A power distribution network cluster adjustment method and system and medium

Abstract

The application discloses a power distribution network cluster adjustment method and system and a medium, and belongs to the technical field of smart grids. The method comprises the following steps: clustering and dividing power distribution network nodes to obtain a plurality of static clusters according to acquired power distribution network topology data, load data and historical photovoltaic output data; when it is monitored that any static cluster meets a preset condition, determining the correlation degree of each node in the cluster with the cluster and adjacent clusters according to the topology data, and then determining the belonging ambiguity of the node; determining the deviation degree of each node deviating from the overall operation trend of the cluster according to meteorological data and historical photovoltaic output data; fusing the belonging ambiguity and the deviation degree to obtain a comprehensive belonging uncertainty, determining a node belonging adjustment scheme according to the comprehensive belonging uncertainty, and adjusting the cluster boundary. Through implementation of the application, the technical problem that a static hard boundary cannot adapt to the severe fluctuation of photovoltaic power can be solved, and the accuracy of power distribution network cluster division is improved.

Description

Technical Field

[0001] This application relates to the field of smart grid technology, and in particular to a method, system and medium for clustered adjustment of distribution networks. Background Technology

[0002] As the proportion of distributed photovoltaic (PV) power connected to distribution networks continues to increase, the inherent randomness, volatility, and spatiotemporal differences in its output lead to frequent problems such as voltage exceeding limits, backflow of power flow, and increased network losses. Cluster partitioning, as an important means to reduce the difficulty of managing high-proportion distributed resources and achieve hierarchical distributed local balance in distribution networks, directly affects the operational stability of the distribution network due to its quality.

[0003] Existing methods for dividing distribution network clusters typically rely on historical cross-sectional data to construct static models, often employing static indicators such as electrical distance and modularity combined with heuristic algorithms to complete the partitioning. This method's analysis is limited to operational characteristics at a fixed point in time, and the resulting partitioning only applies to steady-state conditions. Once the cluster boundaries are defined, they remain fixed and lack the ability to dynamically adjust to fluctuations in photovoltaic (PV) output. When PV output fluctuates drastically, fixed boundaries can easily lead to a situation where one adjacent cluster experiences severe power surplus while the other has idle adjustment margins, making it difficult to match the dynamic operational needs of the power grid. Therefore, improving the accuracy of distribution network cluster partitioning under the premise of distributed PV integration is a pressing technical problem that needs to be solved. Summary of the Invention

[0004] This application provides a method, system, and medium for cluster adjustment in distribution networks, which solves the problem that static hard-boundary clusters cannot adapt to severe photovoltaic fluctuations and improves the accuracy of distribution network cluster division.

[0005] This application provides a clustered adjustment method for a distribution network, including: Based on the acquired distribution network topology data, distribution network load data, and historical photovoltaic power output data, the distribution network nodes are clustered to obtain several static clusters; When any static cluster in the distribution network is detected to meet the preset conditions, the first degree of association between each node in the static cluster and the static cluster, and the second degree of association between each node in the static cluster and the adjacent static cluster are determined according to the distribution network topology data, the distribution network load data, and the historical photovoltaic output data; the affiliation ambiguity of each node in the static cluster is determined according to the first degree of association and the second degree of association. Obtain meteorological data of the power distribution network, and determine the degree of deviation of each node in the static cluster from the overall operating trend of the static cluster based on the historical photovoltaic output data and the meteorological data of the power distribution network; The attribution ambiguity and the deviation are fused to obtain the comprehensive attribution uncertainty of each node in the static cluster; Based on the comprehensive attribution uncertainty, a node attribution adjustment scheme is determined within the static cluster, and the cluster boundary of the distribution network is adjusted according to the node attribution adjustment scheme.

[0006] This application constructs several static clusters through clustering, forming the basic control framework of the distribution network. When any static cluster meets preset conditions, it directly calculates the correlation degree of each node relative to its own cluster and adjacent clusters, thereby determining the node's affiliation ambiguity and quantifying the degree to which the node is physically torn apart by the clusters on both sides. Simultaneously, it determines the deviation of the node from the overall operating trend of the cluster based on historical photovoltaic output data and meteorological data, quantifying the node's instability under dynamic fluctuations. The affiliation ambiguity and deviation are integrated into a comprehensive affiliation uncertainty. Based on this comprehensive index, a node affiliation adjustment scheme is determined, and the cluster boundary is adjusted, minimizing the search range and action cost of cluster adjustment. This achieves adaptive adjustment of the cluster boundary without shaking the entire static network framework, solving the fundamental defect that static hard boundaries cannot adapt to photovoltaic fluctuations and improving the adaptability of distribution network cluster division to photovoltaic fluctuations. Compared with existing technologies, this application identifies boundary nodes by integrating physical structure ambiguity and operational fluctuation deviation, and directly determines the adjustment scheme based on the comprehensive affiliation uncertainty, achieving precise response to photovoltaic fluctuations and on-demand adjustment of the cluster boundary.

[0007] Further, determining the first degree of association between each node in the static cluster and the static cluster, and the second degree of association between each node in the static cluster and adjacent static clusters, based on the distribution network topology data, the distribution network load data, and the historical photovoltaic output data, includes: Based on the distribution network topology data, the distribution network load data, and the historical photovoltaic power output data, a first electrical coupling similarity between distribution network nodes is determined. For each node in the static cluster, the sum of the second electrical coupling similarities between the current node and all other nodes in the static cluster is calculated to obtain the first degree of association; For each node in the static cluster, the sum of the third electrical coupling similarity between the current node and all other nodes in the adjacent static cluster is calculated to obtain the second degree of association.

[0008] This application determines the first electrical coupling similarity between distribution network nodes based on the aforementioned distribution network topology data, distribution network load data, and historical photovoltaic power output data, quantifying the tightness of the electrical connections between nodes. For each node within the static cluster, the sum of the second electrical coupling similarities between that node and all other nodes within the cluster is calculated to obtain the first association degree, reflecting the embedding depth of the node within its cluster. The sum of the third electrical coupling similarities between that node and all nodes in adjacent static clusters is calculated to obtain the second association degree, reflecting the traction force of the node on the adjacent clusters. By comparing the first and second association degrees, nodes located at electrical boundaries and subjected to bidirectional pulling from both sides of the clusters can be accurately located, providing a reliable comparative basis for the subsequent quantification of attribution ambiguity.

[0009] Further, determining the affiliation ambiguity of each node within the static cluster based on the first association degree and the second association degree includes: Calculate the difference between the first degree of association and the second degree of association, and determine a first intermediate value based on the difference; Calculate the sum of the first degree of association and the second degree of association to obtain the second intermediate value; The affiliation ambiguity of each node in the static cluster is determined based on the ratio of the first intermediate value to the second intermediate value.

[0010] This application determines a first intermediate value based on the absolute value of the difference between the first and second correlation degrees, quantifying the degree of imbalance in the traction forces of the nodes on both sides of the cluster; it determines a second intermediate value based on the sum of the first and second correlation degrees, serving as a normalization benchmark to reflect the total coupling strength experienced by the node; and it determines the node's affiliation ambiguity based on the ratio of the first and second intermediate values. By using a negative correlation where a ratio closer to 0 indicates a clearer affiliation and a ratio closer to 1 indicates a more ambiguous affiliation, the bidirectional tearing degree of the node on the electrical boundary is transformed into a dimensionless index between 0 and 1, providing a quantitative basis for the subsequent accurate identification of nodes in a boundary oscillation state.

[0011] Furthermore, determining the deviation of each node within the static cluster from the overall operating trend of the static cluster based on the historical photovoltaic output data and the distribution network meteorological data includes: The historical photovoltaic output data and the distribution network meteorological data are input into a preset prediction model. Based on the time series data of light intensity and temperature in the distribution network meteorological data, several sets of predicted photovoltaic output data sequences are generated. Based on the predicted photovoltaic power output data sequence, determine the probability value corresponding to each set of the predicted photovoltaic power output data sequence; For each set of photovoltaic power output data sequences, the correlation coefficient between the net power time series data of each node in the static cluster and the overall net power time series data of the static cluster is calculated; wherein, the net power time series data of the node is determined based on the photovoltaic power output data and node load of the node; the overall net power time series data of the static cluster is determined based on the net power of all nodes in the static cluster. Based on the correlation coefficient and the probability value, the deviation of each node in the static cluster from the overall operating trend of the static cluster is calculated.

[0012] This application generates several sets of predicted photovoltaic (PV) output data sequences based on the historical PV output data and the distribution network meteorological data, and determines the probability value of each sequence. It uses the illumination and temperature information in the meteorological data to simulate PV output changes under different weather conditions, quantifying the uncertainty of future PV output into multiple typical scenarios and their probability distributions. For each set of predicted data, it calculates the correlation coefficient between the net power time-series data of each node in the static cluster and the overall net power time-series data of the cluster, quantifying the degree of synchronization or deviation between node power fluctuations and the overall cluster trend. Based on the correlation coefficient and the probability values ​​of each scenario, it comprehensively evaluates the deviation of each node, incorporating the instability of nodes under normal and extreme operating conditions into a unified metric through multi-scenario fusion. This application can accurately identify nodes that are prone to cluster imbalance during severe PV fluctuations, overcoming the shortcomings of traditional methods that rely on single-scenario or deterministic data and ignore the impact of meteorological uncertainty on node deviation behavior, providing a quantitative basis for subsequent boundary adjustments under dynamic operating conditions.

[0013] Furthermore, the fusion of the attribution ambiguity and the deviation to obtain the comprehensive attribution uncertainty of each node in the static cluster includes: For each node in the static cluster, calculate the first product of the attribution ambiguity and the preset first weight coefficient, and the second product of the deviation and the preset second weight coefficient; Add the first product to the second product to obtain the overall attribution uncertainty of each node in the static cluster.

[0014] By calculating the first product of the node's ambiguity and a preset first weighting coefficient, and the second product of the node's deviation and a preset second weighting coefficient, independent and adjustable contribution weights are assigned to the physical structure boundary attributes and operational state fluctuation attributes, respectively. The first product and the second product are added together to obtain the node's comprehensive ambiguity. By linear weighting, the two types of ambiguity features with different dimensions are fused into a composite index with a unified dimension. This allows the node's ambiguity on the electrical boundary and its instability under photovoltaic fluctuations to be simultaneously included in the same evaluation system, providing a comprehensive quantitative basis for subsequent boundary node selection and adjustment scheme decisions.

[0015] Further, determining the node affiliation adjustment scheme within the static cluster based on the comprehensive affiliation uncertainty includes: Nodes whose overall attribution uncertainty is greater than a preset threshold and which have a direct topological connection with adjacent static clusters are selected from the static clusters and used as the adjustment node set. Based on the adjusted node set, a dual-objective optimization model is constructed to minimize the first objective and the second objective; wherein, the first objective is the sum of the degree to which the voltage of all nodes in the static cluster and the adjacent static cluster deviates from the preset range, and the second objective is the number of transferred nodes; Solve the bi-objective optimization model to obtain the node affiliation adjustment scheme for each node in the adjusted node set.

[0016] This application selects nodes from the static cluster whose overall attribution uncertainty exceeds a preset threshold and which have direct topological connections with adjacent static clusters as the adjustment node set. This compresses the scope of boundary adjustment from the entire cluster to a very small number of key nodes, avoiding computational redundancy in global reconstruction. Based on the adjustment node set, a dual-objective optimization model is constructed, aiming to minimize the sum of voltage deviations from the preset range and the number of transferred nodes. This model achieves a Pareto tradeoff between voltage over-limit suppression and action costs, balancing adjustment effectiveness and operational costs. Solving the dual-objective optimization model yields the attribution adjustment scheme for each node. Through mathematical programming, the optimal boundary fine-tuning strategy is obtained, eliminating voltage over-limits by adjusting only a very small number of boundary nodes. This realizes an adaptive boundary adjustment mechanism that restores system operational safety at minimal cost.

[0017] Furthermore, based on the acquired distribution network topology data, distribution network load data, and historical photovoltaic power output data, the distribution network nodes are divided to obtain several static clusters, including: Based on the first electrical coupling similarity, a genetic algorithm is used to randomly allocate power grid nodes and generate several candidate allocation schemes. Based on the candidate partitioning scheme and the first electrical coupling similarity, calculate the intra-cluster electrical distance value, inter-cluster power complementarity value, and self-balancing margin value of the candidate partitioning scheme; The overall fitness of each candidate partitioning scheme is calculated based on the internal electrical distance value, the inter-cluster power complementarity value, and the self-balancing margin value. Based on the comprehensive fitness, the optimal partitioning scheme is determined to obtain several static clusters.

[0018] Based on the first electrical coupling similarity, this application uses a genetic algorithm to randomly allocate power grid nodes to generate several candidate partitioning schemes. A multi-objective optimization framework is used to search for various possible cluster partitioning structures in parallel. The internal electrical distance value, inter-cluster power complementarity value, and self-balancing margin value of each candidate scheme are calculated to quantify the cluster's cohesion, inter-cluster mutual assistance potential, and anti-fluctuation capability, forming mutually balancing evaluation dimensions. These three values ​​are weighted and fused into a comprehensive fitness value, and a dynamic trade-off mechanism is used to automatically balance the Pareto competition between cluster autonomy and inter-interval interconnection. The optimal partitioning scheme is selected based on the comprehensive fitness value, and node affiliation is determined, resulting in several static clusters. This provides a stable framework foundation for subsequent dynamic boundary adjustments, characterized by tight internal electrical coupling, balanced inter-interval complementarity, and self-balancing margin.

[0019] Further, the step of calculating the intra-cluster electrical distance value, inter-cluster power complementarity value, and self-balancing margin value of the candidate partitioning scheme based on the candidate partitioning scheme and the first electrical coupling similarity includes: For each node in the candidate partitioning scheme, the sum of the fourth electrical coupling similarity between the current node and all other nodes in the candidate partitioning scheme is calculated to obtain the weighted degree of each node; the total weight of the candidate partitioning scheme is determined based on the weighted degree; and the electrical distance value within the cluster is calculated based on the weighted degree and the total weight. Based on the distribution network load data and the predicted photovoltaic output data, determine the net active power time series data and net reactive power time series data for each static cluster in the candidate partitioning scheme; based on the net active power time series data and net reactive power time series data of any two different static clusters, determine the active power correlation value and reactive power correlation value for each pair of static clusters; based on the two correlation values ​​of each pair of static clusters, calculate the power complementarity value between clusters. Based on the net active power time-series data of each static cluster, determine the power fluctuation standard deviation of each static cluster in the candidate partitioning scheme; based on the number of nodes contained in each static cluster and the power fluctuation standard deviation, determine the weighted average standard deviation; based on the power fluctuation standard deviation and the weighted average standard deviation of all nodes in the candidate partitioning scheme, calculate the self-balancing margin value.

[0020] This application calculates the weighted degree and total weight of each node based on the electrical coupling similarity of nodes in the candidate partitioning schemes, and then calculates the electrical distance value within the cluster to quantify the tightness of electrical connections between nodes within the cluster. Based on distribution network load data and predicted photovoltaic output data, it determines the net active power time-series data and net reactive power time-series data for each static cluster, calculates the active power correlation value and reactive power correlation value between any two clusters, and then obtains the power complementarity value between clusters to quantify the complementary and mutually supportive potential of different clusters in power fluctuations. Based on the net active power time-series data of each static cluster, it calculates the power fluctuation standard deviation of each cluster, determines the weighted average standard deviation based on the number of nodes in the cluster, and then calculates the self-balancing margin value to quantify the cluster's self-regulation capability in response to photovoltaic fluctuations. These three values ​​comprehensively evaluate the merits of the candidate partitioning schemes from three dimensions: cluster cohesion, interval complementarity, and anti-fluctuation resilience, providing a quantitative basis for subsequent multi-objective optimization and the selection of the optimal static cluster.

[0021] Another embodiment of the present invention provides a dynamic adjustment system for distribution network cluster boundaries, wherein the distribution network cluster adjustment system is used in the distribution network cluster adjustment method described in this application, comprising: Based on the acquired distribution network topology data, distribution network load data, and historical photovoltaic power output data, the distribution network nodes are divided to obtain several static clusters; When any static cluster in the distribution network is detected to meet the preset conditions, the first degree of association between each node in the static cluster and the static cluster, and the second degree of association between each node in the static cluster and the adjacent static cluster are determined according to the distribution network topology data, the distribution network load data, and the historical photovoltaic output data; the affiliation ambiguity of each node in the static cluster is determined according to the first degree of association and the second degree of association. Obtain meteorological data of the power distribution network, and determine the degree of deviation of each node in the static cluster from the overall operating trend of the static cluster based on the historical photovoltaic output data and the meteorological data of the power distribution network; The attribution ambiguity and the deviation are fused to obtain the comprehensive attribution uncertainty of each node in the static cluster; Based on the comprehensive attribution uncertainty, a node attribution adjustment scheme is determined within the static cluster, and the cluster boundary of the distribution network is adjusted according to the node attribution adjustment scheme.

[0022] Another embodiment of the present invention provides a computer-readable storage medium, including: a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the distribution network clustering adjustment method as described in this application. Attached Figure Description

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

[0024] Figure 1 This is a flowchart illustrating one embodiment of the cluster adjustment of the distribution network provided in this application; Figure 2 This is a schematic diagram of a static cluster partitioning process provided in one embodiment of this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0027] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0028] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0029] In scenarios with high penetration of distributed photovoltaic (PV) power, the high proportion of distributed PV grid connections leads to frequent fluctuations in the operating status of the distribution network. The static hard boundaries formed by traditional cluster partitioning methods cannot adapt to real-time fluctuations in PV output, easily causing voltage exceedance issues. To address this, this application proposes a distribution network cluster adjustment method that can identify boundary nodes located at electrical boundaries and exhibiting dynamic swing characteristics, and perform local boundary adjustments. This improves the accuracy of distribution network cluster partitioning and enhances its adaptability to PV fluctuations.

[0030] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating one embodiment of the distribution network cluster adjustment method provided in this application. To address the technical problem that static hard boundaries cannot adapt to drastic photovoltaic fluctuations, one embodiment of this application provides a distribution network cluster adjustment method, including steps S1 to S5, each step as follows: Step S1: Based on the obtained distribution network topology data, distribution network load data, and historical photovoltaic power output data, the distribution network nodes are divided to obtain several static clusters; Step S2: When any static cluster in the distribution network meets the preset conditions, based on the distribution network topology data, distribution network load data, and historical photovoltaic power output data, determine the first degree of association between each node in the static cluster and the static cluster, and the second degree of association between each node in the static cluster and the adjacent static cluster; based on the first degree of association and the second degree of association, determine the ambiguity of the affiliation of each node in the static cluster. Step S3: Obtain meteorological data of the distribution network, and determine the degree of deviation of each node in the static cluster from the overall operating trend of the static cluster based on historical photovoltaic output data and meteorological data of the distribution network. Step S4: Fuse the attribution ambiguity and deviation to obtain the comprehensive attribution uncertainty of each node in the static cluster; Step S5: Based on the comprehensive uncertainty of node affiliation, determine the node affiliation adjustment scheme within the static cluster, and adjust the cluster boundary of the distribution network according to the node affiliation adjustment scheme.

[0031] Please refer to Figure 2 In some embodiments, based on the acquired distribution network topology data, distribution network load data, and historical photovoltaic power output data, the distribution network nodes are divided to obtain several static clusters, including steps S11 to S14, each step of which is as follows: Step S11: Based on the first electrical coupling similarity, a genetic algorithm is used to randomly allocate power grid nodes and generate several candidate partitioning schemes; Step S12: Based on the candidate partitioning scheme and the first electrical coupling similarity, calculate the intra-cluster electrical distance value, inter-cluster power complementarity value, and self-balancing margin value of the candidate partitioning scheme; Step S13: Calculate the overall fitness of each candidate partitioning scheme based on the internal electrical distance value, the inter-cluster power complementarity value, and the self-balancing margin value. Step S14: Determine the optimal partitioning scheme based on the comprehensive fitness to obtain several static clusters.

[0032] Electrical coupling similarity refers to the degree of coupling between nodes calculated based on the voltage-reactive power sensitivity matrix, used to characterize the electrical proximity of two nodes. Candidate partitioning schemes refer to a method of partitioning a distribution network node cluster corresponding to each chromosome in the genetic algorithm.

[0033] The intra-cluster electrical distance value is a comprehensive indicator measuring the tightness of electrical coupling between nodes within a cluster; a higher value indicates tighter electrical connections within the cluster. The inter-cluster power complementarity value is a comprehensive indicator measuring the degree of negative correlation between the net power curves of different clusters; a higher value indicates stronger potential for mutual complementarity and support between clusters. The self-balancing margin value is a comprehensive indicator measuring the cluster's ability to self-regulate in response to fluctuations in photovoltaic output; a higher value indicates stronger cluster autonomy. The overall fitness is the final indicator used in the genetic algorithm to evaluate the merits of candidate partitioning schemes, obtained by weighted summation of the above three sub-objectives.

[0034] In some embodiments, step S12 further includes steps S121 to S123, each of which is as follows: Step S121: For each node in the candidate partitioning scheme, calculate the sum of the fourth electrical coupling similarity between the current node and all other nodes in the candidate partitioning scheme to obtain the weighted degree of each node; determine the total weight of the candidate partitioning scheme based on the weighted degree; calculate the electrical distance value within the cluster based on the weighted degree and the total weight. Step S122: Based on the distribution network load data and predicted photovoltaic output data, determine the net active power time series data and net reactive power time series data of each static cluster in the candidate partitioning scheme; based on the net active power time series data and net reactive power time series data of any two different static clusters, determine the active power correlation value and reactive power correlation value of each pair of static clusters; based on the two correlation values ​​of each pair of static clusters, calculate the power complementarity value between clusters. Step S123: Based on the net active power time series data of each static cluster, determine the power fluctuation standard deviation of each static cluster in the candidate partitioning scheme; based on the number of nodes and power fluctuation standard deviation of each static cluster, determine the weighted average standard deviation; based on the power fluctuation standard deviation and weighted average standard deviation of all nodes in the candidate partitioning scheme, calculate the self-balancing margin value.

[0035] In specific implementations, such as Figure 2 As shown, Figure 2This is a schematic diagram of a static cluster partitioning process according to an embodiment of this application. Step S1 requires dividing the distribution network nodes into several static clusters with tight internal electrical coupling, strong source-load complementarity, and a certain power self-balancing capability, as the basic topology framework for subsequent flexible boundary adjustments.

[0036] In step S11, based on the photovoltaic power output time-series data of each node in the distribution network under multiple typical scenarios, combined with the distribution network topology, line impedance parameters, and load data of each node, the power flow calculation engine is invoked to perform numerical simulation of the distribution network, obtaining the ground state and global state variables under each evolution scenario; the state variables include at least node voltage amplitude, branch power flow, and system network loss, used to verify the satisfaction of the grid's operating constraints under each scenario; and the voltage-reactive power sensitivity matrix is ​​calculated: Where H, N, M, and L are Jacobian submatrices. Based on Matrix, any two nodes and The electrical coupling similarity between them is defined as: The value range is [-1, 1], with values ​​closer to 1 indicating stronger electrical coupling between the two nodes. To facilitate the calculation of modularity and other metrics in subsequent genetic algorithms, the similarity is mapped to the [0, 1] interval to obtain the electrical coupling similarity weight: in, This is a scaling factor used to adjust the discriminative power of the weights. Represents a node When the injected unit reactive power changes, the node The change in voltage amplitude reflects the reactive voltage coupling strength between nodes. The final weight matrix is ​​then obtained. This serves as the foundational data for subsequent static cluster partitioning and ambiguity calculation.

[0037] A landmark-based coding method is used to randomly generate multiple chromosomes, each corresponding to a candidate partitioning scheme. The chromosome length is equal to the total number of distribution network nodes. , No. Value of each gene locus Represents a node With nodes There is a connection relationship, meaning the two belong to the same cluster or are connected by a path. The initial population contains... Such chromosomes (e.g.) .

[0038] For each chromosome in the current population, a depth-first search algorithm is used to traverse the graph structure it maps to, identifying all connected components. Each connected component corresponds to a static cluster, thus obtaining a specific partitioning scheme. ,in The number of clusters (automatically determined by decoding, no need to specify in advance).

[0039] In step S12, in each iteration of the genetic algorithm, for each candidate partitioning scheme, i.e., the cluster partitioning scheme obtained after decoding each chromosome, the following three sub-objective values ​​are calculated in parallel: In step S121, for each node in the candidate partitioning scheme Calculate the sum of the electrical coupling similarities between the current node and all other nodes in the candidate partitioning scheme to obtain the weighted degree of each node. : in, For nodes With nodes The electrical coupling similarity weights between nodes are used to determine the total weight of the candidate partitioning scheme based on the weighted degrees of all nodes. According to the weighting degree and the total weight Calculate the electrical distance value within the cluster. Specifically, a weighted modularity is adopted. : in, For the Kronecker function, when the node and The value is 1 if the elements belong to the same cluster, and 0 otherwise. After normalization, the electrical distance within the cluster is obtained: The larger the value, the tighter the electrical coupling within the cluster.

[0040] In step S122, based on the distribution network load data and the predicted photovoltaic output data, the net active power time series data of each static cluster in the candidate partitioning scheme is determined. Net reactive power time series data Wherein, net power = sum of photovoltaic output within the cluster minus sum of load within the cluster. Based on any two different static clusters... and Using net active power time-series data and net reactive power time-series data, determine the active power correlation value for each pair of static clusters. Reactive power correlation value For example, the Pearson correlation coefficient can be used for calculation. Based on the two correlation values ​​for each pair of static clusters, the power complementarity value between clusters is calculated. : in, This represents the total number of clusters. The larger this value, the stronger the potential for power complementarity between clusters.

[0041] In step S123, each static cluster in the candidate partitioning scheme is determined based on the net active power time-series data of each static cluster. power fluctuation standard deviation . Specifically, For clusters The standard deviation of net power imbalance (total PV output within the cluster minus total load) under all typical evolution scenarios. Then, based on the number of nodes contained in each static cluster. and power fluctuation standard deviation Determine the weighted average standard deviation : in, This represents the total number of nodes in the distribution network.

[0042] Calculate the self-balancing margin value based on the power fluctuation standard deviation and weighted average standard deviation of all nodes in the candidate partitioning scheme. .Pick The self-balancing margin value is the maximum possible estimate of the system's imbalance (e.g., the standard deviation of the total network net load). : The larger the self-balancing margin value, the stronger the cluster's ability to self-regulate in response to photovoltaic fluctuations.

[0043] After the above three sub-objective values ​​are calculated, the process proceeds to step S13, the comprehensive fitness calculation module. The comprehensive fitness of each candidate partitioning scheme is obtained by weighted summation using the dynamic entropy weight method. Based on this, subsequent genetic operations such as selection, crossover, and mutation are performed to finally output the optimal cluster partitioning scheme.

[0044] To avoid biased partitioning caused by fixed weights, this application introduces a dynamic entropy weight mechanism: based on the dispersion of the three sub-objective values ​​of all candidate partitioning schemes in the current population, the weight of each sub-objective is automatically calculated. Specific step S13 is as follows: Let the population size be... For the first The first candidate partitioning scheme Sub-target values ( Corresponding to , , The dimensionless processing was performed using the range method: Then calculate the first... Information entropy of each sub-target : The dynamic weights of each sub-objective are determined based on information entropy: The overall fitness of each candidate partitioning scheme Defined as a weighted sum of the above three sub-objective values: In step S14, the candidate splitting scheme with the highest overall fitness in the current population (i.e., the optimal individual) is recorded. Then, genetic operations are performed: a roulette wheel selection strategy is used to select parent individuals from the current population, perform single-point crossover with a certain crossover probability, and perform basic position mutation with a certain mutation probability to generate a new generation of population. This process is repeated until the stopping condition is met (such as reaching the preset maximum number of iterations or the fitness value no longer improving for several consecutive generations).

[0045] When the iteration stops, the candidate partitioning schemes corresponding to the historical best individuals are output as the optimal partitioning scheme. This scheme divides the distribution network nodes into several static clusters. ,in This represents the number of clusters.

[0046] Through the above process, the static cluster obtained in this step has tight internal electrical coupling. High), with complementary and mutually supportive potential among clusters ( (High), and each cluster has a certain degree of self-balancing ability ( (High), providing a stable basic topological framework for subsequent quantification of attribution uncertainty and adaptive adjustment of boundaries.

[0047] In some embodiments, step S2 includes steps S21 to S23, each of which is as follows: Step S21: Determine the first electrical coupling similarity between distribution network nodes based on distribution network topology data, distribution network load data, and historical photovoltaic power output data; Step S22: For each node in the static cluster, calculate the sum of the second electrical coupling similarity between the current node and all other nodes in the static cluster to obtain the first degree of association; Step S23: For each node in the static cluster, calculate the sum of the third electrical coupling similarity between the current node and all other nodes in the adjacent static cluster to obtain the second degree of association.

[0048] The first degree of association refers to the sum of the electrical coupling similarities between an edge node in a static cluster and all other nodes in that cluster, used to quantify the electrical attraction of that node to its own cluster. The second degree of association refers to the sum of the electrical coupling similarities between an edge node in a static cluster and all nodes in adjacent static clusters, used to quantify the electrical attraction of that node to its neighboring clusters. Edge nodes are nodes in a static cluster that have direct physical topological connections (i.e., at least one connecting branch) with other clusters; these nodes are the primary candidates for cluster boundary adjustments.

[0049] Attribution ambiguity refers to the degree of uncertainty in the physical topology of an edge node, which is between its own cluster and the adjacent clusters. Its value ranges from [0,1], and the closer the value is to 1, the more likely the node is to cross the cluster boundary.

[0050] The preset conditions refer to the operational boundary warning criteria that trigger cluster boundary adjustments, including node voltage exceeding the limit and the duration exceeding the delay immunity threshold, or the cluster self-balancing margin being exhausted and adjacent clusters having excess adjustment margin.

[0051] In some embodiments, after step S23, steps S24 and S25 are further included, and the specific details of each step are as follows: Step S24: Calculate the difference between the first degree of association and the second degree of association, and determine the first intermediate value based on the difference; calculate the sum of the first degree of association and the second degree of association to obtain the second intermediate value; Step S25: Determine the ambiguity of the affiliation of each node in the static cluster based on the ratio of the first intermediate value to the second intermediate value.

[0052] In the specific implementation process, after completing the static cluster partitioning in step S1, the system continuously monitors the distribution network in real time. When any static cluster meets preset conditions, for example, if the voltage of a node within the cluster continuously exceeds the delay immunity threshold, or if the real-time net power deficit / surplus of the cluster exceeds the limit capacity of its internal adjustable resources and adjacent clusters have surplus adjustment margin, then the boundary adjustment process is triggered, and step 2 is executed. For each edge node in the triggered cluster, the degree to which it is physically affected by the "two-way tearing" between its own cluster and adjacent clusters is quantified, i.e., the affiliation ambiguity is calculated. The specific implementation process is as follows.

[0053] In step S21, the electrical coupling similarity weight matrix calculated in step S1 is directly called. ,in That is, a node With nodes The first electrical coupling similarity between them. This matrix has integrated the global voltage-reactive power sensitivity coupling relationship of the distribution network, and its element values ​​range from... A larger value indicates a tighter electrical connection between the two nodes.

[0054] In step S22, the first correlation degree is calculated for the static cluster where the triggering electrical event is located. Identify all nodes within the cluster. For each edge node in the cluster... (i.e., nodes that have physical topological connections with adjacent clusters), calculate the current node With the static cluster The sum of the second electrical coupling similarities between all other nodes within the same node yields the first degree of association. : First degree of correlation Indicators reflect nodes The degree of electrical connection with the overall cluster. The larger the value, the more "embedded" the node is within the cluster, and the stronger its attraction to the cluster.

[0055] In step S23, while calculating the second degree of correlation, the relationship with... Adjacent static clusters (i.e., clusters directly connected via link branches). This applies to the same edge node. Calculate the current node With the adjacent static cluster The sum of the third electrical coupling similarity among all nodes within the node yields the second degree of association. : Second degree of correlation Indicators reflect nodes The strength of the electrical connection with adjacent clusters. If A relatively large value indicates that the node is electrically close to the cluster boundary and has the potential for cross-regional mutual assistance.

[0056] In step S24, the first intermediate value and the second intermediate value are calculated based on the first correlation degree obtained above. Second degree of correlation .

[0057] The first median value is the absolute value of the difference between the first degree of association and the second degree of association, i.e. The second median value is the sum of the first and second degrees of association, i.e. .

[0058] In step S25, the affiliation ambiguity is determined based on the ratio of the first median value to the second median value. Using the ratio of the first median value to the second median value, the node affiliation is calculated according to the following formula. Ambiguity of Attribution : The physical meaning of this formula is: the more balanced the attraction of a node to its own cluster and neighboring clusters (i.e., the better). ),ratio Approaching 0, thus A value close to 1 indicates that the node is located at an "electrical watershed" in physical space, with high ambiguity in its affiliation; conversely, if the node is deeply embedded in the cluster body ( If the ratio approaches 1, then the ratio is close to 1. If the value approaches 0, the attribution is highly certain.

[0059] Through the steps described above, this application quantifies the physical topology-based ownership uncertainty of each edge node. This ownership ambiguity will serve as one of the bases for fusion with dynamic deviation in subsequent steps to form the final comprehensive ownership uncertainty, which is used to identify active boundary nodes that need to be transferred.

[0060] In some embodiments, step S3 includes steps S31 to S34, each of which is as follows: Step S31: Input historical photovoltaic power output data and distribution network meteorological data into the preset prediction model, and generate several sets of predicted photovoltaic power output data sequences based on the time series data of light intensity and temperature in the distribution network meteorological data. Step S32: Determine the probability value corresponding to each set of predicted photovoltaic power output data sequences based on the predicted photovoltaic power output data sequences; Step S33: For each set of photovoltaic power output data sequences, calculate the correlation coefficient between the net power time series data of each node in the static cluster and the overall net power time series data of the static cluster; wherein, the net power time series data of the node is determined based on the photovoltaic power output data and node load of the node; the overall net power time series data of the static cluster is determined based on the net power of all nodes in the static cluster. Step S34: Calculate the deviation of each node in the static cluster from the overall operating trend of the static cluster based on the correlation coefficient and probability value.

[0061] The pre-defined prediction model refers to the set of models used in this application to generate photovoltaic power output data under different scenarios. Specifically, it includes the BASS diffusion model (for predicting the total installed photovoltaic capacity in a region), a node allocation model (for allocating the total capacity to each node), a irradiance beta distribution model (for describing the randomness of irradiance), a photovoltaic power output conversion model (for converting irradiance into power output), a spatiotemporal correlation model (for reflecting the dynamic correlation between nodes), and a scenario reduction algorithm (for extracting photovoltaic power output data for typical scenarios). This model takes historical photovoltaic power output data and meteorological data as input and outputs several sets of photovoltaic power output time-series scenarios under typical scenarios and their corresponding probabilities. Scenario reduction refers to the technique of selecting a few typical scenarios from a large number of original scenarios and redistributing their probabilities. The aim is to reduce the estimation complexity while retaining the main statistical characteristics of the original scenarios (such as mean, variance, and extreme value boundaries). Typical scenarios refer to the representative photovoltaic power output time-series data sequences retained after scenario reduction. Each typical scenario corresponds to an occurrence probability (i.e., the likelihood that the scenario may occur in the future).

[0062] Net power time-series data refers to the sequence of changes in the difference between photovoltaic output and load power of a node or cluster over a specific time interval (e.g., 24 hours). For a node, net power = photovoltaic output - node load; for a cluster, net power = the sum of the net power of all nodes within the cluster. A positive net power value indicates a power surplus, while a negative value indicates a power deficit.

[0063] The correlation coefficient is a statistical indicator that measures the degree of linear correlation between two variables, with a value ranging from -1 to 1. In this application, it is used to quantify the waveform similarity between the node net power time series curve and the overall cluster net power time series curve. A correlation coefficient close to 1 indicates that the two trends are highly consistent, while a correlation coefficient close to -1 indicates that the trends are opposite.

[0064] Deviation is used to quantify the degree to which a node deviates from the overall operating trend of the cluster. The larger the value, the more "abnormal" the node is under extreme conditions, and the more likely it is to become an active node at the boundary that needs to be cut.

[0065] In the specific implementation process, after completing step 2 of static cluster partitioning and quantifying the ambiguity of node physical space affiliation, the degree to which each node deviates from the overall trend of the cluster during dynamic operation is further evaluated, i.e., the dynamic source-load state deviation. This deviation can identify potential nodes that perform normally under normal operating conditions but deviate significantly from the cluster trend under extreme fluctuation conditions.

[0066] Specifically, step S3 involves acquiring meteorological data (including time-series data of solar intensity and ambient temperature) within the distribution network area and combining it with historical photovoltaic power output data to calculate the deviation of each node.

[0067] In steps S31 and S32, the BASS diffusion evolution model is constructed by inputting historical photovoltaic power output data (i.e., the historical installed capacity sequence of the target area) and distribution network meteorological data into a preset prediction model. A distributed photovoltaic evolution model based on the BASS dynamic equation is established: in, for The cumulative installed photovoltaic capacity at any given time To achieve market saturation capacity, For innovation coefficient, The simulation coefficient is determined through a three-stage parameter calibration method (combining roof area, grid capacity, and policy objectives). Then, through linear transformation and nonlinear optimization fitting... , The model parameters are obtained.

[0068] The node allocation model allocates the cumulative photovoltaic installed capacity to each node. A node attractiveness factor is defined by considering the node's return on investment, voltage margin, and roof area constraints. ,: in, For the weights of the analytic hierarchy process, As evaluation indicators, This is a standardized function. After normalization, we get... The predicted installed capacity of each node is obtained by solving a quadratic programming model. This minimizes the deviation between the actual allocation and the distribution of the attraction factor.

[0069] in, This represents the total installed capacity of the region predicted by the BASS diffusion model, i.e., the cumulative photovoltaic installed capacity. For nodes Maximum installable capacity.

[0070] Based on the predicted installed capacity By combining a probability model of solar irradiance (Beta distribution), a photovoltaic power conversion model, and a spatiotemporal correlation model (exponential decay spatial correlation + first-order autoregressive temporal correlation), a large number (e.g., 1000 sets) of 24-hour photovoltaic power output curves with spatiotemporal correlation are generated using the Cholesky decomposition method. Specifically, the shape parameters of the Beta distribution are first fitted based on historical meteorological data. Then, multivariate normal random numbers with specified spatial correlation are generated through Cholesky decomposition. After probability integral transformation, a solar irradiance sequence conforming to the Beta distribution is obtained, which is finally substituted into the photovoltaic power conversion model. This allows us to obtain photovoltaic power output data for each node, each moment, and each scenario. ,in The scene is assigned a number. The resulting large amount of data constitutes the original set of predicted photovoltaic power output data sequences.

[0071] Due to the large volume of the original data, using it directly for subsequent calculations would result in excessive computation. Therefore, a synchronous back-substitution reduction method was adopted to extract the data. This group contains representative predicted photovoltaic power output data sequences. (In the original sequence set...) Based on the initial equal probability, iteratively perform the following operations: calculate the distance matrix between each sequence, find the sequence that minimizes the global probability loss, merge its probability with the nearest sequence, and then delete that sequence. Repeat the above process until the number of sequences is reduced to a minimum. Ultimately obtained A set of predicted photovoltaic power output data sequences, denoted as Each group Includes photovoltaic power output data for each node over a 24-hour period. ( Each sequence corresponds to an updated probability. (satisfy These predicted photovoltaic output data sequences and their probabilities will serve as the basis for subsequent static cluster partitioning and dynamic deviation calculations.

[0072] In step S33, for each set of photovoltaic power output data sequences under typical scenarios after reductions in steps S31 and S32, the nodes... Net power timing data Defined as the light output of the node minus the load: in, For nodes In the scene Next moment Photovoltaic power output, This is for node load. Static cluster. Overall net power timing data Defined as the sum of the net power of all nodes in the cluster: in, Positive values ​​indicate power surplus, while negative values ​​indicate power deficit.

[0073] For each set of typical photovoltaic power output data sequences, the computing nodes Pearson correlation coefficient between the net power time series curve of the individual and the overall net power time series curve of the cluster : in, The time series length (e.g., 24 hours). and These are the mean values ​​of the corresponding time series curves. The range of values ​​for this correlation coefficient is... .when When the value approaches 1, the fluctuation trend of the node is highly consistent with that of the overall cluster; when When the value is negative, the node fluctuation direction is opposite to that of the cluster as a whole, which will exacerbate the difficulty of cluster tuning.

[0074] Preferably, in order to highlight the impact of extreme operating conditions on deviation, this application introduces a severity penalty factor. First, for each set of typical photovoltaic power output data sequences, the computing cluster... Maximum absolute value of net power: This value reflects the maximum impact of this data set on the cluster. Let... Given the rated adjustment capacity of the adjustable devices within the cluster, the penalty factor is defined as follows: in, This is a penalty amplification factor (typically 2-5). When the shock exceeds the cluster's own adjustment capacity, Furthermore, it grows exponentially, thus giving this scenario a very high weight in subsequent calculations, ensuring that extreme operating conditions have a dominant influence on node deviation.

[0075] In step S34, the probability values ​​corresponding to the photovoltaic power output data sequence are combined. With penalty factor compute nodes The original dynamic deviation : in, This is the deviation coefficient of the node under this set of data.

[0076] Normalize the original deviation of all nodes in the cluster and map it to The interval is used to obtain the final deviation. : Deviation The larger the value, the more the node deviates from the overall trend of the cluster under extreme conditions, and the more likely it is to become a boundary active node that needs to be transferred.

[0077] In some embodiments, step S4 includes steps S41 and S42, each of which is as follows: Step S41: For each node in the static cluster, calculate the first product of the attribution ambiguity and the preset first weight coefficient, and the second product of the deviation and the preset second weight coefficient; Step S42: Add the first product and the second product to obtain the comprehensive attribution uncertainty of each node in the static cluster.

[0078] Among them, the comprehensive attribution uncertainty refers to the comprehensive index obtained by fusing attribution ambiguity and dynamic deviation, which is used to quantify the dual uncertainty of a node in terms of physical topology and operational status. The larger the value, the more likely the node is to cross the cluster boundary.

[0079] In the specific implementation process, the static physical space affiliation ambiguity of each node is calculated separately. (Step S2) and the deviation of the dynamic source load state Following step S3, step S41 merges these two indicators to form a comprehensive indicator that fully reflects the dual uncertainties of the node in terms of physical topology and operational status—the comprehensive attribution uncertainty. .

[0080] In step S41, a first weighting coefficient is preset. Second weighting coefficient These are used to adjust the contribution weights of attribution ambiguity and deviation in the final comprehensive index, respectively. The two weighting coefficients satisfy... All values ​​are positive. The specific values ​​of the weighting coefficients can be set according to actual project requirements. This applies to each node within the static cluster. Perform the following two multiplication operations respectively: [Calculate the ambiguity of the attribution] With the first weighting coefficient Multiply, we get Deviation With the second weighting coefficient Multiply, we get .

[0081] In step S42, the first product and the second product are added together to obtain the node. Comprehensive attribution uncertainty : because and All have been mapped to the previous normalization steps. interval, and ,therefore The range of values ​​is also .

[0082] Overall Attribution Uncertainty This refers to the node in the electrical topology ( ) and dynamic operation ( The degree of "swinging" in both dimensions. The larger the value, the more likely the node is to cross the cluster boundary, and it belongs to the "boundary active node" that needs to be closely monitored and may be subject to affiliation adjustment; The smaller the value, the more stable the node's affiliation, and it should remain within the current cluster.

[0083] In practical applications, a preset threshold can be set. , will satisfy The nodes are identified as the set of active nodes at the global boundary. .

[0084] In some specific embodiments, step S5 includes steps S51 to S53, and the specific details of each step are as follows: Step S51: Select nodes from the static cluster whose overall attribution uncertainty is greater than a preset threshold and which have a direct topological connection with adjacent static clusters, and use them as the adjustment node set; Step S52: Based on the adjusted node set, construct a dual-objective optimization model that minimizes the first objective and the second objective; wherein, the first objective is the sum of the degree to which the voltage of all nodes in the static cluster and the adjacent static cluster deviates from the preset range, and the second objective is the number of transferred nodes; Step S53: Solve the bi-objective optimization model to obtain the node affiliation adjustment scheme for each node in the adjustment node set.

[0085] The adjustment node set refers to the set of nodes between the triggering cluster and its neighboring clusters that satisfy the conditions that "the overall attribution uncertainty is greater than a preset threshold" and "there is a direct physical topological connection with the neighboring clusters". This set is the source of decision variables for the subsequent optimization model.

[0086] The sum of voltage deviations from the preset range refers to the sum of the voltage over-limit penalty values ​​for all nodes in the trigger cluster and its adjacent clusters. The value is 0 when the voltage is within the safe range, and positive when the voltage exceeds the limit, with the value increasing as the deviation becomes more severe. The primary objective is to sum this penalty value for all relevant nodes.

[0087] The number of nodes transferred refers to the total number of nodes that are decided to be transferred from the original cluster to the adjacent cluster in the adjustment node set. It directly corresponds to the number of physical switch operations or communication link reconfigurations and is an indicator for measuring the adjustment cost.

[0088] In the specific implementation process, execution is triggered when a static cluster in the distribution network meets preset conditions (such as node voltage exceeding the limit or cluster self-balancing margin being exhausted).

[0089] In step S51, a preset threshold for the overall attribution uncertainty is set. Will satisfy All nodes are marked as global boundary active nodes, forming a set. .

[0090] In step S52, a subset of the flexible transition zone is extracted in a directional manner when a certain static cluster... When boundary adjustments are triggered due to voltage over-limit or power imbalance, identify adjacent clusters with excess regulation capacity. .from The following nodes are selected through targeted screening as the candidate node set for this adjustment, i.e., the adjustment node set. This node belongs to the trigger cluster. This node and its neighboring clusters There exists a direct topological connection (i.e., there is at least one connecting branch). This can be expressed as: in, To adjust the node set, it only includes nodes located at the cluster edge, with high overall uncertainty of affiliation, and with physical connections to the target cluster, thus avoiding the blind selection of nodes from the entire network.

[0091] For adjusting the node set Each node in Define binary decision variables .in, Indicates that the node The affiliation relationship is changed from the original cluster Temporarily transferred to adjacent cluster ; Indicates maintaining nodes In the original cluster The ownership remains unchanged.

[0092] The primary objective is to trigger the cluster. and its neighboring clusters Minimize the sum of the deviations of the voltage of all nodes from the preset safe range. Define nodes. Voltage deviation penalty function: in, , These represent the lower and upper voltage limits, respectively. The function is positive when the voltage exceeds the limit, and the value increases with the severity of the voltage exceedance; therefore, the first objective expression is: The second objective is to minimize the total number of nodes that undergo ownership transfer, i.e.: The second objective directly corresponds to the number of times the physical contact switch is operated or the number of times the communication link is reconfigured, reflecting the design philosophy of achieving boundary adjustment with minimal action cost.

[0093] The two objectives mentioned above are combined into a single comprehensive minimization objective function: in and This is the normalized weighting coefficient. Since ensuring grid voltage safety is the primary task, it is typically set to... .

[0094] To ensure the safe operation of the system after the transfer, the following constraints are added: Receiver overload constraint: After the transfer, the receiver cluster The new net load (or net power) must not result in The voltage exceeds the limit or the capacity is overloaded.

[0095] Connectivity and radial constraints: transfer variables The combination must ensure that the changed cluster topology still maintains the connectivity graph attribute, and that the entire distribution network maintains a radial operating structure.

[0096] The specific expressions for the above constraints can be linearized based on the actual distribution network model or by using power flow verification.

[0097] In step S53, the above optimization model is a 0-1 integer programming problem, and the number of decision variables is equal to the set of adjustment nodes. The cardinality can be determined using the following method: Exact algorithms: directly solve for the global optimum using branch and bound or Gaussian solvers (such as Gurobi, CPLEX, MATLAB's intlinprog); Heuristic algorithms: For scenarios with high real-time requirements, greedy algorithms or local search can be used to quickly obtain approximate optimal solutions.

[0098] The optimal decision variable sequence is obtained after solving. ,in The node whose affiliation needs to be adjusted is the node whose affiliation needs to be adjusted.

[0099] Perform boundary adjustments according to the optimal action sequence: Physical layer segmentation: For nodes equipped with physical communication switches, the receiving-side cluster is closed remotely. The communication switch, disconnect the original cluster The sectionalizing switch physically integrates the boundary nodes and their associated loads / power supplies into the grid. The power supply area.

[0100] Following layered handover (for soft switching or virtual power plant scenarios): High The node's control command receiving rights are transferred from Seamless transfer of regional controllers to The area controller enables logical switching of ownership within seconds.

[0101] After execution, the cluster boundary adaptively expands or contracts, precisely relieving the pressure on over-limit clusters without affecting the backbone topology of the entire network.

[0102] Preferably, to avoid frequent boundary switching and degradation of switch mechanical life caused by high-frequency fluctuations in photovoltaic power, a state recovery observation window is established. (For example, 10 minutes). When the extreme conditions that triggered the adjustment are resolved, and the original cluster... The voltage, power, and other states are restored to the safe dead zone and maintained. After a certain period, the system automatically issues a rollback command: resetting the transferred nodes back to their original static backbone network affiliation as determined in step S1. This mechanism ensures the long-term stability and management order of the distribution network's basic topology.

[0103] In some embodiments, one embodiment of the distribution network cluster adjustment system provided in this application includes: Based on the acquired distribution network topology data, distribution network load data, and historical photovoltaic power output data, the distribution network nodes are divided to obtain several static clusters; When any static cluster in the distribution network meets the preset conditions, based on the distribution network topology data, distribution network load data, and historical photovoltaic power output data, the first degree of association between each node in the static cluster and the static cluster, and the second degree of association between each node in the static cluster and the adjacent static cluster are determined; based on the first degree of association and the second degree of association, the ambiguity of the affiliation of each node in the static cluster is determined. Obtain meteorological data of the power distribution network, and determine the degree of deviation of each node in the static cluster from the overall operating trend of the static cluster based on historical photovoltaic power output data and meteorological data of the power distribution network; The ambiguity of attribution and the degree of deviation are fused to obtain the comprehensive attribution uncertainty of each node in the static cluster; Based on the comprehensive uncertainty of node affiliation, a node affiliation adjustment scheme is determined within the static cluster, and the cluster boundary of the distribution network is adjusted according to the node affiliation adjustment scheme.

[0104] It is understood that the above system item embodiments correspond to the method item embodiments of this application, and can implement the distribution network cluster adjustment method provided by any of the above method item embodiments of this application.

[0105] Based on the above-described method embodiments, another embodiment of this application provides a computer-readable storage medium, including: a stored computer program, wherein, when the computer program is running, it controls the device where the computer-readable storage medium is located to execute the distribution network cluster adjustment method provided by any of the above-described method embodiments of this application.

[0106] The above are preferred embodiments of this application. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principles of this application, and these improvements and modifications are also considered to be within the scope of protection of this application.

Claims

1. A method for clustered adjustment of a power distribution network, characterized in that, include: Based on the acquired distribution network topology data, distribution network load data, and historical photovoltaic power output data, the distribution network nodes are divided to obtain several static clusters; When any static cluster in the distribution network is detected to meet the preset conditions, the first degree of association between each node in the static cluster and the static cluster, and the second degree of association between each node in the static cluster and the adjacent static cluster are determined according to the distribution network topology data, the distribution network load data, and the historical photovoltaic output data; the affiliation ambiguity of each node in the static cluster is determined according to the first degree of association and the second degree of association. Obtain meteorological data of the power distribution network, and determine the degree of deviation of each node in the static cluster from the overall operating trend of the static cluster based on the historical photovoltaic output data and the meteorological data of the power distribution network; The attribution ambiguity and the deviation are fused to obtain the comprehensive attribution uncertainty of each node in the static cluster; Based on the comprehensive attribution uncertainty, a node attribution adjustment scheme is determined within the static cluster, and the cluster boundary of the distribution network is adjusted according to the node attribution adjustment scheme.

2. The method for cluster adjustment of a distribution network according to claim 1, characterized in that, The step of determining the first degree of association between each node in the static cluster and the static cluster, and the second degree of association between each node in the static cluster and adjacent static clusters, based on the distribution network topology data, the distribution network load data, and the historical photovoltaic output data, includes: Based on the distribution network topology data, the distribution network load data, and the historical photovoltaic power output data, a first electrical coupling similarity between distribution network nodes is determined. For each node in the static cluster, the sum of the second electrical coupling similarities between the current node and all other nodes in the static cluster is calculated to obtain the first degree of association; For each node in the static cluster, the sum of the third electrical coupling similarity between the current node and all other nodes in the adjacent static cluster is calculated to obtain the second degree of association.

3. The method for cluster adjustment of a distribution network according to claim 2, characterized in that, The step of determining the affiliation ambiguity of each node in the static cluster based on the first association degree and the second association degree includes: Calculate the difference between the first degree of association and the second degree of association, and determine a first intermediate value based on the difference; Calculate the sum of the first degree of association and the second degree of association to obtain the second intermediate value; The affiliation ambiguity of each node in the static cluster is determined based on the ratio of the first intermediate value to the second intermediate value.

4. The method for cluster adjustment of a distribution network according to claim 1, characterized in that, The step of determining the deviation of each node in the static cluster from the overall operating trend of the static cluster based on the historical photovoltaic output data and the distribution network meteorological data includes: The historical photovoltaic output data and the distribution network meteorological data are input into a preset prediction model. Based on the time series data of light intensity and temperature in the distribution network meteorological data, several sets of predicted photovoltaic output data sequences are generated. Based on the predicted photovoltaic power output data sequence, determine the probability value corresponding to each set of the predicted photovoltaic power output data sequence; For each set of photovoltaic power output data sequences, the correlation coefficient between the net power time series data of each node in the static cluster and the overall net power time series data of the static cluster is calculated; wherein, the net power time series data of the node is determined based on the photovoltaic power output data and node load of the node; the overall net power time series data of the static cluster is determined based on the net power of all nodes in the static cluster. Based on the correlation coefficient and the probability value, the deviation of each node in the static cluster from the overall operating trend of the static cluster is calculated.

5. The method for cluster adjustment of a distribution network according to claim 1, characterized in that, The process of fusing the attribution ambiguity and the deviation to obtain the comprehensive attribution uncertainty of each node within the static cluster includes: For each node in the static cluster, calculate the first product of the attribution ambiguity and the preset first weight coefficient, and the second product of the deviation and the preset second weight coefficient; Add the first product to the second product to obtain the overall attribution uncertainty of each node in the static cluster.

6. The method for cluster adjustment of a distribution network according to claim 1, characterized in that, The step of determining the node affiliation adjustment scheme within the static cluster based on the comprehensive affiliation uncertainty includes: Nodes whose overall attribution uncertainty is greater than a preset threshold and which have a direct topological connection with adjacent static clusters are selected from the static clusters and used as the adjustment node set. Based on the adjusted node set, a dual-objective optimization model is constructed to minimize the first objective and the second objective; wherein, the first objective is the sum of the degree to which the voltage of all nodes in the static cluster and the adjacent static cluster deviates from the preset range, and the second objective is the number of transferred nodes; Solve the bi-objective optimization model to obtain the node affiliation adjustment scheme for each node in the adjusted node set.

7. The method for cluster adjustment of a distribution network according to claim 2, characterized in that, The distribution network nodes are divided based on the acquired distribution network topology data, distribution network load data, and historical photovoltaic power output data to obtain several static clusters, including: Based on the first electrical coupling similarity, a genetic algorithm is used to randomly allocate power grid nodes and generate several candidate allocation schemes. Based on the candidate partitioning scheme and the first electrical coupling similarity, calculate the intra-cluster electrical distance value, inter-cluster power complementarity value, and self-balancing margin value of the candidate partitioning scheme; The overall fitness of each candidate partitioning scheme is calculated based on the internal electrical distance value, the inter-cluster power complementarity value, and the self-balancing margin value. Based on the comprehensive fitness, the optimal partitioning scheme is determined to obtain several static clusters.

8. The method for cluster adjustment of a distribution network according to any one of claims 7, characterized in that, The step of calculating the intra-cluster electrical distance value, inter-cluster power complementarity value, and self-balancing margin value of the candidate partitioning scheme based on the candidate partitioning scheme and the first electrical coupling similarity includes: For each node in the candidate partitioning scheme, the sum of the fourth electrical coupling similarity between the current node and all other nodes in the candidate partitioning scheme is calculated to obtain the weighted degree of each node; the total weight of the candidate partitioning scheme is determined based on the weighted degree; and the electrical distance value within the cluster is calculated based on the weighted degree and the total weight. Based on the power distribution network load data and the predicted photovoltaic output data, determine the net active power time series data and net reactive power time series data of each static cluster in the candidate partitioning scheme; based on the net active power time series data and net reactive power time series data of any two different static clusters, determine the active power correlation value and reactive power correlation value of each pair of static clusters; based on the two correlation values ​​of each pair of static clusters, calculate the power complementarity value between clusters. Based on the net active power time-series data of each static cluster, determine the power fluctuation standard deviation of each static cluster in the candidate partitioning scheme; based on the number of nodes contained in each static cluster and the power fluctuation standard deviation, determine the weighted average standard deviation; based on the power fluctuation standard deviation and the weighted average standard deviation of all nodes in the candidate partitioning scheme, calculate the self-balancing margin value.

9. A cluster adjustment system for a power distribution network, characterized in that, The distribution network cluster adjustment system is used to execute the distribution network cluster adjustment method as described in any one of claims 1-8, including: Based on the acquired distribution network topology data, distribution network load data, and historical photovoltaic power output data, the distribution network nodes are divided to obtain several static clusters; When any static cluster in the distribution network is detected to meet the preset conditions, the first degree of association between each node in the static cluster and the static cluster, and the second degree of association between each node in the static cluster and the adjacent static cluster are determined according to the distribution network topology data, the distribution network load data, and the historical photovoltaic output data; the affiliation ambiguity of each node in the static cluster is determined according to the first degree of association and the second degree of association. Obtain meteorological data of the power distribution network, and determine the degree of deviation of each node in the static cluster from the overall operating trend of the static cluster based on the historical photovoltaic output data and the meteorological data of the power distribution network; The attribution ambiguity and the deviation are fused to obtain the comprehensive attribution uncertainty of each node in the static cluster; Based on the comprehensive attribution uncertainty, a node attribution adjustment scheme is determined within the static cluster, and the cluster boundary of the distribution network is adjusted according to the node attribution adjustment scheme.

10. A computer-readable storage medium, characterized in that, include: A stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the distribution network clustering adjustment method as described in any one of claims 1 to 8.

Images