula, The typhoon wind speed at the location of the power transmission line; This refers to the power flow sudden change caused by net load fluctuations in transmission lines. This refers to the operating status of the power transmission line.

[0092] Based on the training data, the following Bayesian network optimization model can be obtained: In the formula, For Bayesian networks about datasets The likelihood function; The Bayesian network parameter values ​​that maximize the likelihood function.

[0093] The conditional probabilities of each node in the Bayesian network are trained using a training dataset; that is, the conditional probabilities of each node in the Bayesian network are learned using the training dataset, thereby determining the parameters of the Bayesian network. The trained Bayesian network can then be obtained. The training process for the Bayesian network can be referenced from existing technologies and will not be elaborated upon here. Real-time power flow data and real-time meteorological data from the power system are input into the trained Bayesian network. By processing the data, we can obtain the risk outage index, which reflects the real-time outage risk of power lines. This index is used to characterize the causal probability prediction of cascading failures in the power system, i.e., the result of the third-level analysis. It is expressed by the following formula: In the formula, For the line The outage risk indicator at time t also serves as a basis for observing the line's wind speed and power flow changes at time t. The conditional probability of a broken wire.

[0094] Finally, the results of the first, second, and third level analyses are combined to obtain the multi-level vulnerability assessment results of the power system.

[0095] This invention incorporates meteorological factors (wind speed), system operating status (power flow), and fault development processes into a single Bayesian network for analysis. This overcomes the potential limitations of traditional analysis methods, enabling a more comprehensive and systematic understanding of the mechanisms of cascading faults and providing a powerful tool for in-depth analysis of cascading faults. Furthermore, it integrates the impacts of extreme weather, power system structure, and source-load output on power system vulnerability, establishing a multi-level assessment method of "surface phenomenon identification - internal structure analysis - deep causal revelation." Compared to traditional single-level assessments, this method reflects the correlation and causality between different levels, and the assessment results are closer to reality. It can provide suggestions and references for power system planning to reduce the risk of extreme events.

[0096] In one embodiment, the present invention uses an improved IEEE 14-bus system as a power system, the structure of which is as follows: Figure 4As shown, to analyze the impact of extreme weather on renewable energy sources such as wind power and photovoltaics, 20% of the installed power capacity was replaced with wind power and 30% with photovoltaics in the verification example. Historical meteorological data of Typhoon Lan (No. 7) was used, and a coordinate system was constructed with the typhoon's landfall point as the origin. It was assumed that the typhoon's path was linear, its center speed was 14.9 km / h, and its initial center pressure was 950 hPa. Dynamic time-series simulations were conducted using the Batts typhoon model to obtain the time-varying fault probabilities of some transmission lines, as shown below. Figure 5 As shown, in the early stages of typhoon landfall, line 5, affected by the typhoon's path, has a high probability of failure, with a maximum probability of 0.713. As the typhoon moves and weakens, the probability of line failure gradually decreases, and transmission lines far from the typhoon's path are basically unaffected.

[0097] Using photovoltaic power output, load power, and meteorological data from selected regions, time-series data on renewable energy power output and load power under typhoon conditions were selected. Typhoon-labeled data was then input into the Time-RpGAN-GP model for training, thereby generating source-load power output scenarios under extreme typhoon conditions. The expanded extreme scenario source-load power output sample results of the Time-RpGAN-GP model are shown below. Figure 6 As shown, solar radiation is blocked by clouds when the typhoon passes through, and photovoltaic power output exhibits obvious low power statistics. The thermal structure of the typhoon causes changes in the temperature field, resulting in differential abrupt changes in load power in the wind and rain zone and the outer area, respectively, with a gradual decrease and a sharp increase. The source-load power time series samples generated by Time-RpGAN-GP can track the model and trend of the actual samples. Compared with CGAN, the Time-RpGAN-GP model has a more significant advantage in capturing the changing trends and fluctuation characteristics of new energy sources.

[0098] Based on wind speed-related time-varying fault probabilities and source-load power time-series scenarios, a sequential Monte Carlo simulation method can be used to generate power system disaster operation scenarios under typhoon weather in batches. This method effectively couples physical-driven models and data-driven methods, significantly improving the completeness of extreme event simulations while ensuring computational accuracy. Theoretically, the more simulations performed, the closer the generated scenarios will be to the true probability distribution of system damage. However, in engineering applications, only the most likely scenarios need to be studied. Furthermore, too many scenarios will increase the solution complexity and slow down the computation speed. This invention uses the load loss power index of each scenario to reduce the number of scenarios, and obtains converged typical disaster-stricken typhoon scenarios by iteratively updating the cluster centers.

[0099] This paper analyzes the evolution path of cascading failures by incorporating network topology changes and source-load output scenarios under the influence of extreme weather into an AC-CFM model. The initial fault set and the fault set after the termination of cascading failures are used as the system operating boundary to analyze the severity of line overload and node low voltage. The distribution of power system severity at the initial moment of typhoon impact is shown in the figure. Figure 7 As shown; among them, line 12 (12-13) has been disconnected in a cascading failure, and although the severity is high, the probability is 0. Lines 2 (2-3) and 4 (4-5) have a probability of line overload, but the severity is 0. The node low-voltage severity and probability of node 1 are the most significant, verifying the operating status of the isolated node. However, considering the actual boundary conditions of the severity assessment, the node low-voltage severity risk of the isolated node does not participate in the ranking of weak nodes. In addition, due to changes in system source load output and network structure, other nodes also showed certain low-voltage risks. The system severity risk under this extreme weather is 21.432.

[0100] The identification results of structurally weak branches and nodes based on spectral analysis are as follows: Figure 8 , Figure 9 As shown, the Bayesian network structure of the IEEE 14-node system is as follows: Figure 10 As shown in the table below, the larger the relative change value, the higher the structural importance of the corresponding branch and node in the system. For example, even if node 1 becomes an isolated node in an AC cascading failure, it does not cause further cascading failures because of its low importance in the system. Combining system scenario vulnerability and structural vulnerability, the two types of indicators are summed and sorted to obtain the identification results of the system's weak links, as shown in the table below: Table 1. Results of IEE 14 System Weakness Identification From Table 1 and the accompanying drawings in the instruction manual Figure 8 , Figure 9 It can be seen that weak branches and weak nodes are topologically related, reflecting the impact of network structure on system function.

[0101] Based on a causal-probabilistic dual-driven prediction method combining Bayesian networks and Monte Carlo simulation, the causal relationship of fault propagation is analyzed, and the weak links in the power system are identified as shown in the table below: Table 2. Bayesian Outage Risk Identification Results for the IEE 14-Node System Table 2 shows that lines with a higher probability of line faults are lines 10, 11, 7, 6, and 13, which also have higher power outage risk indicators; combined with Figure 10The Bayesian network structure diagram analysis shows that the probability of failure for this type of line is low. However, since a break in this line can easily trigger multiple line overloads and breakages, the cascading risk is high. Therefore, the power outage risk index for this type of line is much higher than the probability of line failure, which fully demonstrates the effectiveness of the "causality-probability" dual-driven prediction and identification.

[0102] This invention addresses the issue of improving the accuracy of power system vulnerability assessment under extreme weather conditions by designing a power system vulnerability assessment method that considers the impact of extreme weather. By constructing a line fault probability model, it can more accurately consider the fault probability of different lines under meteorological conditions, significantly improving the accuracy of line fault probability prediction. Through a gradient-penalized time-series relative matching generative adversarial network model and Monte Carlo simulation, it generates typical scenarios covering various operating states of the power system under extreme weather conditions, facilitating a more comprehensive assessment of the power system's vulnerability under different circumstances. By simulating the cascading failure evolution process using an AC cascading failure model, it clarifies the scope of fault impact and the final operating state of the system, providing important reference for developing targeted preventative measures. Based on the fault evolution results, through spectral theory and Bayesian networks, combined with severity analysis, structural vulnerability assessment, and causal-probability prediction, it achieves multi-level vulnerability assessment and quantification of the power system, comprehensively identifying surface phenomena, analyzing internal structures, and revealing deep causal relationships. This solves the problems of insufficient single-level assessment and neglect of causal relationships in existing technologies, achieving a comprehensive and accurate assessment of the power system's vulnerability under extreme weather conditions.

[0103] It should be noted that although the steps in the flowchart above are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order requirement for the execution of these steps, and they can be executed in other orders.

[0104] In another embodiment, such as Figure 11 As shown, a second aspect of the present invention provides a power system vulnerability assessment system that takes into account the impact of extreme weather, comprising: The probability calculation module 10 is used to acquire the topology data of the power system and its meteorological data and source-load data under extreme weather conditions, and to construct the line fault probability model of the power system based on the topology data and the meteorological data in order to calculate the time-varying fault probability of the line. The scenario generation module 20 is used to expand the source load output sample set under extreme weather conditions based on the source load data by using a time series relative matching generative adversarial network model with gradient penalty, and combine the expansion result with the time-varying fault probability of the line to generate typical scenarios through Monte Carlo simulation. The fault evolution module 30 is used to take the typical scenario as input, simulate the cascading fault evolution process under line fault and source load fluctuation through the AC cascading fault model, and output the cascading fault evolution path and the system operating status at the time of termination. The multi-level assessment module 40 is used to perform first-level surface vulnerability analysis by constructing a severity model based on the cascading failure evolution path and the system operating state at the time of termination, second-level internal structural vulnerability analysis by using spectral theory, and third-level cascading failure causal probability analysis based on Bayesian networks, so as to obtain the multi-level vulnerability assessment results of the power system.

[0105] In one embodiment, the step of constructing a line fault probability model of the power system based on the topology data and the meteorological data to calculate the time-varying line fault probability includes: A meteorological impact model is constructed based on the meteorological data, and geographical data of each transmission line in the power system is obtained. The meteorological impact model is then corrected using the geographical data to obtain the time-series corrected wind speed for each transmission line. Based on the time-series corrected wind speed and the topology data, an overall failure probability model for each of the transmission lines is constructed, and the service life of each of the overall failure probability models is corrected to obtain the line fault probability model of the power system in order to quantify the time-varying fault probability of the lines.

[0106] In one embodiment, the step of expanding the source load output sample set under extreme weather conditions based on the source load data using a time-series relative matching generative adversarial network model with gradient penalty includes: Features are extracted from the source load data and the meteorological data respectively to obtain a feature set, which is then used to reduce the dimensionality of the feature set through an embedding function to generate low-dimensional features. The low-dimensional features are reconstructed using a recovery function to obtain high-dimensional features, and the reconstruction loss between the low-dimensional features and the high-dimensional features is calculated. A feature vector set is constructed by random sampling, and the feature vector set is input into a generator for processing to obtain sample features. The supervision loss between the sample features and the low-dimensional features is calculated. The sample features and the low-dimensional features are input together into the discriminator for processing to obtain the sample probability and the true probability, and the gradient-penalized relative matching unsupervised loss between the sample probability and the true probability is calculated. Based on the reconstruction loss, the supervision loss, and the relative matching unsupervised loss, a total loss function is constructed. With the goal of minimizing the total loss function, the generator and the discriminator are alternately optimized through the backpropagation algorithm until a preset number of training iterations are reached. The optimized generator is then used to expand the source load output sample set under extreme weather conditions, resulting in an expanded extreme weather source load output sample set as the expansion result.

[0107] In one embodiment, combining the augmentation results with the time-varying fault probability of the line to generate typical scenarios through Monte Carlo simulation includes: Based on the time-varying fault probability of the line and the extended sample set of extreme weather source load output, several fault scenarios are obtained by constructing multi-time-period scenarios through sequential Monte Carlo. For each of the aforementioned fault scenarios, the power loss of the scenario is calculated as a clustering feature, and based on the clustering feature, the K-means clustering algorithm is used to cluster the fault scenarios to obtain typical scenarios.

[0108] In one embodiment, the fault evolution module 30 includes: Based on the typical scenario and the topology data, a node power balance equation is constructed to perform optimal AC power flow calculation and obtain node voltage and line power flow data. The state of the power system is checked based on the node voltage and the line power flow data. When the check result exceeds the limit, a protection mechanism is triggered to update the typical scenario. Based on the updated typical scenario, iteratively execute the optimal AC power flow calculation steps and system state verification steps until the preset iteration termination condition is met, and output the cascading failure evolution path and the system running state at the time of termination.

[0109] In one embodiment, the step of performing a first-level surface vulnerability analysis by constructing a severity model based on the cascading failure evolution path and the system operating state at the time of termination, and performing a second-level internal structural vulnerability analysis using spectral theory, includes: The line's rated capacity percentage is determined based on the system's operating state at the time of termination. A line overload severity model is constructed based on the line's rated capacity percentage, and a node low-voltage severity model is constructed based on the system's operating state at the time of termination. Based on the line overload severity model and the node low voltage severity model, the expected risk corresponding to the surface weak link of the power system is quantified to obtain the first-level analysis results. The power flow weights of each transmission line are determined by the system operating state at the time of termination, in order to construct a weight matrix, and a Laplace matrix is ​​constructed based on the weight matrix. Based on the topology data and the cascading failure evolution path, a network adjacency matrix is ​​constructed and combined with the Laplace matrix to quantify the network structural vulnerability index corresponding to the inherent structural weaknesses of the power system, thereby obtaining secondary analysis results.

[0110] In one embodiment, the three-level cascading failure causal probability analysis based on Bayesian networks includes: Based on the time-series corrected wind speed, the power flow data of the line, and the cascading fault evolution path, wind speed nodes, power flow mutation nodes, and fault nodes are determined respectively. Edges are set based on the dependency relationship of the corresponding fault mechanism between the wind speed nodes, the power flow mutation nodes, and the fault nodes to construct a Bayesian network. A training dataset is constructed based on the wind speed node, the tidal current change node, and the fault node. The conditional probability of each node in the Bayesian network is trained using the training dataset to obtain the trained Bayesian network. The real-time power flow data and real-time meteorological data of the power system are input into the trained Bayesian network for processing, so as to realize the causal probability prediction of the cascading failures of the power system and obtain the three-level analysis results.

[0111] It should be noted that the modules in the aforementioned power system vulnerability assessment system considering the impact of extreme weather can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module. For specific limitations regarding the power system vulnerability assessment system considering the impact of extreme weather, please refer to the limitations of the power system vulnerability assessment method considering the impact of extreme weather mentioned above; both have the same function and role, and will not be repeated here.

[0112] A third aspect of the present invention provides an electronic device comprising: Processor, memory, and bus; The bus is used to connect the processor and the memory; The memory is used to store operation instructions; The processor is configured to execute operations corresponding to a power system vulnerability assessment method considering the impact of extreme weather, as shown in the first aspect of the present invention, by invoking the operation instructions.

[0113] In one alternative embodiment, an electronic device is provided, such as Figure 12 As shown, Figure 12The illustrated electronic device 5000 includes a processor 5001 and a memory 5003. The processor 5001 and the memory 5003 are connected, for example, via a bus 5002. Optionally, the electronic device 5000 may also include a transceiver 5004. It should be noted that in practical applications, the transceiver 5004 is not limited to one type, and the structure of this electronic device 5000 does not constitute a limitation on the embodiments of the present invention.

[0114] Processor 5001 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 5001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0115] Bus 5002 may include a path for transmitting information between the aforementioned components. Bus 5002 may be a PCI bus or an EISA bus, etc. Bus 5002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 12 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0116] The memory 5003 may be a ROM or other type of static storage device capable of storing static information and instructions, RAM or other type of dynamic storage device capable of storing information and instructions, or it may be an EEPROM, CD-ROM or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0117] The memory 5003 is used to store application code that executes the present invention, and its execution is controlled by the processor 5001. The processor 5001 is used to execute the application code stored in the memory 5003 to implement the content shown in any of the foregoing method embodiments.

[0118] Among them, electronic devices include, but are not limited to: mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), and in-vehicle terminals (such as in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers.

[0119] The fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a method for assessing the vulnerability of a power system considering the impact of extreme weather, as described in the first aspect of the present invention.

[0120] Another embodiment of the present invention provides a computer-readable storage medium storing a computer program that, when run on a computer, enables the computer to execute the corresponding content in the foregoing method embodiments.

[0121] Furthermore, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0122] In summary, this invention relates to the field of power system assessment technology, and discloses a method and system for assessing the vulnerability of power systems considering the impact of extreme weather. It constructs a line fault probability model for the power system under extreme weather conditions to determine the time-varying fault probability of lines under dynamic meteorological influences. Typical scenarios are generated using a gradient-penalized time-series relative matching generative adversarial network model and Monte Carlo simulation. Using the generated typical scenarios as input, the evolution of cascading faults is simulated through an AC cascading fault model, outputting the cascading fault evolution path. Furthermore, based on the fault evolution results, multi-level vulnerability quantification of the power system is achieved through spectral graph theory and Bayesian networks, combined with severity analysis, structural vulnerability assessment, and causal-probability prediction. This solves the problems of insufficient single-level assessment and neglect of causal relationships in existing technologies, improving the accuracy of risk assessment and disaster prevention capabilities of power systems under extreme weather conditions.

[0123] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on its differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0124] The above-described embodiments are merely preferred embodiments of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention. Therefore, the scope of protection of this invention should be determined by the scope of the claims.

Claims

1. A method for assessing the vulnerability of power systems considering the impact of extreme weather, characterized in that, include: The topology data of the power system and its meteorological data and source-load data under extreme weather conditions are obtained, and a line fault probability model of the power system is constructed based on the topology data and the meteorological data to calculate the time-varying fault probability of the line. Based on the source load data, the source load output sample set under extreme weather conditions is expanded by a time series relative matching generative adversarial network model with gradient penalty, and the expansion result is combined with the time-varying fault probability of the line to generate typical scenarios through Monte Carlo simulation. Using the aforementioned typical scenario as input, the cascading failure evolution process under line faults and source load fluctuations is simulated through an AC cascading failure model, and the cascading failure evolution path and system operating status at termination are output. Based on the cascading failure evolution path and the system operating state at the time of termination, a first-level surface vulnerability analysis is performed by constructing a severity model, a second-level intrinsic structural vulnerability analysis is performed using spectral theory, and a third-level cascading failure causal probability analysis is performed based on Bayesian networks to obtain the multi-level vulnerability assessment results of the power system.

2. The method for assessing the vulnerability of power systems considering the impact of extreme weather as described in claim 1, characterized in that, The step of constructing a line fault probability model for the power system based on the topology data and the meteorological data to calculate the time-varying fault probability of the lines includes: A meteorological impact model is constructed based on the meteorological data, and geographical data of each transmission line in the power system is obtained. The meteorological impact model is then corrected using the geographical data to obtain the time-series corrected wind speed for each transmission line. Based on the time-series corrected wind speed and the topology data, an overall failure probability model for each of the transmission lines is constructed, and the service life of each of the overall failure probability models is corrected to obtain the line fault probability model of the power system in order to quantify the time-varying fault probability of the lines.

3. The method for assessing the vulnerability of power systems considering the impact of extreme weather as described in claim 1, characterized in that, The expansion of the source load output sample set under extreme weather conditions based on the source load data, using a time-series relative matching generative adversarial network model with gradient penalty, includes: Features are extracted from the source load data and the meteorological data respectively to obtain a feature set, which is then used to reduce the dimensionality of the feature set through an embedding function to generate low-dimensional features. The low-dimensional features are reconstructed using a recovery function to obtain high-dimensional features, and the reconstruction loss between the low-dimensional features and the high-dimensional features is calculated. A feature vector set is constructed by random sampling, and the feature vector set is input into a generator for processing to obtain sample features. The supervision loss between the sample features and the low-dimensional features is calculated. The sample features and the low-dimensional features are input together into the discriminator for processing to obtain the sample probability and the true probability, and the gradient-penalized relative matching unsupervised loss between the sample probability and the true probability is calculated. Based on the reconstruction loss, the supervision loss, and the relative matching unsupervised loss, a total loss function is constructed. With the goal of minimizing the total loss function, the generator and the discriminator are alternately optimized through the backpropagation algorithm until a preset number of training iterations are reached. The optimized generator is then used to expand the source load output sample set under extreme weather conditions, resulting in an expanded extreme weather source load output sample set as the expansion result.

4. The power system vulnerability assessment method considering the impact of extreme weather according to claim 3, characterized in that, The process of combining the augmented results with the time-varying fault probability of the line to generate typical scenarios through Monte Carlo simulation includes: Based on the time-varying fault probability of the line and the extended sample set of extreme weather source load output, several fault scenarios are obtained by constructing multi-time-period scenarios through sequential Monte Carlo. For each of the aforementioned fault scenarios, the power loss of the scenario is calculated as a clustering feature, and based on the clustering feature, the K-means clustering algorithm is used to cluster the fault scenarios to obtain typical scenarios.

5. The method for assessing the vulnerability of power systems considering the impact of extreme weather according to claim 2, characterized in that, The process takes the typical scenario as input, simulates the cascading failure evolution process under line faults and source-load fluctuations using an AC cascading failure model, and outputs the cascading failure evolution path and the system operating state at termination, including: Based on the typical scenario and the topology data, a node power balance equation is constructed to perform optimal AC power flow calculation and obtain node voltage and line power flow data. The state of the power system is checked based on the node voltage and the line power flow data. When the check result exceeds the limit, a protection mechanism is triggered to update the typical scenario. Based on the updated typical scenario, iteratively execute the optimal AC power flow calculation steps and system state verification steps until the preset iteration termination condition is met, and output the cascading failure evolution path and the system running state at the time of termination.

6. The method for assessing the vulnerability of a power system considering the impact of extreme weather, as described in claim 5, is characterized in that... The process involves analyzing primary surface weaknesses by constructing a severity model based on the cascading failure evolution path and the system's operating state at termination, and analyzing secondary internal structural weaknesses using spectral theory, including: The line's rated capacity percentage is determined based on the system's operating state at the time of termination. A line overload severity model is constructed based on the line's rated capacity percentage, and a node low-voltage severity model is constructed based on the system's operating state at the time of termination. Based on the line overload severity model and the node low voltage severity model, the expected risk corresponding to the surface weak link of the power system is quantified to obtain the first-level analysis results. The power flow weights of each transmission line are determined by the system operating state at the time of termination, in order to construct a weight matrix, and a Laplace matrix is ​​constructed based on the weight matrix. Based on the topology data and the cascading failure evolution path, a network adjacency matrix is ​​constructed and combined with the Laplace matrix to quantify the network structural vulnerability index corresponding to the inherent structural weaknesses of the power system, thereby obtaining secondary analysis results.

7. A method for assessing the vulnerability of a power system considering the impact of extreme weather, as described in claim 5, is characterized in that... The three-level cascading failure causal probability analysis based on Bayesian networks includes: Based on the time-series corrected wind speed, the power flow data of the line, and the cascading fault evolution path, wind speed nodes, power flow mutation nodes, and fault nodes are determined respectively. Edges are set based on the dependency relationship of the corresponding fault mechanism between the wind speed nodes, the power flow mutation nodes, and the fault nodes to construct a Bayesian network. A training dataset is constructed based on the wind speed node, the tidal current change node, and the fault node. The conditional probability of each node in the Bayesian network is trained using the training dataset to obtain the trained Bayesian network. The real-time power flow data and real-time meteorological data of the power system are input into the trained Bayesian network for processing, so as to realize the causal probability prediction of the cascading failures of the power system and obtain the three-level analysis results.

8. A power system vulnerability assessment system that considers the impact of extreme weather, characterized in that, include: The probability calculation module is used to acquire the topology data of the power system and its meteorological data and source-load data under extreme weather conditions, and to construct the line fault probability model of the power system based on the topology data and the meteorological data in order to calculate the time-varying fault probability of the line. The scenario generation module is used to expand the source load output sample set under extreme weather conditions based on the source load data by using a time series relative matching generative adversarial network model with gradient penalty, and combine the expansion result with the time-varying fault probability of the line to generate typical scenarios through Monte Carlo simulation. The fault evolution module is used to simulate the cascading fault evolution process under line faults and source load fluctuations through an AC cascading fault model, taking the typical scenario as input, and outputting the cascading fault evolution path and the system operating status at the time of termination. The multi-level assessment module is used to perform first-level surface vulnerability analysis by constructing a severity model based on the cascading failure evolution path and the system operating state at the time of termination, second-level internal structural vulnerability analysis by using spectral theory, and third-level cascading failure causal probability analysis based on Bayesian networks, so as to obtain the multi-level vulnerability assessment results of the power system.

9. A power system vulnerability assessment system considering the impact of extreme weather according to claim 8, characterized in that, The step of constructing a line fault probability model for the power system based on the topology data and the meteorological data to calculate the time-varying fault probability of the lines includes: A meteorological impact model is constructed based on the meteorological data, and geographical data of each transmission line in the power system is obtained. The meteorological impact model is then corrected using the geographical data to obtain the time-series corrected wind speed for each transmission line. Based on the time-series corrected wind speed and the topology data, an overall failure probability model for each of the transmission lines is constructed, and the service life of each of the overall failure probability models is corrected to obtain the line fault probability model of the power system in order to quantify the time-varying fault probability of the lines.

10. A power system vulnerability assessment system considering the impact of extreme weather according to claim 8, characterized in that, The expansion of the source load output sample set under extreme weather conditions based on the source load data, using a time-series relative matching generative adversarial network model with gradient penalty, includes: Features are extracted from the source load data and the meteorological data respectively to obtain a feature set, which is then used to reduce the dimensionality of the feature set through an embedding function to generate low-dimensional features. The low-dimensional features are reconstructed using a recovery function to obtain high-dimensional features, and the reconstruction loss between the low-dimensional features and the high-dimensional features is calculated. A feature vector set is constructed by random sampling, and the feature vector set is input into a generator for processing to obtain sample features. The supervision loss between the sample features and the low-dimensional features is calculated. The sample features and the low-dimensional features are input together into the discriminator for processing to obtain the sample probability and the true probability, and the gradient-penalized relative matching unsupervised loss between the sample probability and the true probability is calculated. Based on the reconstruction loss, the supervision loss, and the relative matching unsupervised loss, a total loss function is constructed. With the goal of minimizing the total loss function, the generator and the discriminator are alternately optimized through the backpropagation algorithm until a preset number of training iterations are reached. The optimized generator is then used to expand the source load output sample set under extreme weather conditions, resulting in an expanded extreme weather source load output sample set as the expansion result.

11. A power system vulnerability assessment system considering the impact of extreme weather according to claim 10, characterized in that, The process of combining the augmented results with the time-varying fault probability of the line to generate typical scenarios through Monte Carlo simulation includes: Based on the time-varying fault probability of the line and the extended sample set of extreme weather source load output, several fault scenarios are obtained by constructing multi-time-period scenarios through sequential Monte Carlo. For each of the aforementioned fault scenarios, the power loss of the scenario is calculated as a clustering feature, and based on the clustering feature, the K-means clustering algorithm is used to cluster the fault scenarios to obtain typical scenarios.

12. A power system vulnerability assessment system considering the impact of extreme weather according to claim 9, characterized in that, The fault evolution module includes: Based on the typical scenario and the topology data, a node power balance equation is constructed to perform optimal AC power flow calculation and obtain node voltage and line power flow data. The state of the power system is checked based on the node voltage and the line power flow data. When the check result exceeds the limit, a protection mechanism is triggered to update the typical scenario. Based on the updated typical scenario, iteratively execute the optimal AC power flow calculation steps and system state verification steps until the preset iteration termination condition is met, and output the cascading failure evolution path and the system running state at the time of termination.

13. A power system vulnerability assessment system considering the impact of extreme weather according to claim 12, characterized in that, The process involves analyzing primary surface weaknesses by constructing a severity model based on the cascading failure evolution path and the system's operating state at termination, and analyzing secondary internal structural weaknesses using spectral theory, including: The line's rated capacity percentage is determined based on the system's operating state at the time of termination. A line overload severity model is constructed based on the line's rated capacity percentage, and a node low-voltage severity model is constructed based on the system's operating state at the time of termination. Based on the line overload severity model and the node low voltage severity model, the expected risk corresponding to the surface weak link of the power system is quantified to obtain the first-level analysis results. The power flow weights of each transmission line are determined by the system operating state at the time of termination, in order to construct a weight matrix, and a Laplace matrix is ​​constructed based on the weight matrix. Based on the topology data and the cascading failure evolution path, a network adjacency matrix is ​​constructed and combined with the Laplace matrix to quantify the network structural vulnerability index corresponding to the inherent structural weaknesses of the power system, thereby obtaining secondary analysis results.

14. A power system vulnerability assessment system considering the impact of extreme weather according to claim 12, characterized in that, The three-level cascading failure causal probability analysis based on Bayesian networks includes: Based on the time-series corrected wind speed, the power flow data of the line, and the cascading fault evolution path, wind speed nodes, power flow mutation nodes, and fault nodes are determined respectively. Edges are set based on the dependency relationship of the corresponding fault mechanism between the wind speed nodes, the power flow mutation nodes, and the fault nodes to construct a Bayesian network. A training dataset is constructed based on the wind speed node, the tidal current change node, and the fault node. The conditional probability of each node in the Bayesian network is trained using the training dataset to obtain the trained Bayesian network. The real-time power flow data and real-time meteorological data of the power system are input into the trained Bayesian network for processing, so as to realize the causal probability prediction of the cascading failures of the power system and obtain the three-level analysis results.

15. An electronic device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the power system vulnerability assessment method considering the impact of extreme weather as described in any one of claims 1 to 7.

16. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein when the device containing the computer-readable storage medium executes the computer program, it implements the power system vulnerability assessment method considering the impact of extreme weather as described in any one of claims 1 to 7.

Images