Variable grid group control optimization method, system and device based on cloud edge collaboration and medium

Abstract

The invention relates to the technical field of intelligent power distribution network control, in particular to a variable grid group control optimization method, system and device based on cloud edge collaboration and a medium. Collecting real-time operation data of the distributed resources and performing state information processing to generate structured state information reflecting a relationship between a physical state and a network structure; the topological similarity and the operation characteristic difference degree are integrated through a clustering algorithm, self-adaptive division is conducted on the structured state information, and a plurality of variable grid groups and topological boundary information are generated; taking each variable grid group as an independent game main body, and combining a revenue function and a constraint condition to construct an initial strategy set of each variable grid group; performing multiple rounds of game interaction and income evaluation through a game algorithm to obtain an optimal variable grid group strategy; edge-side collaborative scheduling and resource allocation are carried out based on the optimal strategy, and an edge control result is obtained; an edge control result is fed back to the cloud for global coordination and integration, and a consistency strategy is generated, issued and executed.

Description

Technical Field

[0001] This invention relates to the field of smart distribution network control technology, and in particular to a variable grid group control optimization method, system, device and medium based on cloud-edge collaboration. Background Technology

[0002] Against the backdrop of rapid development of new power infrastructure, various types of resources, such as distributed renewable energy, electric vehicles, and energy storage, are being connected to the distribution network on a large scale, driving the gradual evolution of distribution operation from centralized management and control to distributed collaborative management. In order to improve resource utilization efficiency and operational flexibility, operation and management concepts such as "grid-based operation," "regional autonomous collaboration," and "cloud-edge collaborative control" have been proposed in recent years.

[0003] Traditional power distribution network control methods often employ a strategy of dividing the network into multiple power grids based on geographical topology or administrative regions, and then configuring edge control units at each grid for local resource scheduling and status control. Simultaneously, a central monitoring node coordinates the entire network by issuing unified policies. In this structure, the edge side primarily handles local sensing and initial decision-making, while the cloud is responsible for policy formulation, optimization, and unified command distribution.

[0004] However, existing methods still have significant shortcomings in addressing the highly dynamic nature of distribution network structures, the strong coupling of resource behaviors, and the conflicts of objectives among multiple stakeholders. Most existing methods employ static partitioning to establish the grid structure, which is ill-suited to the dynamic evolution processes caused by frequent changes in operating conditions, such as load migration and power disturbances. This leads to a mismatch between the partitioning results and actual operational needs, reducing the effectiveness of control strategies. Furthermore, traditional methods lack effective multi-stakeholder coordination and optimization mechanisms; strategies are often formulated independently between different grids, making it difficult to achieve global optimization. Summary of the Invention

[0005] In view of the problems existing in the prior art, the inventors have proposed the present invention.

[0006] Therefore, the problem this invention aims to solve is how to address the difficulty in achieving dynamic grid partitioning based on operational status in existing power distribution network control technologies, and the poor overall system performance caused by the lack of effective coordination and optimization among multiple grids. By constructing a comprehensive similarity index that integrates network structure and operational status, adaptive grid partitioning based on current physical connections and real-time operational status is achieved. Furthermore, game theory methods are used to achieve coordination and optimization among multiple grid groups, thereby overcoming the shortcomings of traditional static partitioning methods that cannot respond to load changes and operational risks in real time.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0008] In a first aspect, embodiments of the present invention provide a variable grid group control optimization method based on cloud-edge collaboration, which includes collecting real-time operation data of distributed resources and processing state information to generate structured state information reflecting the relationship between physical state and network structure.

[0009] By combining topological similarity and operational feature differences through clustering algorithms, the structured state information is adaptively divided to generate multiple variable grid groups and topological boundary information.

[0010] Each variable grid group is treated as an independent game entity, and an initial strategy set for each variable grid group is constructed by combining the payoff function and constraints.

[0011] By using a game theory algorithm, multiple rounds of game interaction and payoff evaluation are conducted on the initial strategy set of each variable grid group, and the selection of the initial strategy of each variable grid group is dynamically optimized to obtain the optimal variable grid group strategy.

[0012] Based on the optimal variable mesh group strategy, edge-side collaborative scheduling and resource allocation are performed, and the operating status of the variable mesh group is adjusted in real time to obtain edge control results.

[0013] The edge control results are fed back to the cloud for global coordination and integration, generating a consistent policy and issuing it for execution.

[0014] As a preferred embodiment of the variable grid group control optimization method based on cloud-edge collaboration described in this invention, the real-time operating data includes: monitoring node voltage, current, active power, reactive power, frequency, status indicators, and communication delay data.

[0015] The state information processing includes missing value imputation and outlier correction processing, as well as feature scaling processing and structured encoding mapping.

[0016] As a preferred embodiment of the variable grid swarm control optimization method based on cloud-edge collaboration described in this invention, the adaptive partitioning includes at least topological feature similarity calculation and runtime feature difference calculation.

[0017] The topological feature similarity calculation is used to extract the physical connection relationships between monitoring nodes;

[0018] The calculation of operational feature differences is used to quantify the differences in operational status between nodes, and the resulting comprehensive similarity index is used for clustering.

[0019] As a preferred embodiment of the variable grid swarm control optimization method based on cloud-edge collaboration described in this invention, the comprehensive similarity index includes:

[0020] Based on the similarity calculation of the topological connection characteristics of the monitoring nodes, the cosine similarity method is used to evaluate the network structure relationship between the nodes.

[0021] Based on the difference in the operating characteristics of monitoring nodes, the Euclidean distance method is used to quantify the differences in the operating status between nodes.

[0022] By fusing the similarity and difference using weighted coefficients, a comprehensive index for cluster analysis is constructed. As a preferred embodiment of the cloud-edge collaborative variable grid swarm control optimization method of this invention, the construction of the initial strategy set for each variable grid swarm includes:

[0023] Identify the resource type, status information, and network topology characteristics of each variable grid group, and determine the control objectives and resource constraints of each grid group;

[0024] Based on game theory, a payoff function model is constructed to quantify the expected payoffs of each variable grid group under different strategy choices.

[0025] An optimization algorithm is used to calculate the feasible policy space of each variable grid group under the current state, and an initial policy set is generated.

[0026] As a preferred embodiment of the variable grid swarm control optimization method based on cloud-edge collaboration described in this invention, the game theory algorithm includes:

[0027] A game model is established based on the initial strategy set of each variable grid group, and game rules and interaction mechanisms are set.

[0028] In each round of the game, the strategy choices and payoff performance of each variable grid group are evaluated, and the payoff results of each grid group are calculated and fed back.

[0029] The strategy selection of each variable grid group is dynamically adjusted based on the game results and payoff feedback, and the optimal equilibrium strategy is found through multiple rounds of iteration.

[0030] As a preferred embodiment of the variable grid swarm control optimization method based on cloud-edge collaboration described in this invention, the edge-side collaborative scheduling includes:

[0031] Based on the optimal variable grid group strategy, a time-series coordination and distributed allocation algorithm is used to coordinate the scheduling of each variable grid group.

[0032] Based on the optimized resource allocation scheme, a distributed edge control strategy is adopted to dynamically adjust the operation control parameters of each variable grid group;

[0033] The global coordination and integration includes transmitting edge control results to the cloud via communication protocols and generating a globally consistent control strategy using a centralized optimization algorithm.

[0034] Secondly, embodiments of the present invention provide a variable grid group control and optimization system based on cloud-edge collaboration, which includes a state awareness module for collecting real-time operating data of distributed resources and processing state information to generate structured state information that reflects the relationship between physical state and network structure.

[0035] The grid partitioning module is used to adaptively partition the structured state information by combining topological similarity and operational feature differences through a clustering algorithm, generating multiple variable grid groups and topological boundary information;

[0036] The strategy modeling module is used to treat each variable grid group as an independent game subject and, in combination with the payoff function and constraints, construct the initial strategy set for each variable grid group.

[0037] The game optimization module is used to perform multiple rounds of game interaction and payoff evaluation on the initial strategy set of each variable grid group through game algorithm, dynamically optimize the selection of the initial strategy of each variable grid group, and obtain the optimal variable grid group strategy.

[0038] The edge scheduling module is used to perform edge-side collaborative scheduling and resource allocation based on the optimal variable grid group strategy, adjust the running status of the variable grid group in real time, and obtain edge control results.

[0039] The consistency control module is used to feed back the edge control results to the cloud for global coordination and integration, generate consistency policies, and issue them for execution.

[0040] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, they implement the steps of the variable grid swarm control optimization method based on cloud-edge collaboration as described in the first aspect of the present invention.

[0041] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of the variable grid swarm control optimization method based on cloud-edge collaboration as described in the first aspect of the present invention.

[0042] The beneficial effects of this invention are as follows: By constructing a comprehensive similarity index that integrates network structure and operational status, this invention achieves adaptive mesh partitioning based on current physical connections and real-time operational status. The partitioning method not only considers the static topological characteristics of the power distribution network but also dynamically integrates the changing trends of operational control parameters such as voltage, current, and power of monitoring nodes. This overcomes the shortcomings of traditional static partitioning methods, which cannot respond to load changes and operational risks in real time, significantly improving the accuracy and adaptability of mesh partitioning.

[0043] This invention employs game theory to achieve coordinated optimization among multiple grid groups. Each variable grid group is treated as an independent decision-making entity, and the optimal equilibrium strategy is found through multiple rounds of game interaction. This effectively solves the problems of independent strategy formulation, conflicting objectives, and difficulty in achieving global optimum in traditional methods. The game theory algorithm fully considers the mutual influence and resource competition among grid groups, ensuring that the overall system performance is maximized while satisfying the interests of each individual group.

[0044] This invention achieves an organic combination of edge-side collaborative scheduling and cloud-based global optimization. Through a cloud-edge collaborative architecture, it fully leverages the real-time response capabilities of edge computing and the global optimization capabilities of cloud computing. The edge side is responsible for the rapid scheduling and real-time control of local resources, while the cloud is responsible for the unified formulation and coordination of global strategies. The two work together to form a hierarchical control system that ensures both real-time control and global optimality of decision-making.

[0045] This invention achieves reliable transmission of edge control results to the cloud via the MQTT communication protocol, ensuring the real-time performance and accuracy of control information. Simultaneously, a centralized optimization algorithm is employed to process and integrate globally coordinated input data, generating a consistent control strategy and issuing it for execution. This ensures coordinated control behavior among different grid groups and avoids global performance loss caused by local optimization.

[0046] This invention effectively addresses complex issues such as the highly dynamic structure of distribution networks, strong coupling of resource behaviors, and conflicts among multiple stakeholders, significantly improving the utilization efficiency of distributed resources and the operational flexibility of distribution networks. Through the organic combination of adaptive grid partitioning, game-theoretic coordination optimization, and cloud-edge collaborative control, it provides effective technical support for the intelligent operation of new power systems, possessing significant theoretical value and practical application significance. Attached Figure Description

[0047] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 The flowchart shows a variable grid swarm control optimization method based on cloud-edge collaboration.

[0049] Figure 2 A computer device diagram for a variable grid swarm control optimization method based on cloud-edge collaboration;

[0050] Figure 3 This is another flowchart of Example 2 of the variable grid group control optimization method based on cloud-edge collaboration. Detailed Implementation

[0051] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0052] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0053] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.

[0054] Example 1

[0055] Reference Figures 1-2 This is the first embodiment of the present invention, which provides a variable grid swarm control optimization method based on cloud-edge collaboration, including:

[0056] S100: Collects real-time operational data of distributed resources and processes state information to generate structured state information that reflects the relationship between physical state and network structure;

[0057] S200: By combining topological similarity and operational feature differences through clustering algorithms, the structured state information is adaptively divided to generate multiple variable grid groups and topological boundary information;

[0058] S300: Treat each variable grid group as an independent game entity, and construct an initial strategy set for each variable grid group by combining the payoff function and constraints;

[0059] S400: Through game theory algorithms, multiple rounds of game interaction and payoff evaluation are performed on the initial strategy set of each variable grid group to dynamically optimize the selection of the initial strategy of each variable grid group and obtain the optimal variable grid group strategy.

[0060] S500: Based on the optimal variable mesh group strategy, it performs edge-side collaborative scheduling and resource allocation, adjusts the running status of the variable mesh group in real time, and obtains edge control results;

[0061] S600: Feeds the edge control results back to the cloud for global coordination and integration, generates consistent policies, and issues them for execution.

[0062] It should be noted that the large-scale integration of various resources such as distributed renewable energy, electric vehicles, and energy storage into the distribution network has led to a gradual evolution of distribution network operation from centralized control to distributed collaborative management. Traditional methods, which establish grid structures using static partitioning, struggle to adapt to the dynamic evolution processes caused by frequent changes in operating conditions, such as load migration and power disturbances, resulting in a mismatch between the partitioning results and actual operational needs. Furthermore, the lack of an effective multi-agent coordination and optimization mechanism among grids means that strategy formulation is often independent, making it difficult to achieve global optimization and thus reducing the effectiveness of control strategies. Therefore, to address the aforementioned operational monitoring and coordinated control problems, this paper constructs a comprehensive similarity index integrating network structure and operating status through steps S100-S600. This enables adaptive grid partitioning based on current physical connections and real-time operational status, and game theory methods are used to achieve coordinated optimization among multiple grid groups.

[0063] Through steps S100-S600, this invention achieves adaptive grid partitioning based on current physical connectivity and real-time operational status. First, step S100 collects operational data such as voltage, current, and power from monitoring nodes and performs structured encoding based on the network topology, providing accurate status information for subsequent analysis. Step S200 performs cluster analysis based on comprehensive similarity indices, overcoming the limitation of traditional static partitioning methods in responding to load changes in real time. Steps S300-S400 treat each grid group as an independent game entity, finding the optimal equilibrium strategy through multi-round game interaction, resolving the problem of multi-entity objective conflict. Steps S500-S600 organically combine edge-side collaborative scheduling with cloud-based global optimization, ensuring the security and economy of system operation. The entire solution, through a cloud-edge collaborative architecture, achieves efficient management and optimized control of distributed resources.

[0064] Example 2

[0065] Reference Figures 2-3 This is the second embodiment of the present invention.

[0066] In this embodiment of the application, step S100 involves collecting real-time operational data of distributed resources and processing their status information to generate structured status information reflecting the relationship between physical status and network structure, including the following steps A1-A4:

[0067] A1: Real-time operating data includes: monitoring node voltage, current, active power, reactive power, frequency, status indicators, and communication delay data;

[0068] State information processing includes missing value imputation and outlier correction, as well as feature scaling and structured encoding mapping.

[0069] Specifically, voltage and current sensors are used to measure voltage and current data in real time at each monitoring node. Active and reactive power at each monitoring node are acquired using power measurement equipment. Frequency at each monitoring node is measured using a frequency meter. The operating status of each monitoring node is identified in real time by status monitoring equipment to determine its status information, such as normal operation, fault, or standby. Data transmission delays at each monitoring node are recorded using communication equipment, and after preliminary processing, a real-time operating dataset is formed. For example, a status sensing module is responsible for collecting this real-time operating data from various monitoring nodes in the power distribution network.

[0070] A2: The missing value imputation and outlier correction algorithms are used to clean the real-time running data and generate a unified dataset.

[0071] Specifically, missing value imputation algorithms are used to handle missing values ​​in the real-time operational data. By analyzing historical data from each monitoring node, appropriate interpolation methods, such as linear interpolation and Lagrange interpolation, are used to fill in missing values, ensuring the completeness of the real-time operational data. Outlier correction algorithms are used to detect and correct abnormal data. Statistical analysis methods, such as Z-score analysis or IQR, are used to identify outlier data points that exceed the normal range, and these points are corrected or removed. After cleaning, all the real-time operational data is integrated into a unified dataset.

[0072] It should be noted that the normal range is determined by statistical analysis of historical data or large sample data of monitoring nodes, such as calculating the mean ± 3σ or interquartile range, to reflect the reasonable fluctuation range of each parameter under typical operating conditions.

[0073] A3: The Min-Max normalization algorithm is used to scale the features of the unified dataset to generate a normalized running feature matrix. Combined with the physical connection topology of each monitoring node, a feature splicing strategy is used to fuse the normalized running features with the topology to generate combined data.

[0074] Specifically, the Min-Max normalization algorithm is used to scale the various operational features in the unified dataset. For each operational feature value, the minimum and maximum values ​​are calculated, and normalization is performed using the formula, that is, scaling the data to the range [0,1]. The expression is:

[0075]

[0076] Where, x ′ Let x represent the normalized unified data, min(x) represent the original data, and max(x) represent the minimum value of the running feature.

[0077] After normalization, the normalized running feature vectors are fused with the topology information using a feature concatenation strategy, taking into account the physical connection topology of each monitoring node. The topology information is represented by the connection relationships between monitoring nodes, typically in the form of an adjacency matrix or edge list.

[0078] A4: The combined data is mapped using a structured coding method to generate structured state information that reflects the relationship between physical state and network structure.

[0079] It should be noted that the normalized operational feature vector and the physical connection topology in the combined data are separated separately to extract the normalized operational features of each monitoring node and its corresponding adjacency information in the topology. For the normalized operational feature vector, vector encoding is used to retain the numerical features of each dimension. For the adjacency information of the monitoring nodes, adjacency vectors or adjacency matrices are constructed to represent the connection state of the monitoring nodes, and the physical connection information is encoded into a structured form, such as by constructing a structural feature representation through monitoring node degree, the number of adjacent monitoring nodes, and connection weights. The encoded results of the normalized operational feature vector and the encoded results of the structural feature representation are concatenated to generate structured state information.

[0080] In an optional implementation, the state information processing in step S100 may further include data denoising processing based on wavelet transform, which decomposes the original signal into different frequency components through wavelet decomposition, removes high-frequency noise components, retains useful low-frequency information, and further improves data quality.

[0081] In another optional implementation, a data quality assessment mechanism may be introduced in step S100. By calculating data integrity, consistency and accuracy indicators, the quality score of the collected real-time running data is given, and only data that reaches the preset quality threshold can enter the subsequent processing flow.

[0082] In this embodiment of the application, step S200 uses a clustering algorithm to integrate topological similarity and operational feature differences to adaptively partition the structured state information, generating multiple variable mesh groups and topological boundary information, including the following steps B1-B2:

[0083] B1: Adaptive partitioning includes at least topological feature similarity calculation and runtime feature difference calculation;

[0084] Topological feature similarity calculation is used to extract the physical connection relationships between monitoring nodes;

[0085] The runtime characteristic difference calculation is used to quantify the differences in runtime status between nodes, and the resulting comprehensive similarity index is used for clustering.

[0086] Normalized operational features and topological connectivity features are extracted from the structured state information to construct the input vector to be clustered. The similarity of topological features and the difference of operational features are calculated separately and then fused to obtain a comprehensive similarity index.

[0087] Specifically, normalized operational features and topological connectivity features of each monitoring node are extracted from the structured state information. Normalized operational feature vectors and topological connectivity feature vectors are constructed separately, and these two vectors are concatenated into the input vector for clustering. Then, similarity calculation methods are used to calculate the similarity between topological connectivity feature vectors and the difference between normalized operational feature vectors. The calculation of topological connectivity feature similarity can use, for example, the cosine similarity formula, with the expression:

[0088]

[0089] in, t represents the similarity of the topological connectivity features between monitoring node i and monitoring node j. i t represents the topological connectivity feature vector of monitoring node i. j Let ||t| represent the topological connectivity feature vector of monitoring node j. i || represents vector t i The Euclidean norm, ||t j || represents vector t j The Euclidean norm;

[0090] The normalized operational characteristic variance can be calculated using the Euclidean distance formula, which is expressed as follows:

[0091]

[0092] in, r represents the difference in normalized operational characteristics between monitoring node i and monitoring node j. i r represents the normalized running feature vector of monitoring node i. j Represents the normalized operational feature vector of monitoring node j, ||r i -r j || indicates r i vector and r j The Euclidean distance between vectors;

[0093] Then, the topological connectivity feature similarity and the normalized operational feature difference are fused according to weighted coefficients to construct a comprehensive similarity index, expressed as:

[0094]

[0095] Among them, M ij This represents the comprehensive similarity index between monitoring node i and monitoring node j, where α represents the fusion weight coefficient, and D...max This represents the maximum normalized operational characteristic difference value among all monitoring node pairs.

[0096] It should be noted that topology connectivity features refer to the connection status and network structure information between nodes in a distribution network. They reflect the power transmission path and the physical relationship between monitoring nodes, and are one of the basic data for power grid analysis and control.

[0097] B2: Based on the comprehensive similarity index, a clustering algorithm is used to classify the monitoring nodes, generate preliminary grid division results, perform boundary consistency analysis, identify and label the boundary monitoring nodes of each grid, and generate multiple variable grid groups and topological boundary information.

[0098] Specifically, based on the comprehensive similarity index, a similarity matrix is ​​constructed between each pair of monitoring nodes. Then, K-means clustering or spectral clustering algorithms are used to process the similarity matrix. All monitoring nodes are clustered and classified according to the index values ​​between each pair of monitoring nodes in the similarity matrix, resulting in preliminary grid division results. For example, the grid division module implements adaptive grid division based on the comprehensive similarity index.

[0099] Based on the initial grid division results, for each cluster set of monitoring nodes, the topological connectivity feature vectors of all monitoring nodes are extracted. The node numbers of all monitoring nodes with physical connections to the target monitoring node are sequentially traversed through the connectivity vectors. For each connected monitoring node, its cluster number is queried. If the cluster number of any connected monitoring node is inconsistent with the cluster number of the target monitoring node, it is determined that the target monitoring node has a cross-cluster connection. The topological connectivity relationships of all boundary monitoring nodes are collected and organized to generate the variable grid group and topological boundary information corresponding to each initial grid division result.

[0100] In an optional implementation, the clustering algorithm in step S200 can also employ a hierarchical clustering method, which dynamically determines the optimal number of clusters by constructing a clustering tree diagram, thereby avoiding the subjective influence of pre-setting the number of clusters.

[0101] In another optional implementation, step S200 may also introduce clustering effectiveness evaluation indicators, such as silhouette coefficient, Calinski-Harabasz index, etc., to evaluate and optimize different clustering results, ensuring the rationality and effectiveness of grid division.

[0102] In this embodiment of the application, step S300 treats each variable grid group as an independent game subject, and constructs an initial strategy set for each variable grid group by combining the payoff function and constraints, including the following steps C1-C2:

[0103] C1: Constructing the initial policy set for each variable mesh group includes:

[0104] Identify the resource type, status information, and network topology characteristics of each variable grid group, and determine the control objectives and resource constraints of each grid group;

[0105] Based on game theory, a payoff function model is constructed to quantify the expected payoffs of each variable grid group under different strategy choices.

[0106] An optimization algorithm is used to calculate the feasible policy space of each variable grid group under the current state, and an initial policy set is generated.

[0107] Specifically, for each variable grid cluster, the physical connection information of all monitoring nodes within the corresponding grid is first extracted based on the generated topological boundary information to form network topology features. The normalized operating features of each monitoring node are then extracted to form status information. For each monitoring node, the resource attributes it possesses are determined based on the type of connected equipment and its operating function, such as photovoltaic power generation, energy storage devices, adjustable loads, or ordinary loads. The nodes are then classified according to the controllability, adjustment capability, and response characteristics of the resources. At the same time, corresponding identification information is added to each type of resource node for unified management and use in the subsequent scheduling optimization process.

[0108] Based on the task objectives corresponding to each variable grid cluster, the control tasks to be completed are clearly defined, such as load reduction, power balancing, or minimizing operating costs. Based on the defined objectives, the current operating status and adjustability of various distributed resources within the variable grid cluster are statistically analyzed. Resource availability constraints are established based on information such as the rated capacity, current load level, and response speed of the resources. The operating boundary parameters of each monitoring node, including the safe operating range of indicators such as voltage, current, and frequency, are analyzed to construct state boundary constraints.

[0109] Further combining the physical connections and network topology between monitoring nodes, path accessibility and potential islanding risks are identified, forming topology feasibility constraints. After the constraints are determined, a revenue function related to resource allocation results is constructed based on the control objectives of the target task. The revenue function may include an operating cost function, a load reduction benefit function, or a comprehensive economic function, ultimately forming the revenue function and constraints.

[0110] C2: Combining the payoff function and constraints, an optimization algorithm is used to calculate the policy set of each variable grid group under the current state, and the initial policy set of each variable grid group is generated.

[0111] Specifically, for each variable grid cluster, based on the payoff function and constraints, and extracting the resource types, current normalized operating characteristics, and physical connection information of all monitoring nodes within the cluster, the variables to be optimized are defined, such as the output power, on / off status, or load adjustment ratio of various adjustable resources. The upper and lower limits of resource availability, the boundaries of monitoring node operating states, and network topology reachability are defined as specific variable constraints. An existing optimization algorithm, such as particle swarm optimization or genetic algorithm, is selected to initialize the variable combinations, and optimization parameters such as population size and number of iterations are set. Following the evolution rules of the optimization algorithm, the objective value is evaluated and constraints are checked for each variable combination, continuously iterating and updating the variable combinations. The optimal variable combination is selected as the initial strategy set for each corresponding variable grid cluster.

[0112] In an optional implementation, the strategy set construction in step S300 can also employ a multi-objective optimization method, simultaneously considering multiple objectives such as economy, reliability, and environmental protection, and obtaining a strategy set that balances the objectives through Pareto front analysis.

[0113] In another optional implementation, uncertainty analysis can be introduced in step S300 to consider uncertainties such as load forecasting errors and fluctuations in renewable energy output, and to construct a more adaptive set of strategies using robust optimization or stochastic optimization methods.

[0114] In this embodiment of the application, step S400 involves multiple rounds of game interaction and payoff evaluation of the initial strategy set of each variable grid group using a game theory algorithm to dynamically optimize the selection of the initial strategy of each variable grid group and obtain the optimal variable grid group strategy, including the following steps D1-D4:

[0115] D1: Game theory algorithms include:

[0116] A game model is established based on the initial strategy set of each variable grid group, and game rules and interaction mechanisms are set.

[0117] In each round of the game, the strategy choices and payoff performance of each variable grid group are evaluated, and the payoff results of each grid group are calculated and fed back.

[0118] The strategy selection of each variable grid group is dynamically adjusted based on the game results and payoff feedback, and the optimal equilibrium strategy is found through multiple rounds of iteration.

[0119] It should be noted that, based on the initial policy sets and corresponding payoff functions of each generated variable grid group, an initial policy is selected as the starting execution plan for each variable grid group from the initial policy set. The expected payoff value that the initial policy can obtain under the current conditions is calculated and allocated by substituting the current state of the variable grid group into the corresponding payoff function. Appropriate game theory algorithms are selected based on the interaction characteristics between the variable grid groups. For example, in scenarios with policy conflicts and resource competition, a non-cooperative Nash equilibrium game algorithm can be used; in scenarios with a cooperative operational objective, a cooperative alliance game algorithm can be used as the basic algorithm for multi-round game interaction. The number of game rounds and the interaction policy update mechanism are set, and the policy set that each variable grid group can choose in each round of the game is defined, including strategies such as minimum operating cost, maximum load response, and balanced supply and demand. Corresponding policy transition rules are also set to guide the policy iteration process.

[0120] D2: In each round of the game, evaluate the strategy choices and payoffs of each variable grid group, calculate and report the payoff results for each grid group.

[0121] It should be noted that in each round of the game, the payoff for each variable grid group is first calculated based on the current strategy set of each group. The payoff calculation for each variable grid group is evaluated based on the interaction results of the selected control strategy with the control strategies of other grid groups, combined with a pre-defined payoff function. For example, for a specific task, it may be necessary to calculate the load reduction benefit or the reduction in operating costs. Based on the calculated payoff results, the payoff results for each variable grid group are fed back. Each grid group adjusts its control strategy according to the feedback payoff results to ensure that the most advantageous variable grid group strategy is selected.

[0122] D3: Dynamically adjust the strategy selection of each variable grid group based on the game results, and perform multiple rounds of iterative optimization.

[0123] It should be noted that, based on the game outcome, the evaluation of the variable grid group strategy includes analyzing the current strategy choice and gains of each variable grid group, assessing relative advantage, stability, and long-term effects, and examining equilibrium in the game. Through a comprehensive evaluation of gains, risks, strategy stability, and adaptability, the effectiveness and sustainability of the current strategy are determined, thereby optimizing resource allocation. Each variable grid group dynamically adjusts its strategy choice based on the gap between its current gain and its target gain. When adjusting the variable grid group strategy, the strategy choices of other variable grid groups and their mutual influences must be considered. For example, if a variable grid group has a low gain, it will try to choose a more favorable strategy to increase its gain. The adjusted variable grid group strategy will be reflected in the next game round, and the gains will be re-evaluated based on the new strategy choice.

[0124] D4: The optimal variable grid group strategy is obtained through multiple rounds of game interaction and payoff evaluation.

[0125] It should be noted that each variable grid group calculates its payoff under its chosen current strategy. The payoff calculation is based on the impact of the strategy adopted by the grid group on resource utilization efficiency, task completion, and overall effectiveness. During evaluation, the payoff function is used to quantitatively analyze the strategy's effectiveness, while considering constraints such as resource limitations and time windows to ensure that the strategy selection optimizes resource allocation and task execution within the constraints. The evaluation result is the specific payoff value of each variable grid group under the current strategy, reflecting the actual effectiveness of the strategy and the degree to which the goal is achieved. If the payoff value of a certain grid group is lower than expected, it indicates that the current strategy has failed to achieve the optimal effect. Therefore, the strategy needs to be adjusted based on feedback information, selecting a more advantageous action. The adjusted strategy will be carried into the next round of the game for re-evaluation and optimization. This continues until the strategy selections of all variable grid groups reach a stable state or the payoff reaches the optimal level. Through continuous game playing and feedback, the optimal variable grid group strategy is finally obtained.

[0126] In an optional implementation, the game algorithm in step S400 can also adopt evolutionary game theory, which simulates the biological evolution process so that the strategies of each grid group continuously evolve during the game and eventually converge to an evolutionarily stable strategy.

[0127] In another optional implementation, step S400 may also introduce a machine learning method to train the policy selection model of each grid group through a reinforcement learning algorithm, so that it can autonomously learn the optimal policy in a complex environment.

[0128] In this embodiment of the application, step S500 involves edge-side collaborative scheduling and resource allocation based on the optimal variable mesh group strategy, adjusting the operating state of the variable mesh group in real time to obtain the edge control result, including the following steps E1-E2:

[0129] E1: Edge-side collaborative scheduling includes:

[0130] Based on the optimal variable grid group strategy, a time-series coordination and distributed allocation algorithm is used to coordinate the scheduling of each variable grid group.

[0131] Based on the optimized resource allocation scheme, a distributed edge control strategy is adopted to dynamically adjust the operation control parameters of each variable grid group;

[0132] Global coordination and integration includes transmitting edge control results to the cloud via communication protocols and generating a globally consistent control strategy using centralized optimization algorithms.

[0133] Specifically, based on the optimal variable grid cluster strategy, the resource requirements, task objectives, and operational status of each variable grid cluster are collected. Based on the urgency, dependencies, and available time windows of the tasks, the resource scheduling tasks of each variable grid cluster are sequentially arranged to ensure that tasks are rationally and orderly in the time dimension. During this process, task priorities are set, a task dependency graph is constructed, and the task order is dynamically adjusted based on game feedback results to achieve scheduling coordination and adaptability. Based on the determined task sequence, the resource allocation amount for each variable grid cluster in each time period is calculated using a distributed allocation algorithm, taking into account the resource capacity boundaries and task execution requirements of each variable grid cluster. For example, in scheduling tasks involving load reduction, the power reduction amount for each variable grid cluster in different time periods is calculated through multiple rounds of iterative optimization, satisfying both the overall scheduling objective and ensuring fair resource sharing and non-conflict among the variable grid clusters. Finally, an optimized resource scheduling scheme covering the entire scheduling cycle and satisfying global constraints is generated.

[0134] E2: Based on the optimized resource allocation scheme, a distributed edge control strategy is adopted. The operation control parameters and control behavior of each variable grid group are dynamically adjusted according to the optimal variable grid group strategy. Resource scheduling instructions are executed and the current operating status is updated to obtain the edge control result.

[0135] Specifically, based on the optimized resource allocation scheme, the latest resource scheduling instructions and related operational control parameters for each variable grid cluster are obtained. These operational control parameters include, but are not limited to, power output, load allocation, scheduling time, priority, and resource availability. A distributed edge control strategy is adopted, dynamically adjusting the operational control parameters of each variable grid cluster, such as power output and load allocation, according to the optimal variable grid cluster strategy. Modification of control behavior refers to adjusting the strategy to adapt to changing system states based on real-time operating conditions and resource demands; for example, adjusting power output to balance load or adjusting task priorities to respond to sudden demands. Based on the adjusted operational control parameters and control behavior, resource scheduling instructions are executed to coordinate task execution across the variable grid clusters. For example, for a specific load reduction task, it may be necessary to adjust the power allocation value of a particular variable grid cluster to ensure that the load reduction requirements of each variable grid cluster are met. Finally, the current operating status of each variable grid cluster is updated, and the adjustment results are fed back to obtain the edge control results.

[0136] In an optional implementation, the edge-side collaborative scheduling in step S500 can also employ a predictive control method, which, based on load forecasting and renewable energy output forecasting, plans resource scheduling schemes for multiple future time periods in advance, thereby improving the foresight and robustness of scheduling.

[0137] In another optional implementation, an adaptive control strategy can be introduced in step S500 to dynamically adjust the control parameters and control strategy according to the real-time changes in the system's operating state, so as to ensure good control performance under various operating conditions.

[0138] In this embodiment of the application, step S600 involves feeding the edge control results back to the cloud for global coordination and integration, generating a consistent policy, and issuing it for execution, including the following steps F1-F3:

[0139] F1: Transmits the edge control results of each variable grid group to the cloud via the MQTT communication protocol, and combines them with the optimal variable grid group strategy to form global coordinated input data.

[0140] It should be noted that, using the MQTT communication protocol, a message transmission structure is established based on the edge control results of each variable grid group, and corresponding topics are configured. Then, the edge control results of each variable grid group are transmitted to the cloud in message format, ensuring that the messages include necessary control data, such as the operating status of each variable grid group, resource allocation information, and the execution status of control commands. Upon receiving the edge control results in the cloud, the received control data is parsed and relevant information is integrated, combined with the optimal variable grid group strategy, to form global coordination input data. For example, if a variable grid group executes a load reduction command, the cloud will calculate the overall benefit of the load reduction and perform a comprehensive analysis of the execution status of other variable grid groups.

[0141] It should be noted that establishing the message transmission structure includes: first, defining a standardized message format containing fields such as grid group ID, control type, and control value; then, constructing a multi-level topic structure based on the grid group number and control task type to achieve message classification and targeted transmission; configuring edge monitoring nodes as MQTT clients to publish control results to the corresponding topics, with central monitoring nodes subscribing to receive them; and simultaneously setting communication quality and security mechanisms, such as TLS encryption and authentication, to ensure the real-time performance, reliability, and security of control information transmission.

[0142] F2: A centralized optimization algorithm is used to process and integrate the globally coordinated input data to generate a globally consistent control strategy.

[0143] It should be noted that the global coordination input data received from the cloud is collected and integrated into a unified global coordination input dataset. Based on a centralized optimization algorithm, various parameters in the dataset are analyzed, including information such as resource allocation, load scheduling, and operational status of each variable grid group. According to the global coordination input dataset, the objective function is defined as maximizing resource utilization efficiency or revenue, with constraints including grid group task completion requirements, resource capacity limits, and time-series scheduling constraints, ensuring that the optimization process conforms to the global coordination objective, such as optimal global resource allocation or minimization of operating costs. Through the centralized optimization algorithm, iterative calculations are repeatedly performed based on the objective function and constraints to find the optimal solution. During this process, the centralized optimization algorithm continuously adjusts the resource allocation of each variable grid group until the optimal effect requirements are met, ultimately generating a consistent global control strategy.

[0144] F3: The cloud distributes the generated consistency control strategy to each variable grid group through a distribution mechanism, uses a distributed execution platform to perform global scheduling and collaborative control of tasks, and monitors the execution effect in real time.

[0145] Specifically, based on the generated globally consistent control strategy, the cloud selects appropriate transmission methods and timing arrangements according to the task requirements, resource capabilities, and priorities of the variable grid clusters, effectively distributing the global control strategy to each variable grid cluster and ensuring timely feedback on execution status. Each variable grid cluster, upon receiving the distributed control strategy, parses its content, including load scheduling, resource allocation, and operational control requirements. Utilizing the distributed execution platform, each variable grid cluster begins executing its local tasks, adjusting its operational control parameters and execution behavior according to the control strategy. To ensure global coordination, each variable grid cluster provides real-time feedback on its operational status and execution results to the cloud. The cloud receives and aggregates the feedback data, performs real-time monitoring and analysis, and evaluates whether the execution effect of the control strategy meets the expected goals. For example, by comparing actual operational data with target results, the accuracy of load allocation and resource utilization efficiency are checked. Based on the monitoring results, the strategy is adjusted as necessary to ensure the smooth execution of global scheduling and collaborative control tasks.

[0146] In an optional implementation, the global coordination and integration in step S600 can also adopt a hierarchical optimization method, which decomposes the global optimization problem into multiple sub-problems and achieves global optimality through coordination between upper and lower levels, thereby reducing computational complexity.

[0147] In another optional implementation, blockchain technology can be introduced in step S600 to establish a decentralized trust mechanism, ensuring the security and credibility of information interaction between grid groups and preventing malicious attacks and data tampering.

[0148] It should be noted that the above-mentioned internal and external wall temperature stress relationship model introduces a thermo-mechanical-thickness coupling term, which avoids the limitations of a single heat transfer model. Key parameters such as thermal conductivity and elastic modulus are adopted as temperature correlation functions rather than empirical fixed values. At the same time, the degradation process is quantified, and dynamic characteristics are considered, providing an accurate basis for judgment in subsequent steps.

[0149] For example, in this application embodiment, operational data from multiple distribution network monitoring nodes are collected, totaling approximately 5,000 data points including parameters such as voltage, current, and power, ensuring coverage of various operating conditions, such as peak load, valley load, and typical scenarios with high and low renewable energy generation.

[0150] The operational status of each monitoring node was labeled using a data annotation tool, generating a training set with 3000 labeled samples and a validation set with 500 samples. Status information processing was then performed to increase the diversity of the dataset. Normalization scaled the original data to a uniform range, and feature concatenation fused operational features with the topological structure, ultimately generating a total of 4500 valid structured status information samples.

[0151] It should be noted that using structured coding methods can transform the original 5000 raw data samples into 4500 high-quality structured state information samples. This data processing technique not only helps subsequent cluster analysis better identify the boundaries of grid groups, but also improves the convergence performance of game optimization algorithms in complex network environments. During training, the algorithm will be exposed to more diverse operational state data, thus enabling it to better adapt to the dynamic changes in the operating state of the distribution network.

[0152] In this embodiment, the calculation of the comprehensive similarity index can also be achieved through other methods, such as introducing graph neural network technology. By using graph convolutional neural networks to perform deep learning on the network topology, higher-level topological feature representations can be automatically extracted, improving the expressive power of the similarity index. This method can not only capture local adjacency relationships but also learn complex association patterns between multi-hop neighbors, providing richer feature information for grid partitioning. Furthermore, to further improve the accuracy of the similarity index, time series analysis methods can be combined to incorporate the temporal correlation of historical operating data into the similarity calculation. Techniques such as dynamic time warping can be used to handle the time alignment problem of operating modes between different nodes, achieving a more accurate similarity measurement effect. Through these measures, it can be ensured that the generated variable grid group can more accurately reflect the actual operating characteristics of the distribution network, providing strong support for subsequent coordinated control.

[0153] For example, setting the weighting coefficient α in the comprehensive similarity index to 0.6 emphasizes the importance of topological similarity, as physical connections have a decisive impact on power transmission paths in distribution networks. Comparative experiments with different weighting coefficients show that when α = 0.6, the grid partitioning results maintain topological connectivity while reflecting the similarity of operating states, achieving a good balance.

[0154] Furthermore, when using the K-means clustering algorithm for grid partitioning, the elbow rule is used to determine the optimal number of clusters as 8, thus dividing the entire distribution network into 8 variable grid groups. Each grid group contains 15 to 25 monitoring nodes, which ensures both the coordinated control effect within the grid group and avoids the control complexity problem caused by an excessively large single grid group.

[0155] Furthermore, in the game optimization process, the game rounds were set to 50 rounds, and the convergence criterion was that the strategy change amplitude was less than 1% for 5 consecutive rounds. Experimental results show that in most cases, the algorithm can converge to a stable Nash equilibrium strategy within 30-40 rounds, significantly improving the coordination and control efficiency of the system.

[0156] Furthermore, the edge control results are transmitted to the cloud via the MQTT protocol, with an average transmission latency of 120 milliseconds, meeting the requirements for real-time control of the power distribution network. The cloud-based centralized optimization algorithm uses the interior-point method for solving, with a single optimization calculation time of approximately 200 milliseconds. The entire cloud-edge collaborative control cycle is controlled within 500 milliseconds, ensuring the system's real-time response capability.

[0157] According to experimental data, after adopting the method of this invention, the load reduction response time of the distribution network is shortened from 10-15 minutes in the traditional method to 2-3 minutes, the system operating cost is reduced by about 12%, and the renewable energy consumption rate is increased by about 8%, which fully verifies the effectiveness and practicality of the variable grid group control optimization method based on cloud-edge collaboration.

[0158] It should be noted that this invention achieves adaptive grid partitioning based on current physical connections and real-time operational status by constructing a comprehensive similarity index that integrates network structure and operational status. The partitioning method not only considers the static topological characteristics of the distribution network but also dynamically integrates the changing trends of operational control parameters such as voltage, current, and power at monitoring nodes, thus overcoming the shortcomings of traditional static partitioning methods that cannot respond to load changes and operational risks in real time. Simultaneously, by using game theory to achieve coordinated optimization among multiple grid groups, it solves the problem in traditional methods where strategies are formulated independently between grids, making it difficult to achieve global optimum, significantly improving the operational efficiency and control effect of the distribution network.

[0159] In summary, this invention clarifies the state perception and structured coding method for distributed resources in power distribution networks. A dynamic grid partitioning model is constructed based on real-time operational data from monitoring nodes, achieving adaptive partitioning and boundary identification of grid groups. A multi-agent coordination optimization strategy based on game theory is determined, realizing strategy coordination and conflict resolution among grid groups. It achieves an organic combination of edge-side collaborative scheduling and cloud-based global optimization, dynamically adjusting the operational control parameters of each grid group, and comprehensively optimizing the coordinated control of the power distribution network. This invention can provide intelligent coordinated control information for power distribution network operation, reduce system operational risks and costs, improve power distribution network operating efficiency and reliability, and achieve efficient utilization of distributed resources.

[0160] Example 3

[0161] The above is a schematic scheme of a variable grid group control optimization method based on cloud-edge collaboration. It should be noted that the technical solution of this system based on cloud-edge collaboration variable grid group control optimization belongs to the same concept as the technical solution of the aforementioned variable grid group control optimization method based on cloud-edge collaboration. Details not described in detail in the technical solution of the system based on cloud-edge collaboration in this embodiment can be found in the description of the technical solution of the aforementioned variable grid group control optimization method based on cloud-edge collaboration.

[0162] This embodiment also provides a variable grid group control and optimization system based on cloud-edge collaboration, including:

[0163] The state awareness module is used to collect real-time operational data of distributed resources and process state information to generate structured state information that reflects the relationship between physical state and network structure.

[0164] The grid partitioning module is used to adaptively partition the structured state information by combining topological similarity and operational feature differences through a clustering algorithm, generating multiple variable grid groups and topological boundary information;

[0165] The strategy modeling module is used to treat each variable grid group as an independent game subject and, in combination with the payoff function and constraints, construct the initial strategy set for each variable grid group.

[0166] The game optimization module is used to perform multiple rounds of game interaction and payoff evaluation on the initial strategy set of each variable grid group through game algorithm, dynamically optimize the selection of the initial strategy of each variable grid group, and obtain the optimal variable grid group strategy.

[0167] The edge scheduling module is used to perform edge-side collaborative scheduling and resource allocation based on the optimal variable grid group strategy, adjust the running status of the variable grid group in real time, and obtain edge control results.

[0168] The consistency control module is used to feed back the edge control results to the cloud for global coordination and integration, generate consistency policies, and issue them for execution.

[0169] This embodiment also provides an electronic device applicable to the case of variable grid group control optimization based on cloud-edge collaboration, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the variable grid group control optimization method based on cloud-edge collaboration as proposed in the above embodiment.

[0170] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the variable grid group control optimization method based on cloud-edge collaboration as proposed in the above embodiments.

[0171] The storage medium proposed in this embodiment and the variable grid group control optimization method based on cloud-edge collaboration proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0172] Based on the above description of the implementation methods, those skilled in the art will clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0173] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A variable grid swarm control optimization method based on cloud-edge collaboration, characterized in that: This includes collecting real-time operational data of distributed resources and processing state information to generate structured state information that reflects the relationship between physical state and network structure; By combining topological similarity and operational feature differences through clustering algorithms, the structured state information is adaptively divided to generate multiple variable grid groups and topological boundary information. Each variable grid group is treated as an independent game entity, and an initial strategy set for each variable grid group is constructed by combining the payoff function and constraints. By using a game theory algorithm, multiple rounds of game interaction and payoff evaluation are conducted on the initial strategy set of each variable grid group, and the selection of the initial strategy of each variable grid group is dynamically optimized to obtain the optimal variable grid group strategy. Based on the optimal variable mesh group strategy, edge-side collaborative scheduling and resource allocation are performed, and the operating status of the variable mesh group is adjusted in real time to obtain edge control results. The edge control results are fed back to the cloud for global coordination and integration, generating a consistent policy and issuing it for execution.

2. The variable grid swarm control optimization method based on cloud-edge collaboration as described in claim 1, characterized in that: The real-time operating data includes: monitoring node voltage, current, active power, reactive power, frequency, status indicators, and communication delay data; The state information processing includes missing value imputation and outlier correction processing, as well as feature scaling processing and structured encoding mapping.

3. The variable grid swarm control optimization method based on cloud-edge collaboration as described in claim 2, characterized in that: The adaptive partitioning includes at least topological feature similarity calculation and runtime feature difference calculation; The topological feature similarity calculation is used to extract the physical connection relationships between monitoring nodes; The calculation of operational feature differences is used to quantify the differences in operational status between nodes, and the resulting comprehensive similarity index is used for clustering.

4. The variable grid swarm control optimization method based on cloud-edge collaboration as described in claim 3, characterized in that: The comprehensive similarity index includes: Based on the similarity calculation of the topological connection characteristics of the monitoring nodes, the cosine similarity method is used to evaluate the network structure relationship between the nodes. Based on the difference in the operating characteristics of monitoring nodes, the Euclidean distance method is used to quantify the differences in the operating status between nodes. By integrating the similarity and difference by weighting coefficients, a comprehensive index for cluster analysis is constructed.

5. The variable grid swarm control optimization method based on cloud-edge collaboration as described in claim 4, characterized in that: The initial strategy set for constructing each variable mesh group includes: Identify the resource type, status information, and network topology characteristics of each variable grid group, and determine the control objectives and resource constraints of each grid group; Based on game theory, a payoff function model is constructed to quantify the expected payoffs of each variable grid group under different strategy choices. An optimization algorithm is used to calculate the feasible policy space of each variable grid group under the current state, and an initial policy set is generated.

6. The variable grid swarm control optimization method based on cloud-edge collaboration as described in claim 5, characterized in that: The game-playing algorithm includes: A game model is established based on the initial strategy set of each variable grid group, and game rules and interaction mechanisms are set. In each round of the game, the strategy choices and payoff performance of each variable grid group are evaluated, and the payoff results of each grid group are calculated and fed back. The strategy selection of each variable grid group is dynamically adjusted based on the game results and payoff feedback, and the optimal equilibrium strategy is found through multiple rounds of iteration.

7. The variable grid swarm control optimization method based on cloud-edge collaboration as described in claim 6, characterized in that: The edge-side collaborative scheduling includes: Based on the optimal variable grid group strategy, a time-series coordination and distributed allocation algorithm is used to coordinate the scheduling of each variable grid group. Based on the optimized resource allocation scheme, a distributed edge control strategy is adopted to dynamically adjust the operation control parameters of each variable grid group; The global coordination and integration includes transmitting edge control results to the cloud via communication protocols and generating a globally consistent control strategy using a centralized optimization algorithm.

8. A variable grid swarm control optimization system based on cloud-edge collaboration, based on the variable grid swarm control optimization method based on cloud-edge collaboration as described in any one of claims 1 to 7, characterized in that: It also includes a state awareness module, which is used to collect real-time operating data of distributed resources and process state information to generate structured state information that reflects the relationship between physical state and network structure. The grid partitioning module is used to adaptively partition the structured state information by combining topological similarity and operational feature differences through a clustering algorithm, generating multiple variable grid groups and topological boundary information; The strategy modeling module is used to treat each variable grid group as an independent game subject and, in combination with the payoff function and constraints, construct the initial strategy set for each variable grid group. The game optimization module is used to perform multiple rounds of game interaction and payoff evaluation on the initial strategy set of each variable grid group through game algorithm, dynamically optimize the selection of the initial strategy of each variable grid group, and obtain the optimal variable grid group strategy. The edge scheduling module is used to perform edge-side collaborative scheduling and resource allocation based on the optimal variable grid group strategy, adjust the running status of the variable grid group in real time, and obtain edge control results. The consistency control module is used to feed back the edge control results to the cloud for global coordination and integration, generate consistency policies, and issue them for execution.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the variable grid group control optimization method based on cloud-edge collaboration as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the variable grid swarm control optimization method based on cloud-edge collaboration as described in any one of claims 1 to 7.

Images