A Dynamic Assessment and Decision Support Method for Distribution Network Resilience

Abstract

This invention discloses a dynamic assessment and decision support method for distribution network resilience, relating to the field of power system security protection technology. The method includes: constructing a dynamic risk field, integrating external threat trajectories and equipment vulnerabilities, and calculating the time-varying failure probability of nodes; setting key weights based on topology and user value to form value loads; generating a large number of fault evolution sequences through Monte Carlo simulation, and extracting impact velocity and impact magnitude features for intelligent clustering and screening to obtain effective fault evolution sequences; performing fine-step simulation on these sequences, plotting resilience curves, and extracting multi-dimensional dynamic resilience indicators such as robustness depth and loss area; finally, adaptively classifying scenarios based on impact index and robustness depth to generate ranked decision support information packages. This invention achieves a shift from static, passive assessment to dynamic, proactive prediction and scientific decision-making, effectively improving the resilience of distribution networks in the face of extreme events.

Description

Technical Field

[0001] This invention relates to the field of power system security protection technology, and more specifically, to a method for dynamic assessment and decision support of distribution network resilience. Background Technology

[0002] The distribution network is the end-user (home, factory, commercial center) power supply network in the power system. It reduces and distributes electrical energy from the high-voltage transmission backbone network step by step, and finally delivers it safely and stably to every electrical device. It is characterized by its wide coverage and complex structure (radial or ring network, containing a large number of switches, transformers, and protection devices). The operating status of the distribution network directly determines the power supply quality for users. Any fault may cause interruption of social production and life. Therefore, conducting a systematic and forward-looking resilience assessment of the distribution network is an inevitable choice to identify weak links, quantify potential socio-economic losses, and scientifically guide investment to defend against known and unknown risks. It has irreplaceable strategic necessity. Currently, in the planning and operation assessment of distribution networks, the quantification of resilience is generally simplified to the calculation and optimization of individual, static, and outcome-based indicators. The most typical approach is to use outage duration, number of outages, or amount of power outages as the core or even the sole evaluation criteria. These indicators are derived from the traditional reliability assessment system, and their essence is a passive statistical analysis of the consequences of failures. When faced with low-probability, high-impact extreme events (such as typhoons and floods), they can lead to serious technical misjudgments and decision-making misguidance, failing to depict the dynamic process of distribution network resilience, resulting in distorted assessment results and unsupported decision-making. Summary of the Invention

[0003] The main objective of this invention is to provide a method for dynamic assessment and decision support of distribution network resilience, so as to overcome the problems mentioned in the background art.

[0004] To achieve the above objectives, a method for dynamic assessment and decision support of distribution network resilience is provided, comprising the following steps: Obtain the topology diagram of the distribution network and identify the transmission routes from the power source to each power terminal based on the topology diagram; preset several threat scenarios, each of which consists of the trajectory of physical threat intensity that evolves with time and space, and combine the geographical location and physical vulnerability characteristics of the distribution network nodes, calculate the failure probability of each node under the threat scenario that changes with time through the corresponding vulnerability function; Based on the node failure probability, Monte Carlo simulation is performed using a rolling time window method to generate a large number of node failure evolution sequences. Representative effective failure evolution sequences are then selected from these sequences through cluster analysis. For each effective failure evolution sequence, hourly distribution network operation simulation is performed to construct a resilience curve of the value load retention rate over time, and resilience indicators are extracted from the resilience curve. Based on the impact characteristics and resilience indicators of effective fault evolution sequences, the fault evolution sequences are classified into different scenarios, and decision support information packages corresponding to different scenario types are generated to assist in the operation and management decisions of the distribution network; the impact characteristics include impact velocity and impact magnitude.

[0005] Furthermore, the failure probability of each node over time under threat scenarios: Obtain the topology diagram of the distribution network, which is marked with all lines from the distribution network to each power terminal and all nodes on the lines. Several threat scenarios are pre-defined, and each threat scenario is defined as a physical threat intensity trajectory that evolves in time and space. The physical threat intensity trajectory is combined with the geographical location and physical vulnerability characteristics of the power grid assets. By using a predefined vulnerability function corresponding to each threat scenario, the real-time failure probability Pn(t) of each node under this threat scenario can be mapped to change with time and space. Here, n represents the index of any node, n∈N, and N represents the total number of nodes in the distribution network topology diagram. Based on the topology diagram of the power distribution network, the power transmission lines are decomposed to obtain the transmission lines for each power consumer. Each transmission line refers to the power consumer and the nodes through which the power is transmitted. A key weight is assigned to each power consumer based on its type. For each power consumer in a transmission line, its physical load power is multiplied by its corresponding key weight to obtain the value load of that power consumer in the transmission line. For nodes that are not directly loaded in the transmission line, their value load is defined as the sum of the value loads of all downstream power consumers, representing the total value loss downstream due to the failure of that node. Then, based on the failure probability Pn(t) of all nodes in the distribution network topology diagram, the fault probability sets corresponding to each transmission line are divided according to the transmission lines, denoted as... , where i is the index of each node in the transmission line; the fault probability set includes the fault probability of each node in the transmission line.

[0006] Further, Monte Carlo simulation: The Monte Carlo simulation is run in a rolling window manner. At each future time step, random fault sampling is performed on the node to obtain possible fault evolution sequences. Each fault evolution sequence represents a possible future encounter. Cluster analysis is performed on a large number of fault evolution sequences to screen out representative fault evolution sequences at the node as valid fault evolution sequences for the node.

[0007] Further, cluster analysis: S201, extract each fault evolution sequence, calculate the change of the cumulative number of fault nodes over time, and obtain the impact velocity characterizing the fault propagation speed through linear fitting; S202, extract the fault timing and value load information of nodes in each fault evolution sequence, and calculate the impact quantity that characterizes the degree of time-series value damage through a preset time enhancement penalty function; S203, construct standardized feature vectors for each sequence based on the impact velocity and the impact amount, divide the sequence into multiple clusters using a clustering algorithm, calculate the screening value for each cluster, and determine the effective fault evolution sequence based on the screening value of each cluster.

[0008] Further, the calculation process for the impact velocity: The fault evolution sequence is extracted, and the cumulative number of fault nodes from the start time to each time step is calculated. A two-dimensional rectangular coordinate system is constructed with the time step as the abscissa and the cumulative number of fault nodes at each time step as the ordinate. The cumulative number of fault nodes at each time step is then input into the coordinate system according to its corresponding time step to form several discrete points. The least squares method is used to perform linear fitting on these discrete points to obtain the linear equation. The slope of the linear equation is extracted as the fault accumulation rate, which represents the average number of newly added fault nodes per time step within the entire observation time window, i.e., the impact speed of the transport route under the threat scenario.

[0009] Further, the calculation process for the impact amount: Extract the time step number of each fault node in the fault evolution sequence, and select the maximum value of the fault occurrence time step among all fault nodes in the sequence. For each fault node i in the fault evolution sequence, subtract the time step number of fault node i from the maximum value of the fault occurrence time step to obtain the fault advance interval, denoted as Δti. This is then calculated using the formula... Calculate the premature failure enhancement penalty to obtain the time value penalty qi of a single node, where Li is the value load of the node, λ is the time enhancement factor, λ>0, which controls the amplification effect of the time advance on the severity of the fault; sum the time value penalties of all fault nodes in the fault evolution sequence to obtain the impact of the fault evolution sequence.

[0010] Furthermore, valid fault evolution sequences are determined based on the screening values ​​for each cluster: A raw two-dimensional vector is constructed from the impact velocity and impact magnitude of each fault evolution sequence. Z-score is used to standardize all raw two-dimensional vectors to obtain standardized two-dimensional vectors. The similarity between the two-dimensional vectors corresponding to any two fault evolution sequences is calculated using a Gaussian kernel function. A Laplacian matrix is ​​constructed based on the similarity, and K-means clustering is performed to determine the optimal number of clusters, m. All fault evolution sequences are divided into m clusters, each representing a common fault impact pattern. After clustering, the number of fault evolution sequences in each cluster is counted, and this number is divided by the reference number of fault evolution sequences to obtain the cluster proportion, denoted as g1. The impact magnitudes of each fault evolution sequence are summed to obtain the total impact, and this total impact is divided by the impact reference to obtain the impact proportion, denoted as g2. The geometric mean formula is then used. The selection value G is calculated. A preset screening threshold is set. When the screening value is greater than the screening threshold, the cluster is recorded as a representative cluster. The number of representative clusters is counted. If the number of representative clusters is zero, the cluster with the largest screening value is selected as the representative cluster. The fault evolution sequence corresponding to the two-dimensional vector closest to the centroid of the cluster in each representative cluster is selected as the effective fault evolution sequence. Thus, the effective fault evolution sequence of each transport route under the threat scenario can be obtained.

[0011] Furthermore, resilience metrics include robustness depth, resilience loss area, recovery time, and recovery speed. Robustness depth refers to the lowest trough of the resilience curve, i.e., the minimum value load retention rate, which characterizes the severity of the system's functional decline under impact. Resilience loss area refers to the area enclosed by the curve and the 100% horizontal line, which comprehensively measures the scale and duration of performance loss. Recovery time refers to the time required for the curve to recover from the bottom to the basic retention rate, which characterizes the speed of restoring core functions. Recovery speed is calculated by subtracting robustness depth from basic retention rate and then dividing by critical recovery time, which is used to quantify the efficiency of recovery actions.

[0012] Furthermore, the scenario type is determined from the fault evolution sequence: A weight is assigned to the impact velocity and impact magnitude of the impact feature vector of the effective fault evolution sequence. Then, the impact velocity and impact magnitude are linearly weighted and fused to obtain the impact index of the effective fault evolution sequence. An impact threshold and a robustness threshold are preset. If the impact index is greater than or equal to the impact threshold and the robustness depth is less than or equal to the robustness threshold, the effective fault evolution sequence is classified as a rapid paralysis type fault. If the impact index is greater than or equal to the impact threshold and the robustness depth is greater than the robustness threshold, the effective fault evolution sequence is classified as a slow erosion type fault. If the impact index is less than the impact threshold and the robustness depth is less than or equal to the robustness threshold, the effective fault evolution sequence is classified as a critical hub failure type fault. If the impact index is less than the impact threshold and the robustness depth is greater than the robustness threshold, the effective fault evolution sequence is classified as a general fault type. The effective fault evolution sequences corresponding to each transport route are sorted in chronological order according to the impact index, and the fault types and resilience assessment indicators corresponding to the effective fault evolution sequences are used as decision support information packages.

[0013] The beneficial effects of this invention are: (1) By backtracking the power supply line of each power supply terminal through the topology diagram, and assigning key weights to each power supply terminal according to the multi-criteria decision-making method, the factors such as socio-economic, public safety, and critical infrastructure are structured into calculable weight coefficients. The physical load is then multiplied by the weights to obtain the value load. For non-terminal nodes, the value of the node is defined by summing the downstream value loads, forming a loss benchmark measured by value rather than simply the number of power users. This can effectively change the operation and maintenance decision-making objective from minimizing the number of power outage users to maximizing the availability of critical services, making resource allocation more in line with social benefits. (2) By defining and calculating the two characteristic dimensions of impact velocity and impact amount from the perspectives of spatiotemporal propagation mode and temporal value damage, the essence of the fault evolution sequence is accurately characterized, providing a clear physical meaning for mathematical clustering; then, based on the designed geometric mean screening value, an adaptive clustering representative screening mechanism is established, and within this screening mechanism, the cluster occurrence scale (probabilistic representativeness) and the severity of consequences (impact representativeness) are considered, and the screening threshold is objectively set to 1, so as to intelligently and automatically screen out the high-frequency risks and extreme threat scenarios that truly need to be focused on, ensuring that subsequent detailed simulation can cover the main risk modes and compress the amount of computation to an acceptable range for engineering. (3) By coupling the impact index, which is a weighted composite of impact velocity and impact amount, with the robustness depth in the resilience index (and using an adaptive threshold based on sample quantiles, such as 75% / 25%), the effective fault evolution sequences are classified into rapid paralysis, slow erosion, critical hub failure and general faults. Then, the sequence is prioritized according to the impact index and the fault type and resilience index of each sequence are packaged into an executable decision support package. On the one hand, classification helps distribution network managers to differentiate the strategies for different impact modes. On the other hand, sorting and probability weights ensure that limited operation and maintenance resources are prioritized for the situation most likely to cause the greatest social loss, thereby realizing the transformation from passive statistics to active prediction. Attached Figure Description

[0014] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0015] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0016] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0017] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of the invention described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0018] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0019] Please see Figure 1 As shown, this invention provides a dynamic assessment and decision support method for the resilience of a distribution network, comprising the following steps: Step 100: Obtain the topology diagram of the distribution network. The topology diagram is marked with all lines from the distribution network to each power consumer terminal and the nodes on the lines. A node refers to a specific device or electrical component in the distribution network. It is the most basic conceptual unit in network modeling and is used to abstractly represent the physical structure and electrical connection relationship of the power grid. Based on the topology diagram of the power transmission of the distribution network, the lines are decomposed to obtain the power transmission lines to each power consumer terminal. The specific transmission line refers to the line consisting of the power consumer terminal and the nodes through which the power is transmitted to the power consumer terminal. A critical weight is set for each power consumer terminal according to its type. Specifically, the critical weight is a value set by a structured value quantification process based on multi-criteria decision-making. Its core objective is to transform the social and economic importance of the power consumer terminal into a calculable weight coefficient, thereby guiding the resilience assessment to focus on ensuring the most critical social functions. For the power consumer terminal in the transmission route, its physical load power is multiplied by its corresponding critical weight to obtain the value load of the power consumer terminal in the transmission route. Then, for nodes that are not directly loaded in the transmission route, their value load is defined as the sum of the value loads of all downstream power consumers terminals, which represents the total value of downstream losses caused by the failure of this node. Step 200: Several threat scenarios are pre-defined, each defined as a physical threat intensity trajectory (e.g., water level, wind speed field) evolving in time and space. The physical threat intensity trajectory is combined with the geographical location and physical vulnerability characteristics (e.g., altitude, wind resistance level) of power grid assets. Using a predefined vulnerability function corresponding to each threat scenario, the real-time failure probability Pn(t) of each node under this threat scenario, varying with time and space, can be mapped to this probability. Here, n represents the index of any node, n∈N, and N represents the total number of nodes in the distribution network topology diagram. It should be noted that in practical applications, based on meteorological and disaster forecasting and... Based on vulnerability information, a set of representative threat scenarios is constructed. Each threat scenario is defined as one or more threat intensity trajectories evolving over time (e.g., the temporal-spatial intensity field of a storm path, the evolution of rainfall and flash floods). In practical applications, under flood scenarios, the real-time inundation probability of each node is calculated based on node elevation and real-time predicted water level (i.e., the physical threat intensity trajectory) and the vulnerability function of each node under the predicted water level. The real-time inundation probability is then the failure probability of each node. Specific physical threat intensity trajectories under other threat scenarios include: the physical threat intensity trajectory under typhoon scenarios refers to the predicted wind speed; the physical threat intensity trajectory under earthquake scenarios refers to the ground motion peak. In the context of acceleration and wildfire scenarios, the physical threat intensity trajectory refers to radiative heat flux. It's important to note that the vulnerability function establishes a deterministic or probabilistic mapping relationship between the intensity of external physical threats and the probability of power grid equipment failure. Its design strictly relies on the physical failure mechanisms of equipment and historical disaster damage statistics. Specifically, for flood scenarios, the vulnerability function typically manifests as a step function or sigmoid function based on a water level-elevation threshold. The principle is that when the flood depth exceeds the installation elevation of critical equipment components (such as transformer bases and switchgear bottoms), the probability of insulation failure or short circuit increases abruptly or continuously. For typhoon scenarios, the function often employs a Weberian function based on extremum theory. The functions, based on the Goethe distribution or log-normal distribution, correlate wind speed with the design wind resistance rating of towers and conductors, reflecting the probability that wind speed exceeds the structural bearing capacity limit. For earthquake scenarios, the functions generally rely on seismic fortification intensity and structural dynamic response, using vulnerability curves from the specifications to estimate the probability of different damage states through peak ground acceleration and equipment natural vibration periods. For wildfire scenarios, the functions are based on thermal radiation flux attenuation models and insulation material tolerance thresholds to calculate the probability of equipment failure due to air breakdown or material overheating at a specific distance. All functions must be calibrated and verified through laboratory tests, field observation data, or historical disaster cases to ensure their scientific validity and engineering applicability. By integrating external threat trajectories with equipment physical vulnerability functions, the probability of failure of each power grid device (node) at any future time is calculated in real time and dynamically. This transforms potential qualitative risks into quantitative data on the probability of their occurrence and when, laying a precise data foundation for forward-looking assessments.

[0020] Based on the fault probability Pn(t) of all nodes in the distribution network topology diagram, the fault probability set corresponding to each transmission line is obtained by dividing the network according to the transmission lines, denoted as . , where i is the index of each node in the transmission line; the fault probability set includes the fault probability of each node in the transmission line; By constructing the transmission lines of each power user terminal through a topology diagram and assigning critical weights to each terminal using a multi-criteria decision-making method, factors such as socio-economic factors, public safety, and critical infrastructure are structured into calculable weight coefficients. The physical load is then multiplied by the weights to obtain the value load. For non-terminal nodes, the node value is defined by summing the downstream value loads, forming a loss benchmark measured by value rather than simply the number of power users. This effectively shifts the operation and maintenance decision-making objective from minimizing the number of power outage users to maximizing the availability of critical services, making resource allocation more aligned with social benefits.

[0021] Monte Carlo simulations are run using a rolling window approach (e.g., updated every 15 minutes; the specific window width can be adjusted based on actual conditions). At each future time step, random fault sampling is performed on the nodes to obtain possible fault evolution sequences, each representing a possible future encounter. In practical applications, running Monte Carlo simulations yields a large number of fault evolution sequences for each node at each future time step. Directly performing full-step simulation on all sequences would result in an unbearable computational burden. Therefore, cluster analysis is needed to select representative fault evolution sequences with high probability of occurrence or extremely severe consequences at the nodes, which serve as the effective fault evolution sequences for the nodes. The specific process of cluster analysis on a large number of fault evolution sequences is as follows: S201. Extract the fault evolution sequence and calculate the cumulative number of fault nodes from the start time to each time step. It should be noted that the cumulative number of fault nodes here is a unique node to avoid distorting the fault rate by repeatedly counting the same node in different time steps. Construct a two-dimensional rectangular coordinate system with the time step as the x-axis and the cumulative number of fault nodes at each time step as the y-axis. Then, input the cumulative number of fault nodes at each time step into the coordinate system according to its corresponding time step to form several discrete points. Use the least squares method to perform linear fitting on these discrete points to obtain a straight line equation. Extract the slope of the straight line equation as the fault accumulation rate, which represents the average number of newly added fault nodes per time step within the entire observation time window, i.e., the impact speed of the transport route under the threat scenario. S202, extract the time step number of each fault node in the fault evolution sequence, and select the maximum value of the fault occurrence time step among all fault nodes in the sequence; for each fault node i in the fault evolution sequence, subtract the time step number of fault node i from the maximum value of the fault occurrence time step to obtain the fault advance interval, denoted as Δti. The larger this value, the earlier the failure of this node occurs relative to the latest fault node in the sequence. If the value load of the node is larger, the reinforcement effect is stronger at this time, which also means that the fault impact on the transportation route is more intense under this threat scenario; through the formula The premature failure amplification penalty is calculated to obtain the time value penalty qi of a single node, where Li is the value load of the node, λ is the time amplification factor, and λ > 0 controls the amplification effect of the time advance on the severity of the failure. The core of this formula is time amplification, which represents a standard exponential growth function, transforming the linear failure advance interval into a nonlinear amplification coefficient. When the failure advance interval is equal to zero, that is, the node fails only at the end of the sequence, the time value penalty at this time is the node value load itself. Then, the time value penalties of all failed nodes in the failure evolution sequence are summed to obtain the impact of the failure evolution sequence. From the above analysis of the impact of the failure evolution sequence, it can be seen that when the loss of the total value load is the same, in sequence A: all nodes with high value loads fail in the early stage of the disaster (i.e., when Δti is larger), while in sequence B: the same high value nodes fail only at the end of the disaster (when Δti is smaller). The impact of sequence A is much higher than that of sequence B, which means that under the same loss scale, the impact of a catastrophic path that occurs faster and catches the system more off guard is greater. S203. An original two-dimensional vector is constructed from the impact velocity and impact amount of each fault evolution sequence. This vector simultaneously captures the spatiotemporal impact pattern (i.e., fault propagation speed) and temporal value damage (failure degree of high-value nodes) of the fault evolution sequence. Thus, all fault evolution sequences of the transport route can be transformed into the original two-dimensional vector. Z-score is used to standardize all the original two-dimensional vectors to obtain standardized two-dimensional vectors. The specific standardization process is as follows: for a certain feature (impact velocity or impact amount) in all the original two-dimensional vectors, its mean and standard deviation are calculated. Then, the mean is subtracted from each feature value, and then the result is divided by the standard deviation to obtain the standardized feature. This completes the standardization of the impact velocity and impact amount in the two-dimensional vector, and the standardized two-dimensional vector is labeled as (V, Q). Since impact velocity and impact amount usually have different dimensions and orders of magnitude, direct clustering will lead to the large-order-of-magnitude feature dominating the analysis results. The two features of the standardized vector are on the same scale, which can accurately reflect the contribution of each vector feature in the clustering. The Gaussian kernel function is used to calculate the similarity between the two-dimensional vectors corresponding to any two fault evolution sequences. A Laplacian matrix is ​​constructed based on the similarity, and then the K-means algorithm is used for clustering to determine the optimal number of clusters, *m*. All fault evolution sequences are divided into *m* clusters, each representing a common fault impact pattern. After clustering, the number of fault evolution sequences in each cluster is counted, and this number is divided by the reference number of fault evolution sequences (the total number of fault evolution sequences divided by the number of clusters *m*) to obtain the cluster ratio, denoted as *g1*. The impact magnitudes of each fault evolution sequence are summed to obtain the total impact, and this total impact is then divided by the impact reference (the sum of the impact magnitudes of all fault evolution sequences divided by the number of clusters *m*) to obtain the impact ratio, denoted as *g2*. The geometric mean formula is then used. The screening value G is calculated. In this formula, the cluster proportion and impact proportion of each cluster comprehensively reflect the excess representativeness of a cluster relative to the average level in terms of quantity and impact. The synthesis method uses geometric mean to ensure that the two proportions are balanced, avoiding the situation where a cluster is selected even when one proportion is extremely high and the other is extremely low. Thus, the screening value of each cluster can be obtained. The screening threshold is preset. The core of the screening threshold is to establish an objective, adaptive statistical discrimination standard with clear physical meaning, which is usually set to one. The screening value is essentially the geometric mean of the cluster proportion (scale representativeness) and the impact proportion (severity of consequences). When the screening value is equal to one, it indicates that the cluster's comprehensive performance in terms of both scale and impact is exactly at the average level of all clusters. When the screening value is greater than one, it indicates that the cluster has excess representativeness above the average level in two or at least one aspect. Therefore, setting the threshold to 1 allows for the automatic statistical selection of clusters that are prominent in either the probability of occurrence or the impact consequences, or balanced in both aspects, as representative clusters. When the selection value is greater than the threshold, it indicates that the cluster is representative in terms of both quantity (scale) and impact degree, and this cluster is recorded as a representative cluster. The number of representative clusters is counted. If the number of representative clusters is zero, the cluster with the largest selection value is selected as the representative cluster. In practical applications, this situation is a very extreme case and almost never occurs; however, for the completeness of this scheme, the case when the number of representative clusters is zero is still given. The cluster with the most prominent value among the representative clusters is selected. The fault evolution sequence corresponding to the two-dimensional vector near the centroid of the cluster is taken as the effective fault evolution sequence. Thus, the effective fault evolution sequence of each transmission route under the threat scenario can be obtained. Through steps S201 to S203, the intelligent compression process from the generation of massive random sequences to the extraction of representative fault modes is completed. This process relies closely on the core concepts such as value load and dynamic fault probability defined in the early stage, ensuring that the selected sequences are not only mathematically representative, but also correspond to high-frequency risks and extreme threats in the physical sense of power grid resilience, laying a solid foundation for subsequent precise prevention and recovery decisions. By defining and calculating the impact velocity and impact magnitude from the perspectives of spatiotemporal propagation patterns and temporal value damage, the essence of the fault evolution sequence is accurately characterized, providing a physically meaningful tool for mathematical clustering. Then, based on the designed geometric mean screening value, an adaptive clustering representative screening mechanism is established. Within this screening mechanism, the cluster occurrence scale (probabilistic representativeness) and the severity of consequences (impact representativeness) are considered, and the screening threshold is objectively set to 1. This achieves intelligent and automated screening of high-frequency risks and extreme threat scenarios that truly require key attention, ensuring that subsequent detailed simulations can cover the main risk patterns while compressing the computational load to an engineering-acceptable range.

[0022] Step 300: Based on the effective fault evolution sequences of each transmission route, perform time-step simulation and accurately plot resilience curves, and accordingly quantify the dynamic performance response of the distribution network under threat scenario modes, generating a resilience assessment map that can support decision-making; the specific process is as follows: S301 establishes an independent simulation environment for each selected valid fault evolution sequence. This environment starts with the current real-time topology and operating status of the distribution network, and fully loads the deterministic fault event list defined by the sequence and unfolded according to future time steps. The specific list clearly records one or more specific fault nodes (such as the collapse of tower A001, short circuit of substation S1 bus) to be injected at each simulation time step, completing the transformation from probabilistic risk extrapolation to deterministic stress test scenarios. The simulation engine advances step by step at fixed time steps (the time step here is consistent with the Monte Carlo simulation window, such as 15 minutes). Within each time step, the following core calculations are executed sequentially, and the process strictly follows the physical laws and operational control logic of the power grid: (1) Fault injection and topology update: Based on the definition of effective fault evolution sequence, the fault event that is planned to occur at this time step is applied to the network model, disconnecting the corresponding line or disconnecting the relevant equipment, thereby changing the physical connection relationship of the power grid; (2) Protection action simulation and fault isolation: Simulate the response of the relay protection system, automatically trip the circuit breaker adjacent to the fault node, isolate the fault node from the system, and form a new network topology without short-circuit faults; (3) Network Reconstruction and Power Restoration Attempts: After isolating the faulty node, the system automatically attempts to reconstruct the network by operating the remaining tie switches and sectionalizing switches. The goal is to restore power to as many non-faulty areas as possible, forming one or more new power supply islands; this process is optimized by rapidly searching to maximize the restoration of the weighted value load at the end of the feeder. (4) Distributed resource scheduling and island operation: In the formed power supply island, if there are distributed photovoltaic, energy storage or emergency generators, they will be scheduled according to their capacity and current status (such as energy storage SOC) in accordance with the predetermined island operation optimization strategy to support the critical loads in the island as much as possible. (5) Power flow calculation and safety verification: Perform power flow calculation on the reconstructed network to verify whether the voltage exceeds the limit and whether the line is overloaded. If the limit is exceeded, adjust the output of distributed resources or perform load shedding to ensure the physical feasibility of the simulation results; (6) Performance index calculation: Calculate the total value load that the entire network can still maintain power supply at the end of this time step (i.e., the sum of the value load of all power terminals that have not lost power). After completing the full-process simulation of a single effective fault evolution sequence, the total value load is divided by the initial total value load of the distribution network before the disaster to obtain the value load retention rate. A two-dimensional coordinate system is constructed with time as the horizontal axis and the system value load retention rate at each step of the effective fault evolution sequence as the vertical axis. Points are plotted to create curves, which are the precise resilience curves under this threat scenario. From each precise resilience curve, multiple quantitative resilience indicators can be decoupled and extracted. The specific resilience indicators include robustness depth d, resilience loss area S, recovery time T, and recovery speed R. Among them, robustness depth... The degree refers to the lowest trough of the curve, i.e., the minimum value load retention rate, which characterizes the severity of the system's functional decline under shock; the resilience loss area S refers to the area enclosed by the 100% level line below the curve, which comprehensively measures the scale and duration of performance loss; the recovery time refers to the time required for the curve to recover from the bottom to the basic retention rate (95% level), which characterizes the speed of restoring core functions; the recovery speed is calculated by subtracting the robustness depth d (i.e., the minimum value load retention rate in the curve) from the basic retention rate and then dividing by the critical recovery time T, which is used to quantify the efficiency of recovery actions; After the simulation of all effective fault evolution sequences is completed, a set of toughness curves covering different impact modes and their corresponding toughness indices are obtained. Based on the impact characteristics (i.e., standardized two-dimensional vectors) of the effective fault evolution sequences and the toughness assessment indices, a comprehensive scenario evaluation and decision support package are performed. The specific process is as follows: The impact velocity and impact magnitude of the impact feature vector of the effective fault evolution sequence are each assigned a weight. Then, the impact velocity and impact magnitude are linearly weighted and fused to obtain the impact index of the effective fault evolution sequence. In practical applications, the weights of the two can be adjusted according to the degree of attention or sensitivity of the application scenario to the impact velocity and impact magnitude. For example, in the typhoon scenario, more attention is paid to the instantaneous destructive capability of the disaster (i.e., impact velocity), so the weight corresponding to the impact velocity is set to 0.6 and the weight corresponding to the impact magnitude is set to 0.4. The system presets an impact threshold and a robustness threshold. The impact threshold is typically set as the upper quartile (e.g., the 75th quartile) of the impact index of all valid fault evolution sequences. This threshold represents the boundary of the top 25% of high-impact scenarios under the current threat scenario, and can adapt to the specific risk distribution of this assessment to screen out sequences with significant external threats. The robustness threshold is typically set as the lower quartile (e.g., the 25th quartile) of the robustness depth d value of all sequences. This threshold identifies the 25% of scenarios with the most severe performance degradation, representing the critical level of severe functional impairment, and is used to identify fault modes that, regardless of the size of the external impact, have caused deep damage to the system's internal structure. If the impact index is greater than or equal to the impact threshold, and the robustness depth is less than or equal to the robustness threshold, it indicates that the valid fault evolution sequence simultaneously possesses strong external impact and weak internal system defenses, and the distribution network faces a high risk of instantaneous collapse. Its characteristics are a sudden drop in function and urgent recovery. In this case, the valid fault evolution sequence is... The effective fault evolution sequence is classified into the rapid paralysis type. If the impact index is greater than or equal to the impact threshold and the robustness depth is greater than the robustness threshold, it indicates that the external impact of the effective fault evolution sequence is continuous and strong, but the initial structure of the distribution network still has a certain resilience. The function is slowly eroded under continuous pressure, the cumulative loss is huge, and the recovery process is long. In this case, the effective fault evolution sequence is classified into the slow erosion type. If the impact index is less than the impact threshold and the robustness depth is less than or equal to the robustness threshold, it indicates that although the external impact range or intensity of the effective fault evolution sequence is not prominent, it accurately hits the key nodes in the power grid topology and has a very serious impact. In this case, the effective fault evolution sequence is classified into the critical hub disabling type. If the impact index is less than the impact threshold and the robustness depth is greater than the robustness threshold, it indicates that the impact of the effective fault evolution sequence is limited and the power grid has strong resistance. It belongs to the category of predictable and controllable conventional faults and can be handled according to standard procedures. In this case, the effective fault evolution sequence is classified into the general fault type. The effective fault evolution sequences corresponding to each transmission route are sorted in chronological order according to the impact index. The fault types and resilience assessment indicators corresponding to the effective fault evolution sequences are used as decision support information packages to help distribution network operation and management personnel shift from passive response and experience-based decision-making to proactive foresight and scientific decision-making. Ultimately, this aims to achieve the core objectives of minimizing power supply losses, maximizing critical load protection, and optimizing social impact under extreme events. Before a disaster occurs, avoidable losses are prevented to the greatest extent possible. After a disaster occurs, unavoidable losses are restored with the highest efficiency, thereby effectively improving the resilience level of the distribution network.

[0023] By coupling the impact index, a weighted composite of impact velocity and impact magnitude, with the robustness depth in the resilience index (and using an adaptive threshold based on sample quantiles, such as 75% / 25%), effective fault evolution sequences are classified into rapid paralysis, slow erosion, critical hub failure, and general faults. Subsequently, the sequences are prioritized according to the impact index, and the fault type and resilience index of each sequence are packaged into an executable decision support package. On the one hand, classification helps distribution network managers to differentiate their strategies for different impact modes; on the other hand, ranking and probability weights ensure that limited operation and maintenance resources are prioritized for situations most likely to cause the greatest social losses, thereby achieving a shift from passive statistics to proactive prediction.

[0024] The above are merely embodiments of the present invention and are not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A method for dynamic assessment and decision support of distribution network resilience, characterized in that, Includes the following steps: Obtain the topology diagram of the power distribution network and identify the transmission routes from the power source to each power consumer based on the topology diagram; Several threat scenarios are preset, each of which consists of a trajectory of physical threat intensity that evolves over time and space. Combined with the geographical location and physical vulnerability characteristics of the distribution network nodes, the failure probability of each node under the threat scenario is calculated over time using the corresponding vulnerability function. Based on node failure probabilities, Monte Carlo simulations are performed using a rolling time window approach to generate a large number of node failure evolution sequences. Representative and effective failure evolution sequences are then selected through cluster analysis. For each effective failure evolution sequence, hourly distribution network operation simulations are performed to construct a resilience curve of the value load retention rate over time, and resilience indicators are extracted from this curve. The cluster analysis includes: S201, extract each fault evolution sequence, calculate the change of the cumulative number of fault nodes over time, and obtain the impact velocity characterizing the fault propagation speed through linear fitting; S202, extract the fault timing and value load information of nodes in each fault evolution sequence, and calculate the impact quantity, which characterizes the degree of time-series value damage, through a preset time-enhancing penalty function; the calculation process of the impact quantity is as follows: Extract the time step number of each fault node in the fault evolution sequence, and select the maximum value of the fault occurrence time step among all fault nodes in the time step sequence; for each fault node in the fault evolution sequence... i Subtract the maximum value of the step when the fault occurs from the value of the fault node. i The fault advance interval obtained from the time step sequence number is denoted as Δ. ti Through formula Calculate the premature failure enhancement penalty to obtain the time value penalty for a single node. qi ,in Li Let λ be the value load of the node, and λ be the time enhancement factor. λ > 0 controls the amplification effect of the time advance on the severity of the fault. The time value penalty of all fault nodes in the fault evolution sequence is summed to obtain the impact of the fault evolution sequence. S203, construct standardized feature vectors for each sequence based on the impact velocity and the impact amount, divide the sequence into multiple clusters using a clustering algorithm, calculate the screening value for each cluster, and determine the effective fault evolution sequence based on the screening value of each cluster; Based on the impact characteristics and resilience indicators of effective fault evolution sequences, the fault evolution sequences are classified into different scenarios, and decision support information packages corresponding to different scenario types are generated to assist in the operation and management decisions of the distribution network; the impact characteristics include impact velocity and impact magnitude.

2. The method for dynamic assessment and decision support of distribution network resilience according to claim 1, characterized in that, Failure probability of each node over time under threat scenarios: Obtain the topology diagram of the distribution network, which is marked with all lines from the distribution network to each power terminal and all nodes on the lines. Several threat scenarios are pre-defined, and each threat scenario is defined as a physical threat intensity trajectory that evolves in time and space. The physical threat intensity trajectory is combined with the geographical location and physical vulnerability characteristics of the power grid assets. Using a predefined vulnerability function corresponding to each threat scenario, the real-time failure probability Pn(t) of each node under this threat scenario is mapped to change with time and space, where n represents the index of any node, n∈N, and N represents the total number of nodes in the distribution network topology diagram. Based on the topology diagram of the power distribution network, the power transmission lines are decomposed to obtain the transmission lines for each power consumer. Each transmission line refers to the power consumer and the nodes through which the power is transmitted. A key weight is assigned to each power consumer based on its type. For each power consumer in a transmission line, its physical load power is multiplied by its corresponding key weight to obtain the value load of that power consumer in the transmission line. For nodes that are not directly loaded in the transmission line, their value load is defined as the sum of the value loads of all downstream power consumers, representing the total value loss downstream due to the failure of that node. Based on the failure probability Pn(t) of all nodes in the distribution network topology diagram, the failure probability set corresponding to each transmission line is obtained by dividing the network according to the transmission lines, denoted as... , where i is the index of each node in the transmission line; the fault probability set includes the fault probability of each node in the transmission line.

3. The method for dynamic assessment and decision support of distribution network resilience according to claim 2, characterized in that, Monte Carlo simulation: The Monte Carlo simulation is run in a rolling window manner. At each future time step, random faults are sampled at the node to obtain a fault evolution sequence. Each fault evolution sequence represents a possible future encounter. Cluster analysis is performed on a large number of fault evolution sequences to screen out representative fault evolution sequences at the node as valid fault evolution sequences for the node.

4. The method for dynamic assessment and decision support of distribution network resilience according to claim 3, characterized in that, The calculation process of impact velocity: The fault evolution sequence is extracted, and the cumulative number of fault nodes from the start time to each time step is calculated. A two-dimensional rectangular coordinate system is constructed with the time step as the abscissa and the cumulative number of fault nodes at each time step as the ordinate. The cumulative number of fault nodes at each time step is then input into the coordinate system according to its corresponding time step to form several discrete points. The least squares method is used to perform linear fitting on the discrete points to obtain the linear equation. The slope of the linear equation is extracted as the fault accumulation rate, which represents the average number of newly added fault nodes per time step within the entire observation time window, i.e., the impact speed of the transport route under the threat scenario.

5. The method for dynamic assessment and decision support of distribution network resilience according to claim 4, characterized in that, The effective fault evolution sequence is determined based on the screening value of each cluster: A raw two-dimensional vector is constructed from the impact velocity and impact magnitude of each fault evolution sequence. Z-score is used to standardize all raw two-dimensional vectors to obtain standardized two-dimensional vectors. The similarity between the two-dimensional vectors corresponding to any two fault evolution sequences is calculated using a Gaussian kernel function. A Laplacian matrix is ​​constructed based on the similarity, and K-means clustering is performed to determine the optimal number of clusters, m. All fault evolution sequences are divided into m clusters, each representing a common fault impact pattern. After clustering, the number of fault evolution sequences in each cluster is counted, and this number is divided by the reference number of fault evolution sequences to obtain the cluster proportion, denoted as g1. The impact magnitudes of each fault evolution sequence are summed to obtain the total impact, and this total impact is divided by the impact reference to obtain the impact proportion, denoted as g2. The geometric mean formula is then used. The selection value G is calculated. A preset screening threshold is set. When the screening value is greater than the screening threshold, the cluster is recorded as a representative cluster. The number of representative clusters is counted. If the number of representative clusters is zero, the cluster with the largest screening value is selected as the representative cluster. The fault evolution sequence corresponding to the two-dimensional vector closest to the centroid of the cluster in each representative cluster is selected as the effective fault evolution sequence. Thus, the effective fault evolution sequence of each transport route under the threat scenario is obtained.

6. The method for dynamic assessment and decision support of distribution network resilience according to claim 5, characterized in that, Resilience metrics include robustness depth, resilience loss area, recovery time, and recovery speed. Robustness depth refers to the lowest trough of the resilience curve, i.e., the minimum value load retention rate. Resilience loss area refers to the area enclosed by the curve and the 100% horizontal line. Recovery time refers to the time required for the curve to recover from the lowest point to the basic retention rate. Recovery speed is calculated by subtracting robustness depth from the basic retention rate and then dividing by the critical recovery time; it is used to quantify the efficiency of recovery actions.

7. The method for dynamic assessment and decision support of distribution network resilience according to claim 6, characterized in that, Determine the scenario type for the fault evolution sequence: The impact velocity and impact amount of the impact feature vector of the effective fault evolution sequence are assigned a weight respectively. Then, the impact velocity and impact amount are linearly weighted and fused to obtain the impact index of the effective fault evolution sequence. An impact threshold and a robustness threshold are preset. If the impact index is greater than or equal to the impact threshold and the robustness depth is less than or equal to the robustness threshold, the effective fault evolution sequence is classified into the rapid paralysis type of fault. If the impact index is greater than or equal to the impact threshold and the robustness depth is greater than the robustness threshold, then the effective fault evolution sequence will be classified into the slow erosion fault type. If the impact index is less than the impact threshold and the robustness depth is less than or equal to the robustness threshold, the effective failure evolution sequence is classified as a critical hub disabling failure type; if the impact index is less than the impact threshold and the robustness depth is greater than the robustness threshold, the effective failure evolution sequence is classified as a general failure type. The effective fault evolution sequences corresponding to each transport route are sorted in chronological order according to the impact index, and the fault types and resilience assessment indicators corresponding to the effective fault evolution sequences are used as decision support information packages.

Images