Cluster optimization method and system based on improved particle swarm algorithm and multi-objective coupling

Abstract

The application discloses a kind of cluster optimization method and system based on improved particle swarm algorithm and multi-objective coupling, applied to power equipment cluster optimization technical field, method is applied to regional power grid;Regional power grid includes multiple source load nodes;Node cluster is composed of source load node;Method includes: the operating data of each source load node, and the weighted analysis of the complementarity and load rate mean square deviation of each node cluster, obtains the multi-objective comprehensive performance index of each node cluster;Determine the cluster division result of a regional power grid as a particle, obtain particle population;Determine the fitness of each particle based on multi-objective comprehensive performance index;According to the linear decreasing inertia weight strategy and subsection adaptive transfer function of improved particle swarm algorithm, and the fitness of each particle, the particle population is iteratively optimized, and the optimal cluster division result of regional power grid is obtained.The application improves the cluster division accuracy by coupling improved particle swarm algorithm and multi-objective analysis.

Description

Technical Field

[0001] This invention relates to the field of power equipment cluster optimization technology, and in particular to a cluster optimization method and system based on an improved particle swarm optimization algorithm coupled with multi-objectives. Background Technology

[0002] In regional power grids composed of source-load nodes with adjustable resources, existing methods for clustering the regional power grid typically rely on its topology information. The specific steps are as follows: First, electrical parameters and geographic coordinates of each source-load node in the regional power grid are collected to construct a topology model including the source-load nodes. Then, the electrical distance between each source-load node is calculated, and the source-load nodes are clustered based on this electrical distance, grouping those with similar electrical distances into the same cluster.

[0003] However, this cluster partitioning method that only considers physical correlation characteristics has the following drawbacks: the cluster partitioning criteria based on a single indicator cannot be applied to the actual operation scenario of the regional power grid, resulting in the cluster partitioning results obtained by the existing cluster partitioning method not being able to accurately adapt to the actual operation needs of the regional power grid, that is, the cluster partitioning results obtained by the existing cluster partitioning method are not accurate enough. Summary of the Invention

[0004] This invention provides a cluster optimization method and system based on an improved particle swarm optimization algorithm coupled with multi-objectives, in order to solve the technical problem that the cluster partitioning results obtained by existing cluster partitioning methods are not accurate enough.

[0005] To address the aforementioned technical problems, this invention provides a cluster optimization method based on an improved particle swarm optimization algorithm coupled with multi-objective coupling, applied to a regional power grid; the regional power grid comprises multiple source-load nodes; the node cluster consists of the source-load nodes; the method includes: A weighted analysis is performed on the operating data of each source load node, the complementarity of each node cluster, and the root mean square error of the load rate of each node cluster to obtain the multi-objective comprehensive performance index of each node cluster. By determining that a cluster partitioning result of the regional power grid is a particle, the particle population corresponding to the regional power grid is obtained. Based on the aforementioned multi-objective comprehensive performance index, the fitness of each particle is determined; Based on the linearly decreasing inertial weight strategy and piecewise adaptive transfer function of the improved particle swarm optimization algorithm, as well as the fitness of each particle, the particle population is iteratively optimized to obtain the optimal cluster partitioning result of the regional power grid.

[0006] As one preferred embodiment, the weighted analysis of the operating data of each source load node, the complementarity of each node cluster, and the root mean square error of the load rate of each node cluster yields the following multi-objective comprehensive performance indicators for each node cluster: Based on the operating power and node relationship data of each source load node, the electrical coupling tightness index of each node cluster is obtained. Based on the complementarity of adjustment capabilities, temporal complementarity, and type complementarity of each node cluster, the complementarity index of each node cluster is obtained. A weighted analysis was performed on the electrical coupling tightness index, complementarity index, and load rate root mean square error of each node cluster to obtain the multi-objective comprehensive performance index of each node cluster.

[0007] As one preferred embodiment, determining a cluster partitioning result of the regional power grid as a particle, and obtaining the particle population corresponding to the regional power grid includes: The result of a cluster partitioning of the regional power grid is determined as a binary encoded string, and a source-load node is a binary encoded unit; Based on the total number of source and load nodes in the regional power grid, the dimension of the binary encoded string is determined; each dimension corresponds to the cluster affiliation status of a source and load node. Under the relevant operational constraints of the regional power grid, the cluster affiliation status of the source and load nodes is adjusted to obtain the binary encoded string corresponding to each cluster partitioning result of the regional power grid; Each binary encoded string is identified as a particle, thus obtaining the particle population corresponding to the regional power grid.

[0008] As one preferred embodiment, determining the fitness of each particle based on the multi-objective comprehensive performance index includes: Determine the target cluster contained in the cluster partitioning result corresponding to each particle; The fitness of each particle is determined based on the multi-objective comprehensive performance index of the target cluster.

[0009] As one preferred embodiment, the step of iteratively optimizing the particle swarm based on the linearly decreasing inertial weight strategy and piecewise adaptive transfer function of the improved particle swarm optimization algorithm, as well as the fitness of each particle, to obtain the optimal cluster partitioning result of the regional power grid includes: Obtain the iteration termination condition, and determine the current iteration data of the particle population based on the fitness of each particle in the current iteration; When the current iteration data does not meet the iteration termination condition, the current particle search speed is determined based on the current iteration data and the linearly decreasing inertial weight strategy of the improved particle swarm algorithm. Based on the piecewise adaptive transfer function of the improved particle swarm algorithm, the current particle search velocity is converted into a particle adjustment probability. Based on the particle adjustment probability, the particle population is iteratively optimized to obtain the updated iteration data corresponding to the current iteration data; When the updated iteration data meets the iteration termination condition, the optimal cluster partitioning result of the regional power grid is determined based on the updated iteration data; When the updated iteration data does not meet the iteration termination condition, the updated iteration data is used as the current iteration data, and the process returns to the step of determining the current particle search velocity based on the current iteration data and the linearly decreasing inertia weight strategy of the improved particle swarm algorithm when the current iteration data does not meet the iteration termination condition.

[0010] Another embodiment of the present invention provides a cluster optimization system based on an improved particle swarm optimization algorithm coupled with multiple objectives, comprising: The multi-objective comprehensive performance index determination module is used to perform weighted analysis on the operating data of each source load node, the complementarity of each node cluster, and the load rate root mean square error of each node cluster to obtain the multi-objective comprehensive performance index of each node cluster. The particle population construction module is used to determine that a cluster partitioning result of the regional power grid is a particle, and to obtain the particle population corresponding to the regional power grid. The particle fitness determination module is used to determine the fitness of each particle based on the multi-objective comprehensive performance index. The particle swarm iterative optimization module is used to iteratively optimize the particle swarm based on the linear decreasing inertia weight strategy and piecewise adaptive transfer function of the improved particle swarm algorithm, as well as the fitness of each particle, to obtain the optimal cluster partitioning result of the regional power grid.

[0011] As one preferred embodiment, the multi-objective comprehensive performance index determination module includes: The electrical coupling tightness index determination unit is used to obtain the electrical coupling tightness index of each node cluster based on the operating power of each source load node and the node relationship data. The complementarity index determination unit is used to obtain the complementarity index of each node cluster based on the complementarity of the adjustment capability, the complementarity of the timing, and the complementarity of the type of each node cluster. The multi-objective comprehensive performance index determination unit is used to perform weighted analysis on the electrical coupling tightness index, complementarity index, and load rate root mean square error of each node cluster to obtain the multi-objective comprehensive performance index of each node cluster.

[0012] As one preferred embodiment, the particle swarm construction module includes: A binary encoded string determination unit is used to determine that a cluster division result of the regional power grid is a binary encoded string, and a source-load node is a binary encoded unit; The encoding dimension determination unit is used to determine the dimension of the binary encoded string based on the total number of source and load nodes in the regional power grid; each dimension encoding corresponds to the cluster affiliation status of a source and load node. The binary encoded string adjustment unit is used to adjust the cluster affiliation status of the source load nodes under the relevant operational constraints of the regional power grid, so as to obtain the binary encoded string corresponding to each cluster partitioning result of the regional power grid. The particle swarm construction unit is used to determine that each of the binary encoded strings is a particle, thereby obtaining the particle swarm corresponding to the regional power grid.

[0013] As one preferred embodiment, the particle fitness determination module includes: The target cluster determination unit is used to determine the target cluster contained in the cluster partitioning result corresponding to each particle. The particle fitness determination unit is used to determine the fitness of each particle based on the multi-objective comprehensive performance index of the target cluster.

[0014] As one preferred embodiment, the particle population iterative optimization module includes: The current iteration data determination unit is used to obtain the iteration termination condition and determine the current iteration data of the particle population based on the fitness of each particle in the current iteration. The current particle search velocity determination unit is used to determine the current particle search velocity based on the current iteration data and the linear decreasing inertia weight strategy of the improved particle swarm algorithm when the current iteration data does not meet the iteration termination condition. The particle adjustment probability determination unit is used to convert the current particle search velocity into a particle adjustment probability according to the piecewise adaptive transfer function of the improved particle swarm algorithm. The iterative data update unit is used to iteratively optimize the particle population based on the particle adjustment probability to obtain the updated iterative data corresponding to the current iterative data; The optimal cluster partitioning result determination unit is used to determine the optimal cluster partitioning result of the regional power grid based on the updated iteration data when the updated iteration data meets the iteration termination condition. The loop unit is used to take the updated iteration data as the current iteration data when the updated iteration data does not meet the iteration termination condition, and return to execute the step of determining the current particle search speed based on the current iteration data and the linearly decreasing inertia weight strategy of the improved particle swarm algorithm when the current iteration data does not meet the iteration termination condition.

[0015] This application improves the binary particle swarm optimization algorithm by adopting a linearly decreasing inertia weight strategy and a piecewise adaptive transfer function. This balances the global exploration capability and local development capability of the binary particle swarm optimization algorithm, avoids the problems of the binary particle swarm optimization algorithm being prone to getting trapped in local optima and slow convergence speed, and improves the solution accuracy of cluster partitioning. Attached Figure Description

[0016] Figure 1 This is one of the flowcharts of the cluster optimization method based on the improved particle swarm optimization algorithm and multi-objective coupling provided by the present invention; Figure 2 This is the second flowchart of the cluster optimization method based on the improved particle swarm optimization algorithm and multi-objective coupling provided by the present invention; Figure 3 This is a schematic diagram of the cluster optimization system based on the improved particle swarm optimization algorithm and multi-objective coupling provided by the present invention.

[0017] Figure label: Among them, 301 is the multi-objective comprehensive performance index determination module; 302 is the particle population construction module; 303 is the particle fitness determination module; and 304 is the particle population iterative optimization module. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] In the description of this application, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0020] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components. The terms "vertical," "horizontal," "left," "right," "upper," "lower," and similar expressions used herein are for illustrative purposes only and do not indicate or imply that the device or component referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0021] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0022] The cluster optimization method based on an improved particle swarm optimization algorithm coupled with multi-objective coupling provided by this invention is applied to a regional power grid, which comprises multiple source-load nodes forming a node cluster. See also... Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the cluster optimization method based on the improved particle swarm optimization algorithm coupled with multi-objectives provided by the present invention. Figure 1 As shown, this embodiment includes steps 100 to 400, and the specific steps are as follows: Step 100: Perform a weighted analysis on the operating data of each source load node, the complementarity of each node cluster, and the root mean square error of the load rate of each node cluster to obtain the multi-objective comprehensive performance index of each node cluster. In this embodiment, operational data refers to various parameter information characterizing the operating status of source-load nodes, which may include the electrical distance between nodes, transmission loss coefficient, and the total active power of the nodes. Complementarity is a quantitative indicator used to measure the degree of complementary matching between source-load nodes within a node cluster in terms of regulation capability, timing characteristics, and type diversity. Load factor mean square error is the mean square error of the load factor (the ratio of actual load power to rated distribution capacity) of each load node within the node cluster, used to measure the degree of balance in load distribution within the node cluster.

[0023] For each node cluster in the regional power grid, firstly, the electrical coupling tightness index of the cluster is calculated based on the operating data of each source-load node within the cluster; then, the complementarity of the cluster is calculated based on the degree of complementarity and matching between the nodes in terms of regulation capacity, timing characteristics, and type diversity; next, the load rate root mean square error of the cluster is calculated based on the distribution area load rate of each source-load node within the cluster; finally, the electrical coupling tightness index, complementarity, and load rate root mean square error are multiplied by their respective weighting coefficients and summed to obtain the multi-objective comprehensive performance index of the node cluster.

[0024] By constructing a multi-objective comprehensive performance index that integrates electrical coupling tightness, complementarity, and load balancing, a multi-dimensional quantitative evaluation of cluster partitioning schemes is achieved, which solves the problem of the one-sidedness of evaluation by a single index and provides a fitness evaluation basis for subsequent cluster optimization.

[0025] Step 200: Determine that a cluster partitioning result of the regional power grid is a particle, and obtain the particle population corresponding to the regional power grid; In this embodiment, a particle refers to an individual in the improved particle swarm optimization algorithm, where each particle represents a complete cluster partitioning scheme for the regional power grid. A particle swarm refers to a collection of multiple particles, where each particle represents a possible cluster partitioning scheme.

[0026] First, the total number of all source-load nodes in the regional power grid is determined, serving as the dimension for particle encoding. Then, a binary encoding method is used to encode the cluster affiliation status of each source-load node, with each node corresponding to a single bit in the code. The value of this bit represents the cluster to which the node belongs. A complete binary encoded string corresponds to a cluster partitioning result, and this binary encoded string is defined as a particle. Multiple particles are generated through random initialization, and these particles collectively constitute the particle population. The size of the particle population (i.e., the number of particles) can be determined based on the node size and computing resources of the regional power grid; for example, it can be set to any integer value between 50 and 200.

[0027] By transforming the cluster partitioning problem into a particle encoding problem, complex cluster building decisions can be expressed and solved, providing a computable basic data structure for subsequent iterative optimization.

[0028] Step 300: Based on the multi-objective comprehensive performance index, determine the fitness of each particle; In this embodiment, fitness refers to a quantitative value used to evaluate the quality of a particle. The higher the fitness value, the better the cluster partitioning scheme corresponding to that particle.

[0029] For each particle in the particle swarm, firstly, the binary code of the particle is decoded to obtain the cluster partitioning result corresponding to the particle, that is, to determine which node cluster each source payload node is assigned to; then, for each partitioned node cluster, the multi-objective comprehensive performance index of the cluster is calculated; finally, the multi-objective comprehensive performance indices of all node clusters contained in the particle are summed and then averaged to obtain the fitness value of the particle. For example, if the partitioning result corresponding to a particle contains 4 node clusters, the multi-objective comprehensive performance indices of the 4 clusters are added together, and the sum is divided by 4 to obtain the fitness value of the particle.

[0030] By defining the fitness of particles, the partitioning quality of any cluster will affect the overall evaluation of the particles, thereby driving the algorithm to take into account the partitioning quality of all clusters during the optimization process and achieve the globally optimal cluster partitioning.

[0031] Step 400: Based on the linear decreasing inertia weight strategy and piecewise adaptive transfer function of the improved particle swarm algorithm, and the fitness of each particle, the particle swarm is iteratively optimized to obtain the optimal cluster partitioning result of the regional power grid.

[0032] The linearly decreasing inertia weight strategy in this embodiment refers to an update strategy in which the inertia weight gradually decreases according to a linear function as the number of iterations increases. Inertia weight is a parameter in particle swarm optimization (PSO) that controls a particle to maintain its previous state of motion. A larger inertia weight results in a stronger global search capability, while a smaller inertia weight results in a stronger local exploitation capability. The piecewise adaptive transfer function in this embodiment refers to a transfer function that maps the continuous velocity values ​​of a particle to binary position flip probabilities. The curvature of this transfer function can be adaptively adjusted according to the iteration stage.

[0033] First, set the parameters of the improved particle swarm optimization algorithm, including particle population size, maximum number of iterations, upper and lower limits of inertia weight, learning factor, and upper and lower limits of time-varying transfer factor. Then, based on the fitness of each particle, perform the following iterative optimization process: In each iteration, calculate a linearly decreasing inertia weight value according to the current iteration number; using this inertia weight, the individual optimal position of the particle, and the global optimal position of the population, calculate the velocity of each particle according to the velocity update formula; substitute the calculated particle velocity into the piecewise adaptive transfer function to obtain the binary position flip probability; update the binary code of the particle according to the flip probability (i.e., update the cluster partitioning scheme); during the update process, perform compliance verification of the particle code with power system-related operational constraints to eliminate non-compliant schemes; when the preset maximum number of iterations is reached or the global optimal fitness value has not been updated for a certain number of consecutive iterations, terminate the iteration, and take the cluster partitioning result corresponding to the current global optimal particle as the optimal cluster partitioning result of the regional power grid.

[0034] Furthermore, mutation operators can be introduced during the iteration process to randomly flip certain bits of the particle code with a preset low probability, thereby increasing population diversity and preventing the algorithm from converging to a local optimum too early.

[0035] This application improves the binary particle swarm optimization algorithm by adopting a linearly decreasing inertia weight strategy and a piecewise adaptive transfer function. This balances the global exploration capability and local development capability of the binary particle swarm optimization algorithm, avoids the problems of the binary particle swarm optimization algorithm being prone to getting trapped in local optima and slow convergence speed, and improves the solution accuracy of cluster partitioning.

[0036] See Figure 2 , Figure 2 This is a flowchart illustrating another embodiment of the cluster optimization method based on improved particle swarm optimization algorithm coupled with multi-objectives provided by the present invention, as shown below. Figure 2 As shown, this embodiment includes steps 110 to 130, and the specific steps are as follows: Step 110: Based on the operating power and node relationship data of each source load node, obtain the electrical coupling tightness index of each node cluster; In this embodiment, the electrical coupling tightness index is used to quantify the degree of electrical connection between source and load nodes within a node cluster. A higher electrical coupling tightness index indicates a stronger electrical connection within the cluster. Node relationship data refers to data characterizing the electrical connection relationships between source and load nodes, and may include the electrical distance between nodes and the transmission loss coefficient. The electrical distance can be determined based on the per-unit value of the line impedance and the path length; the transmission loss coefficient can be obtained from the operating parameters of the transmission or distribution nodes.

[0037] For each node cluster, obtain the electrical distance and transmission loss coefficient between any two nodes in the cluster, as well as the total active power of the two nodes; then, calculate the electrical coupling tightness index of the cluster based on the following formula (1). For the electrical coupling tightness of cluster g; This represents the total number of nodes in cluster g; This represents the electrical distance between node i and node j; This represents the transmission loss coefficient between node i and node j; This represents the total active power of node i; This represents the total active power of node j.

[0038] (1) Step 120: Based on the complementarity of adjustment capabilities, temporal complementarity, and type complementarity of each node cluster, obtain the complementarity index of each node cluster. In this embodiment, the complementarity of adjustment capabilities is an indicator used to measure the degree of complementarity and matching between the adjustable power output capabilities of generation nodes and the adjustable load capabilities of load nodes within a node cluster. The temporal complementarity is an indicator used to measure the degree of negative correlation between the time-series curves of power generation output and the time-sharing curves of load within a node cluster. The type complementarity is an indicator used to measure the diversity of source and load node types within a node cluster; the richer the types, the stronger the complementarity.

[0039] For each node cluster, the complementarity of adjustment capability, temporal complementarity, and type complementarity of the cluster are calculated respectively. Then, the three are multiplied by their corresponding normalized weight coefficients and summed to obtain the complementarity index of the cluster, as shown in the following formula (2), where, This indicates the functional complementarity of cluster g; , and These represent the complementarity of the adjustment capabilities, the complementarity of the timing, and the complementarity of the types of cluster g, respectively. , and These are the corresponding normalized weighting coefficients. .

[0040] (2) (3) (4) (5) The complementarity of adjustment capabilities can be calculated based on the normalized standard deviation of the adjustable output limits of the power generation nodes and the adjustable load power of the load nodes within the cluster, as shown in formula (3) above. Wherein, and These are the set of power generation nodes and the set of load nodes within cluster g, respectively. and These are the upper and lower limits of the adjustable output of the i-th power generation node in the cluster, respectively; The average adjustable output range of all power generation nodes within the cluster; Let be the adjustable load power of the j-th load node in the cluster; Average adjustable load power of all load nodes within the cluster.

[0041] The temporal complementarity can be obtained by normalizing the negative correlation coefficient between the power generation time-series curve and the load time-sharing load curve within the cluster, as shown in formula (4) above. The total output time-series curve is formed by superimposing all power generation nodes within cluster g. The time-sharing load curve is formed by superimposing all load nodes within cluster g; The covariance function of a time series characterizes the degree of correlation between power generation and load time series changes; The variance function of the time series characterizes the temporal fluctuation amplitude of output and load.

[0042] Type complementarity can be obtained from the entropy calculation of the unique thermal coding information of the power generation type and load type within the cluster, as shown in the above formula (5). Wherein, m is the total number of node types in the entire domain, including thermal power, hydropower, wind power and photovoltaic power generation, as well as coding types such as residential, commercial, industrial and energy storage loads; The proportion of nodes of type k within cluster g is obtained by statistical analysis using one-hot encoding of node types, satisfying the following condition. .

[0043] Step 130: Perform a weighted analysis on the electrical coupling tightness index, complementarity index, and load rate root mean square error of each node cluster to obtain the multi-objective comprehensive performance index of each node cluster.

[0044] For each node cluster, the electrical coupling tightness index, complementarity index, and load rate mean square error of the cluster obtained above are multiplied by their respective weighting coefficients and then summed to obtain the multi-objective comprehensive performance index of the cluster. As shown in formula (6). Wherein, The fitness value represents the overall performance index of global multi-objectives, and the higher the value, the better the cluster partitioning result. Indicates the total number of clusters; , and They represent , and The corresponding weighting coefficients, ; This represents the mean squared error of the load rate within cluster g, calculated based on the load rate of the distribution node area, and is used to balance the cluster size and smooth load fluctuations.

[0045] (6) This embodiment organically integrates multiple dimensions of indicators such as electrical coupling tightness, complementary adjustment capabilities, timing complementarity, type complementarity, and load balancing, achieving a comprehensive and scientific quantitative evaluation of cluster partitioning schemes and providing accurate fitness calculation basis for subsequent optimization.

[0046] In another embodiment of the cluster optimization method based on improved particle swarm optimization algorithm and multi-objective coupling provided by the present invention, step 200 specifically includes: Step 210: Determine that a cluster partitioning result of the regional power grid is a binary encoded string, and that a source-load node is a binary encoded unit; Step 220: Based on the total number of source and load nodes in the regional power grid, determine the dimension of the binary encoded string; each dimension corresponds to the cluster affiliation status of a source and load node. Step 230: Under the relevant operational constraints of the regional power grid, adjust the cluster affiliation status of the source and load nodes to obtain the binary encoded string corresponding to each cluster partitioning result of the regional power grid; Step 240: Determine each of the binary encoded strings as a particle to obtain the particle population corresponding to the regional power grid.

[0047] In this embodiment, the binary encoded string refers to a sequence of 0s and 1s used to digitally represent a cluster partitioning scheme. Each binary encoded unit corresponds to a source-load node, and the value of the encoded unit represents the cluster to which the node belongs. Relevant operational constraints refer to the physical constraints that the regional power grid must meet in actual operation. These constraints may include transmission direction constraints, transmission power constraints, rated transmission capacity constraints, and regional distribution capacity constraints, etc., to ensure that the generated cluster partitioning scheme is physically feasible.

[0048] First, the cluster partitioning result of the regional power grid is represented as a binary encoded string, with one source-load node corresponding to one binary encoded unit. Specifically, the total number of all source-load nodes in the regional power grid that need to participate in cluster partitioning is counted, and this total number is determined as the dimension (i.e., length) of the binary encoded string. For example, if the regional power grid contains 100 source-load nodes, the binary encoded string of each particle is 100 bits, with each bit corresponding to a specific source-load node. The value of the i-th bit in the encoded string (0 or 1) indicates which cluster the i-th source-load node is assigned to. When the number of clusters to be partitioned is greater than two, a combination of multiple binary codes can be used to represent different cluster labels, for example, using 00, 01, 10, and 11 to correspond to four different clusters; alternatively, a multiple binary partitioning method can be used, that is, each time the algorithm is run, the nodes are divided into two clusters, and then each sub-cluster is recursively partitioned until the preset number of clusters is reached.

[0049] After determining the coding dimensions and coding rules, and under the relevant operational constraints of the regional power grid, the cluster affiliation status of source and load nodes is adjusted to obtain the binary coding string corresponding to each cluster partitioning result. Specifically, when generating or adjusting the binary coding string, the power system operation constraints are verified for each cluster partitioning scheme. The verification includes: whether the two ends of the transmission line branch maintain only a single fixed transmission direction, prohibiting simultaneous bidirectional power transmission; whether the distribution line follows the preset planned power supply direction, prohibiting reverse power transmission; whether the line transmission power is within the allowable operating range, with the transmission power of the transmission line between the preset minimum and maximum transmission power, and the lower limit of the transmission power of the distribution line being 0, and not exceeding the preset maximum transmission power of the line; whether the absolute value of the actual incoming power and the absolute value of the outgoing power of any transmission node exceed the upper limit of the rated transmission capacity of that node; and whether the total load power in the region is less than the predetermined total distribution capacity within the region. If any of the above constraints are not met, the partitioning scheme corresponding to the coding string is determined to be non-compliant, and the scheme needs to be removed or the coding needs to be corrected until all constraints are deemed compliant before the coding string can be included in the valid population set.

[0050] Finally, each binary encoded string that passes the compliance check is designated as a particle, and the set of all compliant particles constitutes a particle population. The size of the particle population can be set according to actual needs. In one implementation, to ensure population diversity, multiple initialization strategies can be used to generate the initial particle population: some particles are generated using a completely random initialization method, while others are generated using a heuristic rule-guided initialization method. For example, the initial encoding of some particles can be set based on the preliminary clustering results of electrical distance, allowing the algorithm to start searching from a more reasonable initial solution space, which helps to improve the convergence speed.

[0051] This embodiment achieves compliance and analyzability of the cluster construction scheme by encoding the cluster partitioning scheme into a binary string and embedding it with power system operation constraint verification. It ensures that all schemes participating in the optimization meet the physical constraints of actual power grid operation, avoids the problem of infeasible optimization results, and provides a compliant and effective initial population for subsequent iterative optimization.

[0052] In another embodiment of the cluster optimization method based on improved particle swarm optimization algorithm and multi-objective coupling provided by the present invention, step 300 specifically includes: Step 310: Determine the target cluster contained in the cluster partitioning result corresponding to each particle; Step 320: Determine the fitness of each particle based on the multi-objective comprehensive performance index of the target cluster.

[0053] For a binary encoded string corresponding to a particle, firstly, it is decoded according to a preset encoding rule to parse out the cluster partitioning result represented by the encoded string, and then all target clusters included in the partitioning result are determined, as well as the specific source payload nodes contained in each target cluster. Specifically, each bit of the particle's encoding is traversed, and the source payload node corresponding to each bit is assigned to the corresponding cluster according to the value of each bit, finally obtaining the complete partitioning scheme corresponding to the particle, which includes M target clusters, and each cluster contains several source payload nodes.

[0054] Then, for each target cluster obtained from decoding, the multi-objective comprehensive performance index of that target cluster is calculated according to the corresponding method described above. Finally, the multi-objective comprehensive performance indices of all target clusters are summed and then averaged to obtain the fitness value of the particle. The purpose of this solution is that the partitioning quality of any cluster will affect the overall fitness of the particle, thereby driving the algorithm to take into account the partitioning effect of all clusters during the optimization process, rather than optimizing only one or a few clusters.

[0055] This embodiment defines the fitness of particles to ensure that the particle fitness can fully reflect the quality of the entire partitioning scheme represented by the particle. A decrease in the quality of any cluster partitioning will lower the overall evaluation of the particle, thereby driving the algorithm to improve the partitioning quality of all clusters in a balanced way during the optimization process, and achieving the globally optimal cluster construction.

[0056] In another embodiment of the cluster optimization method based on improved particle swarm optimization algorithm and multi-objective coupling provided by the present invention, step 400 specifically includes: Step 410: Obtain the iteration termination condition and determine the current iteration data of the particle population based on the fitness of each particle in the current iteration; Step 420: When the current iteration data does not meet the iteration termination condition, determine the current particle search speed based on the current iteration data and the linear decreasing inertia weight strategy of the improved particle swarm algorithm. Step 430: Based on the piecewise adaptive transfer function of the improved particle swarm algorithm, convert the current particle search velocity into a particle adjustment probability; Step 440: Based on the particle adjustment probability, perform iterative optimization on the particle population to obtain the updated iterative data corresponding to the current iterative data; Step 450: When the updated iteration data meets the iteration termination condition, determine the optimal cluster division result of the regional power grid based on the updated iteration data; Step 460: When the updated iteration data does not meet the iteration termination condition, the updated iteration data is used as the current iteration data, and the process returns to the step of determining the current particle search speed based on the current iteration data and the linear decreasing inertia weight strategy of the improved particle swarm algorithm when the current iteration data does not meet the iteration termination condition.

[0057] In this embodiment, the iteration termination condition refers to the preset conditions used to determine whether the algorithm stops iterating, including reaching the preset maximum number of iterations, or the global optimal fitness value of the population not being updated for a number of consecutive times (e.g., 5 to 10 times).

[0058] First, a preset iteration termination condition is obtained, and the current iteration data of the particle population is determined based on the fitness of each particle in the current iteration. At the beginning of each iteration, it is first determined whether the preset iteration termination condition is met; if not, the position code and fitness value of each particle in the current iteration are obtained, and the individual optimal position of each particle (i.e., the position corresponding to the highest fitness reached by the particle in history) and the global optimal position of the population (i.e., the position corresponding to the highest fitness reached by all particles in history) are updated accordingly. The above information together constitutes the current iteration data of the particle population.

[0059] When the current iteration data does not meet the iteration termination condition, the current particle search velocity is determined based on the current iteration data and the linearly decreasing inertia weight strategy of the improved particle swarm optimization algorithm. Specifically, firstly, the inertia weight value of the current iteration round is calculated according to the linearly decreasing inertia weight strategy. As shown in formula (10), where, This represents the inertia weight for the current iteration round; and These are the preset upper and lower limits of the inertia weight, respectively; t is the current iteration number; The maximum number of iterations is preset. As can be seen from formula (10), in the early stages of iteration, when t is small, near The particles maintain a large inertia, which is beneficial for global exploration in a vast search space; in the later stages of iteration, t approaches... hour, near The reduced inertia of particle motion facilitates a more precise search near the optimal solution.

[0060] (7) (8) (9) (10) Then, based on the current position, individual optimal position, and global optimal position of the particles in the current iteration data, combined with the calculated inertial weights and the preset learning factor, the velocity of each particle is calculated according to the standard particle swarm velocity update formula, as shown in formula (7). For the first The velocity of the i-th particle in the d-th dimension in the next iteration; For dynamic inertia weights; Let be the particle velocity in the t-th iteration; and All are learning factors; and All are random numbers uniformly distributed between 0 and 1; The optimal position for each individual particle; The optimal position for the entire population; Let be the position of the particle in the t-th iteration.

[0061] After obtaining the particle search velocity, the current particle search velocity is converted into a particle adjustment probability according to the piecewise adaptive transfer function. Specifically, the calculated particle motion velocity is substituted into the piecewise adaptive transfer function, which maps continuous velocity values ​​to binary position flip probabilities between 0 and 1, as shown in formula (8), where, This represents the probability of particle position flipping. and All are adaptive adjustment coefficients, which change linearly with the number of iterations and are limited to a certain range; this piecewise adaptive transfer function is used to map continuous velocity quantities into binary flip probabilities in the range of 0 to 1. According to formula (8), the curvature of the piecewise adaptive transfer function can be adaptively adjusted according to the iteration stage. In the early stage of iteration, the function curvature is relatively gentle, and the velocity is mapped to probability in a relatively mild way, which is conducive to maintaining population diversity; in the later stage of iteration, the function curvature tends to be steep, which is conducive to accelerating convergence.

[0062] After obtaining the particle adjustment probability, the particle population is iteratively optimized based on this probability to obtain the updated iteration data corresponding to the current iteration data. For each bit in the binary code of each particle, a random number between 0 and 1 is generated, and this random number is compared with the calculated flip probability of that bit. If the random number is less than the flip probability, the code value of that bit is flipped; if the random number is greater than or equal to the flip probability, the code value of that bit remains unchanged. After performing the above operations on all particles, the updated particle population is obtained. After completing the code update, compliance verification of the updated particle codes is also required.

[0063] After obtaining the updated iteration data, it is determined again whether the updated iteration data meets the iteration termination condition. If it does, the iteration stops, and the optimal cluster partitioning result of the regional power grid is determined based on the decoding of the binary code string corresponding to the current global best particle. If the iteration termination condition is not met, the updated iteration data is used as the current iteration data, and the process returns to step S420 above to enter the next iteration loop.

[0064] This embodiment achieves a dynamic balance between the algorithm's global exploration capability and local exploitation capability through the synergistic effect of a linearly decreasing inertia weight strategy and a piecewise adaptive transfer function. The linearly decreasing inertia weight strategy solves the problem of rigid search strategies caused by fixed inertia weights in existing algorithms, while the piecewise adaptive transfer function solves the problem of the limited mapping method of existing fixed transfer functions. The combination of the two enables the improved particle swarm optimization algorithm provided in this application to have strong global exploration capability in the early stage of iteration to discover high-quality solution regions, and strong local exploitation capability in the later stage of iteration to accurately lock the optimal solution. This effectively avoids the problems of existing particle swarm optimization algorithms being prone to getting trapped in local optima and slow convergence speed, and improves the solution accuracy of cluster partitioning.

[0065] The following describes the cluster optimization system based on the improved particle swarm optimization algorithm and multi-objective coupling provided by the present invention. The cluster optimization system based on the improved particle swarm optimization algorithm and multi-objective coupling described below can be referred to in correspondence with the cluster optimization method based on the improved particle swarm optimization algorithm and multi-objective coupling described above.

[0066] Please refer to Figure 3 The present invention also provides a cluster optimization system based on an improved particle swarm optimization algorithm coupled with multi-objectives, comprising: The multi-objective comprehensive performance index determination module 301 is used to perform weighted analysis on the operating data of each source load node, the complementarity of each node cluster, and the load rate root mean square error of each node cluster to obtain the multi-objective comprehensive performance index of each node cluster. The particle population construction module 302 is used to determine that a cluster partitioning result of the regional power grid is a particle, and to obtain the particle population corresponding to the regional power grid. The particle fitness determination module 303 is used to determine the fitness of each particle based on the multi-objective comprehensive performance index. The particle swarm iterative optimization module 304 is used to iteratively optimize the particle swarm based on the linear decreasing inertia weight strategy and piecewise adaptive transfer function of the improved particle swarm algorithm, as well as the fitness of each particle, to obtain the optimal cluster partitioning result of the regional power grid.

[0067] Optionally, the multi-objective comprehensive performance index determination module includes: The electrical coupling tightness index determination unit is used to obtain the electrical coupling tightness index of each node cluster based on the operating power of each source load node and the node relationship data. The complementarity index determination unit is used to obtain the complementarity index of each node cluster based on the complementarity of the adjustment capability, the complementarity of the timing, and the complementarity of the type of each node cluster. The multi-objective comprehensive performance index determination unit is used to perform weighted analysis on the electrical coupling tightness index, complementarity index, and load rate root mean square error of each node cluster to obtain the multi-objective comprehensive performance index of each node cluster.

[0068] Optionally, the particle swarm construction module includes: A binary encoded string determination unit is used to determine that a cluster division result of the regional power grid is a binary encoded string, and a source-load node is a binary encoded unit; The encoding dimension determination unit is used to determine the dimension of the binary encoded string based on the total number of source and load nodes in the regional power grid; each dimension encoding corresponds to the cluster affiliation status of a source and load node. The binary encoded string adjustment unit is used to adjust the cluster affiliation status of the source load nodes under the relevant operational constraints of the regional power grid, so as to obtain the binary encoded string corresponding to each cluster partitioning result of the regional power grid. The particle swarm construction unit is used to determine that each of the binary encoded strings is a particle, thereby obtaining the particle swarm corresponding to the regional power grid.

[0069] Optionally, the particle fitness determination module includes: The target cluster determination unit is used to determine the target cluster contained in the cluster partitioning result corresponding to each particle. The particle fitness determination unit is used to determine the fitness of each particle based on the multi-objective comprehensive performance index of the target cluster.

[0070] Optionally, the particle population iterative optimization module includes: The current iteration data determination unit is used to obtain the iteration termination condition and determine the current iteration data of the particle population based on the fitness of each particle in the current iteration. The current particle search velocity determination unit is used to determine the current particle search velocity based on the current iteration data and the linear decreasing inertia weight strategy of the improved particle swarm algorithm when the current iteration data does not meet the iteration termination condition. The particle adjustment probability determination unit is used to convert the current particle search velocity into a particle adjustment probability according to the piecewise adaptive transfer function of the improved particle swarm algorithm. The iterative data update unit is used to iteratively optimize the particle population based on the particle adjustment probability to obtain the updated iterative data corresponding to the current iterative data; The optimal cluster partitioning result determination unit is used to determine the optimal cluster partitioning result of the regional power grid based on the updated iteration data when the updated iteration data meets the iteration termination condition. The loop unit is used to take the updated iteration data as the current iteration data when the updated iteration data does not meet the iteration termination condition, and return to execute the step of determining the current particle search speed based on the current iteration data and the linearly decreasing inertia weight strategy of the improved particle swarm algorithm when the current iteration data does not meet the iteration termination condition.

[0071] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A cluster optimization method based on an improved particle swarm optimization algorithm coupled with multi-objective coupling, characterized in that, Applied to regional power grids; The regional power grid includes multiple source-load nodes; The node cluster consists of the source payload nodes; the method includes: A weighted analysis is performed on the operating data of each source load node, the complementarity of each node cluster, and the root mean square error of the load rate of each node cluster to obtain the multi-objective comprehensive performance index of each node cluster. By determining that a cluster partitioning result of the regional power grid is a particle, the particle population corresponding to the regional power grid is obtained. Based on the aforementioned multi-objective comprehensive performance index, the fitness of each particle is determined; Based on the linearly decreasing inertial weight strategy and piecewise adaptive transfer function of the improved particle swarm optimization algorithm, as well as the fitness of each particle, the particle population is iteratively optimized to obtain the optimal cluster partitioning result of the regional power grid.

2. The cluster optimization method based on improved particle swarm optimization algorithm and multi-objective coupling as described in claim 1, characterized in that, The weighted analysis of the operating data of each source load node, the complementarity of each node cluster, and the root mean square error of the load rate of each node cluster yields the following multi-objective comprehensive performance indicators for each node cluster: Based on the operating power and node relationship data of each source load node, the electrical coupling tightness index of each node cluster is obtained. Based on the complementarity of adjustment capabilities, temporal complementarity, and type complementarity of each node cluster, the complementarity index of each node cluster is obtained. A weighted analysis was performed on the electrical coupling tightness index, complementarity index, and load rate root mean square error of each node cluster to obtain the multi-objective comprehensive performance index of each node cluster.

3. The cluster optimization method based on improved particle swarm optimization algorithm and multi-objective coupling as described in claim 1, characterized in that, The step of determining a cluster partitioning result of the regional power grid as a particle, and obtaining the particle population corresponding to the regional power grid, includes: The result of a cluster partitioning of the regional power grid is determined as a binary encoded string, and a source-load node is a binary encoded unit; Based on the total number of source and load nodes in the regional power grid, the dimension of the binary encoded string is determined; each dimension corresponds to the cluster affiliation status of a source and load node. Under the relevant operational constraints of the regional power grid, the cluster affiliation status of the source and load nodes is adjusted to obtain the binary encoded string corresponding to each cluster partitioning result of the regional power grid; Each binary encoded string is identified as a particle, thus obtaining the particle population corresponding to the regional power grid.

4. The cluster optimization method based on improved particle swarm optimization algorithm and multi-objective coupling as described in claim 1, characterized in that, The determination of the fitness of each particle based on the multi-objective comprehensive performance index includes: Determine the target cluster contained in the cluster partitioning result corresponding to each particle; The fitness of each particle is determined based on the multi-objective comprehensive performance index of the target cluster.

5. The cluster optimization method based on improved particle swarm optimization algorithm and multi-objective coupling as described in claim 1, characterized in that, The step of iteratively optimizing the particle swarm based on the linearly decreasing inertia weight strategy and piecewise adaptive transfer function of the improved particle swarm algorithm, as well as the fitness of each particle, to obtain the optimal cluster partitioning result of the regional power grid includes: Obtain the iteration termination condition, and determine the current iteration data of the particle population based on the fitness of each particle in the current iteration; When the current iteration data does not meet the iteration termination condition, the current particle search speed is determined based on the current iteration data and the linearly decreasing inertial weight strategy of the improved particle swarm algorithm. Based on the piecewise adaptive transfer function of the improved particle swarm algorithm, the current particle search velocity is converted into a particle adjustment probability. Based on the particle adjustment probability, the particle population is iteratively optimized to obtain the updated iteration data corresponding to the current iteration data; When the updated iteration data meets the iteration termination condition, the optimal cluster partitioning result of the regional power grid is determined based on the updated iteration data; When the updated iteration data does not meet the iteration termination condition, the updated iteration data is used as the current iteration data, and the process returns to the step of determining the current particle search velocity based on the current iteration data and the linearly decreasing inertia weight strategy of the improved particle swarm algorithm when the current iteration data does not meet the iteration termination condition.

6. A cluster optimization system based on an improved particle swarm optimization algorithm coupled with multi-objective coupling, characterized in that, include: The multi-objective comprehensive performance index determination module is used to perform weighted analysis on the operating data of each source load node, the complementarity of each node cluster, and the load rate root mean square error of each node cluster to obtain the multi-objective comprehensive performance index of each node cluster. The particle population construction module is used to determine that a cluster partitioning result of the regional power grid is a particle, and to obtain the particle population corresponding to the regional power grid. The particle fitness determination module is used to determine the fitness of each particle based on the multi-objective comprehensive performance index. The particle swarm iterative optimization module is used to iteratively optimize the particle swarm based on the linear decreasing inertia weight strategy and piecewise adaptive transfer function of the improved particle swarm algorithm, as well as the fitness of each particle, to obtain the optimal cluster partitioning result of the regional power grid.

7. The cluster optimization system based on improved particle swarm optimization algorithm and multi-objective coupling as described in claim 6, characterized in that, The multi-objective comprehensive performance index determination module includes: The electrical coupling tightness index determination unit is used to obtain the electrical coupling tightness index of each node cluster based on the operating power of each source load node and the node relationship data. The complementarity index determination unit is used to obtain the complementarity index of each node cluster based on the complementarity of the adjustment capability, the complementarity of the timing, and the complementarity of the type of each node cluster. The multi-objective comprehensive performance index determination unit is used to perform weighted analysis on the electrical coupling tightness index, complementarity index, and load rate root mean square error of each node cluster to obtain the multi-objective comprehensive performance index of each node cluster.

8. The cluster optimization system based on improved particle swarm optimization algorithm and multi-objective coupling as described in claim 6, characterized in that, The particle swarm construction module includes: A binary encoded string determination unit is used to determine that a cluster division result of the regional power grid is a binary encoded string, and a source-load node is a binary encoded unit; The encoding dimension determination unit is used to determine the dimension of the binary encoded string based on the total number of source and load nodes in the regional power grid; each dimension encoding corresponds to the cluster affiliation status of a source and load node. The binary encoded string adjustment unit is used to adjust the cluster affiliation status of the source load nodes under the relevant operational constraints of the regional power grid, so as to obtain the binary encoded string corresponding to each cluster partitioning result of the regional power grid. The particle swarm construction unit is used to determine that each of the binary encoded strings is a particle, thereby obtaining the particle swarm corresponding to the regional power grid.

9. The cluster optimization system based on improved particle swarm optimization algorithm and multi-objective coupling as described in claim 6, characterized in that, The particle fitness determination module includes: The target cluster determination unit is used to determine the target cluster contained in the cluster partitioning result corresponding to each particle. The particle fitness determination unit is used to determine the fitness of each particle based on the multi-objective comprehensive performance index of the target cluster.

10. The cluster optimization system based on improved particle swarm optimization algorithm and multi-objective coupling as described in claim 6, characterized in that, The particle population iterative optimization module includes: The current iteration data determination unit is used to obtain the iteration termination condition and determine the current iteration data of the particle population based on the fitness of each particle in the current iteration. The current particle search velocity determination unit is used to determine the current particle search velocity based on the current iteration data and the linear decreasing inertia weight strategy of the improved particle swarm algorithm when the current iteration data does not meet the iteration termination condition. The particle adjustment probability determination unit is used to convert the current particle search velocity into a particle adjustment probability according to the piecewise adaptive transfer function of the improved particle swarm algorithm. The iterative data update unit is used to iteratively optimize the particle population based on the particle adjustment probability to obtain the updated iterative data corresponding to the current iterative data; The optimal cluster partitioning result determination unit is used to determine the optimal cluster partitioning result of the regional power grid based on the updated iteration data when the updated iteration data meets the iteration termination condition. The loop unit is used to take the updated iteration data as the current iteration data when the updated iteration data does not meet the iteration termination condition, and return to execute the step of determining the current particle search speed based on the current iteration data and the linearly decreasing inertia weight strategy of the improved particle swarm algorithm when the current iteration data does not meet the iteration termination condition.

Images