A power distribution network resilience improvement method based on a data fusion algorithm

Abstract

The application belongs to the technical field of power distribution network, and provides a power distribution network resilience improvement method based on a data fusion algorithm, which comprises the following steps: collecting and fusing multi-dimensional data of the power distribution network, identifying regional-level resilience blocks with similar resilience characteristics by using a clustering algorithm, combining complex network theory and fault simulation to identify system-level resilience key nodes that have the greatest impact on the overall survivability of the system, and forming a resilience region and key node set; for the resilience region and key node set, evaluating node-level, regional-level and system-level resilience index values corresponding to different distributed resource space configuration schemes under a preset extreme scenario; and calculating resilience increment through comparison. An optimization model that fuses verified causal relationships and multi-objective coordination is established, and a scientific resource allocation strategy that guarantees the maximization of the overall resilience of the system and takes into account the balance of resilience between regions is output, so as to realize accurate, reliable and collaborative improvement of the resilience of the power distribution network.

Description

Technical Field

[0001] This invention belongs to the field of power distribution network technology, specifically a method for improving the resilience of power distribution networks based on data fusion algorithms. Background Technology

[0002] Distribution network resilience refers to the ability of a power grid to prevent, resist, recover quickly, and adapt after being subjected to extreme disturbances (such as extreme weather or cyberattacks). It emphasizes that the system must not only not be knocked down (reliability), but also be able to quickly get back up after being knocked down, ensuring continuous power supply and maintaining critical functions to the greatest extent possible. With the widespread access of high proportion of distributed power sources and electric vehicles, the form and operation paradigm of distribution networks have undergone fundamental changes: distributed resources bring considerable flexibility and local support potential, but the strong randomness and spatiotemporal coupling characteristics of the source-load sides also cause the complexity of system operation to increase dramatically. Therefore, its resilience characterization presents new characteristics of cross-scale and nonlinearity.

[0003] However, traditional planning and analysis methods are mostly based on deterministic static models and single data dimensions, making it difficult to characterize the aforementioned dynamic evolution process. Their inherent logic often implicitly assumes that local enhancements will necessarily linearly improve the overall system. This can lead to the resilience paradox in practice, where local optima may harm the overall system. For example, excessively concentrated resource deployment at a few nodes may create strong resilience islands, but it may also sever support paths to vulnerable areas due to fault isolation, thus weakening the overall collaborative recovery capability of the system. A deeper contradiction lies in the dilemma of improving resilience: regional resilience requires resource dispersion to achieve balance and avoid systemic weaknesses; while system-level resilience requires resources to be tilted towards topological and functionally critical nodes to curb cascading failures. Existing methods lack a quantitative understanding of the nonlinear mapping relationship between multi-level resilience indicators. When faced with this contradiction, they often fall into the dilemma of empirical trade-offs or single-objective optimization, thus failing to cope with real extreme risks.

[0004] Therefore, this invention provides a method for improving the resilience of distribution networks based on data fusion algorithms. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0006] The technical solution adopted by this invention to solve its technical problem is: a method for improving the resilience of a distribution network based on a data fusion algorithm, comprising:

[0007] Collect and integrate multi-dimensional data of the distribution network, use clustering algorithms to identify regional resilient blocks with similar resilience characteristics, and combine complex network theory and fault simulation to identify system-level resilient key nodes that have the greatest impact on the overall survivability of the system, forming a set of resilient regions and key nodes.

[0008] For resilient regions and key node sets, evaluate the node-level, region-level, and system-level resilience index values ​​corresponding to different distributed resource space configuration schemes under preset extreme scenarios; calculate the resilience increment by comparison and identify spurious candidate hotspots that may be affected by hidden variables;

[0009] Based on resilience index values, a comparative analysis is conducted on the competition or conflict in space and capacity between the resources required to improve regional resilience balance and the resources required to strengthen key nodes of system resilience, to determine whether there are contradictory phenomena; and causal diagnosis is performed on pseudo-correlated candidate hotspots. Through hierarchical comparison and counterfactual simulation, real correlations and pseudo-correlations are distinguished, and a verified resource-resilience mapping relationship table is output.

[0010] If a contradiction exists, a multi-objective optimization model that integrates regional equilibrium requirements is constructed with the goal of maximizing system-level resilience. The solution outputs the optimal resource allocation strategy that coordinates regional equilibrium and critical protection.

[0011] The beneficial effects of this invention are as follows:

[0012] This invention addresses two core contradictions in improving the resilience of power distribution networks under conditions of high-proportion distributed power sources and electric vehicle access: the conflict between local optimal damage and global and regional equilibrium and key protection objectives. It is driven by deep fusion of multi-level data, which integrates multi-source data such as power grid topology, operation, geography and transportation to identify resilient blocks and key system nodes.

[0013] This invention introduces incremental analysis and a pseudo-correlation hotspot screening mechanism into quantitative assessment, and then uses causal diagnosis through hierarchical comparison and counterfactual simulation to uncover the real nonlinear mapping relationship between resources and resilience, avoiding cognitive biases caused by hidden variables in traditional analysis.

[0014] This invention establishes an optimization model that integrates verified causal relationships and multi-objective collaboration, and outputs a scientific resource allocation strategy that maximizes the overall resilience of the system while taking into account the balance of resilience between regions, thereby achieving precise, reliable and collaborative improvement of the resilience of the distribution network. Attached Figure Description

[0015] The invention will now be further described with reference to the accompanying drawings.

[0016] Figure 1 This is a flowchart of the steps in Embodiment 1 of the present invention;

[0017] Figure 2 This is a system module architecture diagram in Embodiment 2 of the present invention. Detailed Implementation

[0018] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0019] Example 1:

[0020] One of the core inventive points of this invention is that, addressing the shortcomings of current technologies, a data fusion algorithm-based method for improving the resilience of distribution networks is proposed. This method constructs a logical closed loop of "identification-evaluation-diagnosis-optimization" to resolve two core contradictions in improving the resilience of distribution networks under conditions of high-proportion distributed power sources and electric vehicle access: the conflict between local optimal damage to the global and regional equilibrium and the conflict with key protection objectives. Driven by multi-level deep data fusion, it integrates multi-source data such as power grid topology, operation, geography, and transportation to identify resilient blocks and key system nodes. In the quantitative evaluation, incremental analysis and a pseudo-correlation hotspot screening mechanism are introduced. Then, through hierarchical comparison and counterfactual simulation causal diagnosis, the true nonlinear mapping relationship between resources and resilience is revealed, avoiding cognitive biases caused by hidden variables in traditional analysis. Finally, an optimization model integrating verified causal relationships and multi-objective collaboration is established to output a scientific resource allocation strategy that maximizes overall system resilience while also considering regional resilience equilibrium, thereby achieving precise, reliable, and coordinated improvement of distribution network resilience.

[0021] Please see Figure 1 As shown in the figure, the method for improving the resilience of a distribution network based on a data fusion algorithm according to an embodiment of the present invention includes the following steps:

[0022] Step 1: Collect and integrate multi-dimensional data of the distribution network, use clustering algorithms to identify regional resilient blocks with similar resilience characteristics, and combine complex network theory and fault simulation to identify system-level resilient key nodes that have the greatest impact on the overall survivability of the system, forming a set of resilient regions and key nodes;

[0023] Specifically, multi-dimensional data from the power distribution network are collected and integrated to construct a unified analysis dataset, including:

[0024] Power grid structure and operation data: including distribution network topology connections, line parameters, transformer capacity, protection device configuration, historical load curves, distributed generation (DG) installed capacity and historical output data;

[0025] Geographic and meteorological data: Acquire node geographical locations, historical meteorological disaster records (such as typhoon paths and rainstorm areas), and topographic data;

[0026] Traffic and Electric Vehicle (EV) Data: Integrating traffic network information, EV charging station / pile distribution, and statistical data on typical travel patterns and charging behavior;

[0027] Preprocessing: The above data is spatiotemporally aligned, missing values ​​are handled, outliers are detected, and normalization is performed to construct a unified spatiotemporal scale power grid-geography-transportation fusion database;

[0028] Specifically, the process of identifying regional-level resilient blocks is as follows:

[0029] A multidimensional resilience feature vector is calculated for each grid node, where the resilience feature dimensions include, but are not limited to:

[0030] Structural vulnerability: the topological importance of nodes in the network (e.g., degree centrality) and power supply radius;

[0031] Historical reliability: average outage frequency and average outage duration based on historical data;

[0032] Risk exposure: meteorological disaster risk level and equipment environmental risk score based on geographic data;

[0033] Resource support potential: Theoretical maximum support capacity of local DG and energy storage, and estimated dispatchable potential of EVs;

[0034] An unsupervised clustering algorithm is used to cluster the resilience feature vectors of power grid nodes, and nodes with similar resilience feature vectors and spatial proximity are divided into a regional resilience block.

[0035] Optionally, the unsupervised clustering algorithm can be K-means, DBSCAN, or hierarchical clustering. In this embodiment, K-means clustering algorithm is used as an example to divide the region-level resilient blocks.

[0036] Furthermore, the multidimensional resilience feature vectors of all power grid nodes are standardized to eliminate the impact of differences in feature dimensions and value ranges on the clustering results.

[0037] The optimal number of clusters K is determined by using the elbow rule or by combining prior knowledge and domain experience. The value of K usually corresponds to the total number of resilient blocks in the region.

[0038] Perform K-means clustering:

[0039] From the standardized feature vector set, K vectors are randomly selected as the initial cluster centers (centroids).

[0040] a. Allocation phase: Calculate the Euclidean distance (or other distance metric) from the feature vector of each power grid node to the K centroids, and assign the node to the cluster of the nearest centroid;

[0041] b. Update phase: Recalculate the mean of the feature vectors of all power grid nodes in each cluster, and use the mean vector as the new centroid of the cluster;

[0042] c. Repeat steps a and b until the position of all centroids changes less than a preset threshold between two iterations, or the maximum number of iterations is reached, at which point the algorithm converges;

[0043] After the algorithm converges, each node is assigned to a unique cluster. A set of nodes that belong to the same cluster and are spatially continuous or adjacent is defined as a regional resilient block.

[0044] Specifically, the process of identifying critical nodes for system-level resilience is as follows:

[0045] Complex network topology analysis involves abstracting the distribution network into a weighted and directed graph and calculating the topological centrality indices of each node, with a focus on electrical betweenness centrality and proximity centrality.

[0046] Furthermore, electrical betweenness centrality is used to measure the degree of hub status of a node in the power flow path;

[0047] Proximity centrality is used to reflect the electrical distance between a node and all other nodes in the network;

[0048] Multi-scenario fault simulation and impact assessment involves constructing and executing preset extreme disturbance scenario simulations to quantitatively assess the impact of failures at various power grid nodes on the overall system resilience.

[0049] The extreme disturbance scenario is constructed, defined, and three representative sets of fault scenarios are generated:

[0050] Random device failure scenarios: Simulate N-1 and Nk random failures of critical equipment such as lines and transformers, covering most vulnerable parts of the network;

[0051] Severe weather event scenarios: Based on historical meteorological data, simulate spatially correlated disasters such as typhoons, hail, and floods. In these scenarios, the location and probability of failure are strongly correlated with the geographical risk exposure of nodes, and failures usually occur in a regional or clustered manner.

[0052] Targeted network attack scenarios: Simulate malicious attacks against power grid nodes with high topological centrality or those connecting critical loads. Such attacks are designed to cause the widest possible system paralysis at minimal cost.

[0053] For each preset extreme scenario, a time-series simulation tool (such as a simulation platform built on OpenDSS, MATLAB / Simulink, etc.) is used for simulation. The simulation process considers the dynamic response of distributed generation (DG) and electric vehicle (EV), the coordinated action of protection devices, and possible network reconfiguration strategies.

[0054] Record and collect dynamic response data of the system throughout the simulation process, including but not limited to: voltage and frequency curves of each power grid node; load rate changes of lines and transformers; power outage and restoration time series of loads; and action records of protection devices.

[0055] Based on simulation output data, the severity index of the impact of each power grid node (or combination of nodes) under various fault scenarios is calculated, mainly including:

[0056] Scope of load loss impact: The total amount of power loss directly and indirectly caused by the failure of a power grid node (or the primary point of failure), as well as the proportion of power loss of critical loads (such as hospitals and emergency centers);

[0057] Recovery path disruption level: assesses the system's ability to reconfigure the network and restore power supply to non-faulty areas via tie switches after a grid node failure. Quantitative indicators can be the decrease in the proportion of recoverable power loss load or the increase in the number of critical switch operations required to restore power supply.

[0058] Risk of cascading failures: Statistical analysis of the probability and average impact range (number of nodes or load) of cascading failure events triggered by the initial failure of a power grid node, such as subsequent protection malfunctions and power flow exceeding limits leading to equipment overload.

[0059] By integrating the topological centrality index derived from complex network topology analysis with the impact severity index derived from multi-scenario fault simulation, a multi-criteria decision-making method (such as TOPSIS) is adopted to rank and screen power grid nodes.

[0060] Furthermore, with all network nodes as rows and topological centrality indicators and impact severity indicators (load loss, recovery interruption, cascading risk) as columns, an initial evaluation matrix is ​​formed;

[0061] The indicators are normalized, and the weights of each indicator are determined by a combination of subjective and objective methods (such as AHP-entropy weight combination weighting) to balance the importance of theory and the objectivity of data.

[0062] Based on the TOPSIS method, the distance between each node and the optimal ideal solution (where all indicators are optimal) and the worst ideal solution is calculated to obtain the relative proximity score that represents the overall criticality.

[0063] All nodes are sorted in descending order based on their proximity scores. By setting a threshold (such as the top 10% or a score > 0.8), the nodes that simultaneously possess high topological hubness and high fault impact severity, i.e., the node group with the highest score, are identified as system-level resilience critical nodes and added to the set.

[0064] The list of regional resilience blocks is integrated with the list of system-level resilience critical nodes to form a set of resilience regions and critical nodes.

[0065] Step 2: For the set of resilient regions and key nodes, evaluate the node-level, region-level, and system-level resilience index values ​​corresponding to different distributed resource space configuration schemes under preset extreme scenarios;

[0066] In the assessment, resilience increments are calculated by comparing the presence or absence of variables, and spurious candidate hotspots that may be affected by hidden variables are identified.

[0067] Specifically, for the set of resilient regions and key nodes, node-level resilience indicators, region-level resilience indicators, and system-level resilience indicators are defined respectively.

[0068] Furthermore, the node-level resilience indicators are: self-sustaining time and voltage support ratio;

[0069] After a failure occurs, the maximum duration for which a node can continuously supply power to critical loads using local distributed resources (DG, EV) is called the self-sustaining time.

[0070] The percentage of time during which the node voltage remains within ±10% of the rated voltage during a fault is the voltage support rate.

[0071] Regional resilience indicators include: island survival rate and internal mutual aid recovery rate.

[0072] The proportion of core load that can be met by distributed resources within the region after a fault causes the region to be disconnected from the main network is the island survival rate.

[0073] The average load power restored per minute by the region through internal network reconstruction and resource allocation is the internal mutual recovery rate.

[0074] System-level resilience metrics include: total load loss and critical load recovery time.

[0075] The total amount of power lost across the entire network from the occurrence of the fault until the system is fully restored is the total load loss.

[0076] The time required to restore power to more than 95% of the critical loads (such as hospitals and emergency command centers) within the system is called the critical load recovery time.

[0077] Based on resilient regions and key node sets, a set of representative distributed resource (DG, EV schedulable capacity) spatial configuration schemes are generated as evaluation objects:

[0078] Option A (Critical Node Enhancement): Deploy more than 60% of the new resource capacity to system-level resilient critical nodes;

[0079] Option B (Weak Area Reinforcement): Prioritize deploying more than 60% of the new resource capacity in the regional resilient blocks with low resilience characteristic values ​​identified in Step 1;

[0080] Among them, regional-level toughness blocks with low toughness eigenvalues ​​refer to those where, in the cluster analysis of the aforementioned steps, the comprehensive score (such as the average score of each dimension) of the multidimensional toughness eigenvectors of the nodes within the regional-level toughness block is in the bottom 20% of all blocks.

[0081] Option C (Uniformly Distributed): New resources are allocated according to the proportion of peak load in each region;

[0082] Baseline approach: Maintain the existing resource configuration unchanged as the baseline for all comparative analyses;

[0083] For each resource configuration scheme, combined with the set of extreme disturbance scenarios preset in step one, a time-series simulation tool is used to simulate the data, and node voltage, load status, switching action sequence and timestamp data are collected throughout the simulation process.

[0084] Based on simulation data, calculate the resilience index values ​​of each node, regional resilience block, and system under each scheme-scenario combination;

[0085] Calculate the increments of resilience indicators at each level (node ​​level, region level, system level) of schemes A, B, and C relative to the baseline scheme;

[0086] Based on the incremental calculation results of resilience indicators at all levels, spurious hotspots are identified, including:

[0087] Screening for high investment and low efficiency: For each resource allocation scheme, identify and record power grid nodes that are allocated resource capacity higher than the average level of the scheme, but whose corresponding node-level, regional-level resilience block-level, or system-level resilience increments are lower than the average increment level of the scheme, and list them as pseudo-related candidate hotspots;

[0088] Scenario-dependent screening: For each power grid node or regional resilient block, calculate the resilience increment sequence under different extreme disturbance scenarios, identify and record objects whose resilience increment values ​​fluctuate (measured by the coefficient of variation) between different scenarios and are higher than the average level of the entire network node or region, and list them as pseudo-related candidate hotspots.

[0089] The above screening results were compiled into a list of pseudo-relevant candidate hotspots.

[0090] Step 3: Based on the resilience index values, compare and analyze whether there is competition or conflict in space and capacity between the resources required to improve the regional resilience balance and the resources required to strengthen the key nodes of system resilience, and determine whether there are contradictory phenomena.

[0091] Furthermore, causal diagnosis is performed on the identified spurious correlation candidate hotspots. Through comparative analysis and counterfactual simulation, the true nonlinear coupling relationship between resource and resilience indicators is distinguished from the spurious correlation caused by a third variable, thereby establishing a reliable resource-resilience mapping relationship.

[0092] Specifically, based on the simulation results of Scheme B, the spatial distribution of incremental resources (DG and EV capacity) required to improve the resilience index values ​​(especially island survival rate and internal mutual aid recovery rate) of each regional resilient block to a predetermined level is identified, forming a resource demand map for improving regional balance.

[0093] Based on the simulation results of Scheme A, the spatial distribution of incremental resources required to enable all system-level resilience key nodes to achieve the predetermined fault support capability (such as meeting the standard of node voltage support rate after fault) is identified, forming a key node reinforcement resource requirement map.

[0094] Comparing the two resource demand maps above, identify nodes or regions that are physically overlapping or adjacent (i.e., two nodes or regions belong to the same regional resilience block, or belong to two regional resilience blocks directly connected by a link line) and whose resource demand directions (i.e., resources should be concentrated to strengthen key nodes, or should be dispersed to balance regional resilience) are opposite.

[0095] Under the constraint of the total new resource budget of the system, if satisfying the resource needs of one party will result in the resource needs of the other party not being fully satisfied, then it is determined that there is a resource contradiction between regional balance and critical node protection, that is, there is a contradiction phenomenon. Record the specific location (node ​​or region, which is the resource contradiction region) where the contradiction occurs and the estimated value of the conflicting resource capacity.

[0096] The objects in the list of spurious correlation candidate hotspots are analyzed in depth to distinguish between true causal relationships and spurious correlations. The process is as follows:

[0097] Select a potential third variable that may cause spurious correlations, such as the node’s geographic risk exposure (from step one) or the inherent resilience characteristics of the regional resilient block to which it belongs.

[0098] All power grid nodes are stratified according to the value of the selected third variable (e.g., divided into high, medium, and low risk layers). Within each layer, the relationship between resource input and resilience increment of hotspot objects is re-analyzed.

[0099] If the hot topic exhibits characteristics of high investment and low efficiency or scenario dependence in each layer, it strengthens the evidence of the weak correlation between resource investment and resilience improvement. If it only shows anomalies in a specific layer (such as the high-risk layer) and normal performance in other layers, it indicates that its anomalies may be mainly caused by the third variable (such as high geographical risk), which is a spurious correlation.

[0100] For any pseudo-correlated candidate hotspot in the pseudo-correlated candidate hotspot list, design a counterfactual simulation scenario: based on the baseline scheme, the resources originally planned to be invested in the power grid node N are virtually reconfigured to a control power grid node M that is similar to it in topological location or load characteristics but has not been marked as a hotspot;

[0101] Keeping other conditions unchanged, rerun the simulation under the same set of extreme disturbance scenarios and calculate the increments of each level of resilience index under the new configuration;

[0102] If a node is a pseudo-correlated candidate hotspot, and the increase in system-level resilience indicators after resource reconfiguration is better than the original solution, then the node is confirmed to be a pseudo-correlated hotspot.

[0103] If the incremental system-level resilience index does not improve significantly or even decreases after resource reallocation, it indicates that the resource input of this node has a real and significant causal effect on system resilience, and it should be listed as a high-priority efficiency enhancement target.

[0104] Based on the above analysis, a structured, verified resource-resilience mapping table is output.

[0105] The verified resource-resilience mapping table integrates three key information categories: conflicting regions, efficiency-enhancing nodes, and inefficient hotspots. First, it clearly identifies the spatial locations where there is direct resource competition between enhancing regional resilience and strengthening key nodes (i.e., resource conflicting regions). Second, it explicitly lists nodes and regions that have been verified by counterfactual simulations to have a real and significant causal effect on improving system resilience, as high-priority efficiency-enhancing targets. Third, it identifies pseudo-related hotspots caused by interference from third variables such as geographical risks or improper configuration, and lists them as targets for avoiding or restricting resource investment.

[0106] Step 4: If contradictions exist, a multi-objective collaborative optimization model that integrates regional balance requirements is established with the core objective of maximizing system-level resilience. Nodes and regions identified as having real high causal effects are set as high-priority optimization objects. Resource constraint upper limits are applied to hotspots identified as pseudo-correlated. The model is solved using an intelligent optimization algorithm, and the optimal distributed resource space configuration strategy that can coordinate regional resilience balance and critical node protection under total resource constraints is output.

[0107] Specifically, based on the verified resource-resilience mapping table, a multi-objective collaborative optimization model is constructed with the core objective of improving system-level resilience while also taking into account regional balance.

[0108] Furthermore, the multi-objective collaborative optimization model is constructed from three parts: decision variables, core objective function, and constraints.

[0109] Decision variables: Define the planned new distributed generation (DG) capacity and dispatchable electric vehicle (EV) capacity values ​​for each grid node;

[0110] Core objective function:

[0111] The main objective (maximizing system-level resilience) is to maximize the expected improvement in system-level resilience indicators. Specifically, it is to maximize the expected reduction in total system load loss or the expected reduction in critical load recovery time under all preset extreme disturbance scenarios.

[0112] Subordinate objective (promoting balanced regional resilience): The subordinate objective is to minimize the difference in the final resilience level of each regional-level resilience block (characterized by the island survival rate of the block), which can be achieved by minimizing the Gini coefficient or variance of the resilience index between blocks.

[0113] Constraints:

[0114] Total resource constraint: The sum of the newly added resource capacity of all power grid nodes shall not exceed the total budgeted capacity for the planning period;

[0115] Node-level resource upper and lower limits constraints: The new resource capacity of each power grid node is limited by physical space, power grid acceptance capacity, etc.

[0116] High-priority efficiency enhancement objects promote constraints: For nodes or regions listed as high-priority efficiency enhancement objects in the mapping relationship table, set a lower limit constraint on the amount of resources allocated to ensure that verified effective investments are guaranteed;

[0117] Inefficient hotspot avoidance constraint: For power grid nodes identified as pseudo-related hotspots in the mapping table, set an upper limit constraint on the amount of resources allocated (usually set to a low value or zero) to avoid ineffective resource allocation;

[0118] Resource conflict region coordination constraints: For resource conflict regions marked in the mapping table, additional coordination constraints are introduced; for example, if a region-level resilience block (which needs balanced improvement) has resource conflicts with a system-level resilience critical node (which needs focused reinforcement), constraints can be added: ;

[0119] The coordination coefficient is set according to the severity of the conflict, aiming to guide the rational allocation of resources between the conflicting parties.

[0120] Power flow safety and voltage quality constraints: Ensure that under typical operating conditions, resource allocation schemes do not cause line overload or node voltage exceedances;

[0121] The above multi-objective optimization model is solved using an intelligent optimization algorithm;

[0122] Optionally, algorithms suitable for multi-objective, high-dimensional, constrained optimization problems can be selected, such as the Non-Dominated Sorting Genetic Algorithm with Elite Strategy (NSGA-II) or the Multi-Objective Particle Swarm Optimization Algorithm.

[0123] The multi-objective collaborative optimization model is encoded into a form that the algorithm can process. A population of random resource allocation schemes is initialized. During the iteration process, the algorithm generates new schemes through operations such as selection, crossover, and mutation.

[0124] For each new scheme (individual), the evaluation process established in step two is invoked (i.e., combining extreme scenarios to perform time-series simulations and calculate multi-level resilience indices) to evaluate the fitness (i.e., the objective function value).

[0125] Based on fitness values, non-dominated ranking, crowding calculations, and other mechanisms, superior individuals are selected and continuously evolved;

[0126] Finally, it converges to a set of Pareto optimal solutions. Each solution in this set represents a resource allocation scheme that achieves a specific optimal trade-off between improving system-level resilience and regional equilibrium under given resources.

[0127] The Pareto optimal solution set obtained by solving is visualized and presented to planners. Typically, a two-dimensional frontier diagram of system resilience improvement value - regional equilibrium index can be drawn to intuitively show the distribution of schemes under different focuses.

[0128] Planners can select a final solution from the frontier solution set based on their actual management preferences (such as a greater emphasis on overall risk control or a greater focus on fairness).

[0129] As another preferred approach in this embodiment: a compromise solution is automatically recommended from the Pareto solution using a decision-making method (such as fuzzy satisfaction method);

[0130] The selected final solution is described in detail as the optimal distributed resource space configuration strategy, which explicitly lists:

[0131] Recommended DG type and capacity, and dispatchable power of EV charging facilities for each planning node;

[0132] The expected improvement in system-level resilience;

[0133] The improvement in resilience level and balance of each regional resilience block under this strategy;

[0134] A performance overview of the strategy through simulation verification (such as key performance in major extreme scenarios).

[0135] This embodiment uses data-driven analysis to reveal and quantify the contradictions and real benefits in resource allocation, thereby guiding the generation of a resource allocation scheme that achieves a balance between improving overall resilience, balancing regional capabilities, and avoiding ineffective investments, ultimately outputting the optimal distributed resource space allocation strategy.

[0136] Example 2:

[0137] Based on the same inventive concept as the data fusion algorithm-based distribution network resilience enhancement method in the foregoing embodiments, such as... Figure 2 As shown, this application provides a distribution network resilience enhancement system based on a data fusion algorithm, wherein the system specifically includes:

[0138] Fusion and extraction module: Collects and fuses multi-dimensional data of the distribution network, uses clustering algorithms to identify regional resilient blocks with similar resilience characteristics, and combines complex network theory and fault simulation to identify system-level resilient key nodes that have the greatest impact on the overall survivability of the system, forming a set of resilient regions and key nodes;

[0139] Evaluation and Comparison Module: For resilient regions and key node sets, evaluate the node-level, region-level, and system-level resilience index values ​​corresponding to different distributed resource space configuration schemes under preset extreme scenarios;

[0140] In the assessment, resilience increments are calculated by comparing the presence or absence of variables, and spurious candidate hotspots that may be affected by hidden variables are identified.

[0141] Judgment and Diagnosis Module: Based on resilience index values, compare and analyze the resources required to improve the regional resilience balance and the resources required to strengthen the key nodes of system resilience, and determine whether there is competition or conflict in space and capacity, and whether there are contradictory phenomena.

[0142] Furthermore, causal diagnosis is performed on the identified spurious correlation candidate hotspots. Through comparative analysis and counterfactual simulation, the true nonlinear coupling relationship between resource and resilience indicators is distinguished from the spurious correlation caused by a third variable, thereby establishing a reliable resource-resilience mapping relationship.

[0143] Multi-objective collaborative optimization module: If contradictory phenomena exist, the core objective is to maximize system-level resilience. A multi-objective collaborative optimization model that integrates regional equilibrium requirements is established. Nodes and regions identified as having real high causal effects are set as high-priority optimization objects. Resource constraint upper limits are applied to hotspots identified as pseudo-correlated. Intelligent optimization algorithms are used to solve the model and output the optimal distributed resource space configuration strategy that can coordinate regional resilience equilibrium and critical node protection under total resource constraints.

[0144] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for improving the resilience of a distribution network based on a data fusion algorithm, characterized in that: include: Collect and integrate multi-dimensional data of the distribution network, use clustering algorithms to identify regional resilient blocks with similar resilience characteristics, and combine complex network theory and fault simulation to identify system-level resilient key nodes that have the greatest impact on the overall survivability of the system, forming a set of resilient regions and key nodes. For resilient regions and key node sets, evaluate the node-level, region-level, and system-level resilience index values ​​corresponding to different distributed resource space configuration schemes under preset extreme scenarios; calculate the resilience increment by comparison with and without, and identify spurious correlation candidate hotspots affected by hidden variables; Based on resilience index values, a comparative analysis is conducted to determine whether there are any contradictions by comparing or analyzing the competition or conflict in space and capacity between the resources required to improve the regional resilience balance and the resources required to strengthen the key nodes of system resilience. Furthermore, causal diagnosis is performed on spurious candidate hotspots. Through hierarchical comparison and counterfactual simulation, real correlations and spurious correlations are distinguished, and a verified resource-resilience mapping relationship table is output. The verified resource-resilience mapping table includes: Resource conflict areas refer to spatial location information where there is direct resource competition between key nodes that enhance regional resilience balance and strengthen system resilience. High-priority efficiency enhancement targets are nodes and regions that have been verified by counterfactual simulations to have a real and significant positive causal effect on the improvement of system resilience through resource investment. False correlations or inefficient hotspots are nodes and regions where resource investment and resilience improvement are weakly or ineffectively correlated due to interference from a third variable or improper original configuration. If a contradiction exists, a multi-objective optimization model that integrates regional equilibrium requirements is constructed with the goal of maximizing system-level resilience, and the optimal resource allocation strategy that coordinates regional equilibrium and critical protection is output. The process of constructing a multi-objective optimization model that integrates regional equilibrium requirements is as follows: Define decision variables: The newly planned distributed power generation capacity and the dispatchable capacity of electric vehicles at each power grid node are the decision variables; Set the core objective function: the primary objective is to maximize the expected improvement in the system-level resilience index, and the secondary objective is to minimize the difference in the final resilience level of each regional resilience block; The constraints are designed to include total resource capacity constraints, upper and lower limit constraints on resources at each node, incentive constraints for high-priority efficiency enhancement objects based on the verified resource-resilience mapping table, constraints to avoid pseudo-correlation or inefficient hotspots, coordination constraints for resource conflict areas, and power flow and voltage constraints to ensure the safe operation of the power grid. The process of finding the optimal resource allocation strategy for coordinating regional equilibrium and critical protection is as follows: The model is solved using an intelligent optimization algorithm, and after convergence, a set of Pareto optimal solutions representing different trade-offs is obtained. The Pareto optimal solution set is visualized using system-level resilience enhancement value and regional resilience balance index as dimensions. From the Pareto optimal solution set, a final solution is selected according to the preset decision rules, and it is expressed as the optimal distributed resource space configuration strategy that includes the specific resource types, capacity and expected resilience benefits of each node.

2. The method for improving the resilience of a distribution network based on a data fusion algorithm according to claim 1, characterized in that: The process of identifying region-level toughness blocks with similar toughness characteristics using clustering algorithms is as follows: For each power grid node, a multidimensional resilience feature vector is calculated, with feature dimensions including structural vulnerability, historical reliability, risk exposure, and resource support potential. Clustering is performed on the standardized feature vectors, and a set of nodes that belong to the same cluster and are spatially continuous or adjacent is defined as a regional resilient block.

3. The method for improving the resilience of a distribution network based on a data fusion algorithm according to claim 1, characterized in that: The process of identifying the system-level resilience critical nodes that have the greatest impact on the overall survivability of the system is as follows: Perform topology analysis on complex networks and calculate the topological centrality indices of each node, including electrical betweenness centrality and proximity centrality. Construct and execute a pre-defined extreme disturbance scenario simulation to quantitatively assess the severity of the impact of each node failure, including the scope of load loss, the degree of disruption to the recovery path, and the risk of triggering cascading failures. By integrating the topological centrality index with the aforementioned impact severity index, nodes are comprehensively ranked using the TOPSIS method. By setting a ranking ratio threshold, the nodes ranked higher are identified as critical nodes for system-level resilience.

4. The method for improving the resilience of a distribution network based on a data fusion algorithm according to claim 1, characterized in that: The process of evaluating the node-level, region-level, and system-level resilience indicators corresponding to different distributed resource space configuration schemes under preset extreme scenarios is as follows: Define a quantitative indicator system for resilience at the node, regional, and system levels; Based on resilient regions and key node sets, a variety of distributed resource space configuration schemes with different resource distribution tendencies are generated; The distributed resource spatial configuration schemes include: key node reinforcement, weak area reinforcement, and uniform distribution. For each of the aforementioned distributed resource space configuration schemes, a time-series simulation is performed in conjunction with a preset set of extreme scenarios; Based on the simulation output data, calculate the resilience index values ​​of each node, each region-level resilience block, and the entire system under each scheme-scenario combination. By comparing the resilience index values ​​of each scheme with the baseline scheme, the increments of the resilience index at each level are calculated.

5. The method for improving the resilience of a distribution network based on a data fusion algorithm according to claim 1, characterized in that: The process of obtaining pseudo-relevant candidate hotspots: Through screening for high investment and low efficiency, identify and record power grid nodes whose allocated resource capacity in the resource allocation scheme is higher than the average level of the scheme, but whose node-level, regional-level resilience block-level, or system-level resilience increments are lower than the average increment level of the scheme. By screening based on scenario dependence, identify and record power grid nodes or regional resilient blocks whose resilience increment values ​​fluctuate more than the average level of all network nodes or regions in different extreme disturbance scenarios. The results of high-investment, low-efficiency screening and scenario-dependent screening were compiled and summarized to form a list of pseudo-relevant candidate hotspots.

6. The method for improving the resilience of a distribution network based on a data fusion algorithm according to claim 4, characterized in that: The process for determining whether a contradiction exists is as follows: Based on simulation results of the weak area reinforcement scheme and the key node enhancement scheme, resource demand maps for improving regional balance and key node enhancement are generated respectively. By comparing two resource demand maps, identify nodes or regions that overlap or are adjacent in physical space and have opposite resource demand directions; Under the constraint of the total new resource budget of the system, determine whether satisfying the needs of one party will result in the inability to fully satisfy the needs of the other party. If so, it is determined that there is a resource conflict.

7. The method for improving the resilience of a distribution network based on a data fusion algorithm according to claim 5, characterized in that: The process of distinguishing between true correlation and spurious correlation is as follows: For objects in the pseudo-relevant candidate hotspot list; Stratified comparative analysis: Select potential third variables, stratify all nodes according to their values, and re-analyze the relationship between resource input and resilience increment of objects within each stratum to determine whether abnormal correlations are dominated by the corresponding variables; Counterfactual simulation verification: For the object to be analyzed, design a simulation scenario and virtually reconfigure the planned resources to a control node with similar topology or load characteristics but not marked as a hotspot; By comparing the incremental changes in system-level resilience indicators before and after resource reallocation, the true causal effect of resource investment on the corresponding objects can be determined, and then they can be distinguished as high-priority efficiency-enhancing objects or pseudo-related hotspots.

Images