Smart park collaborative regulation method and system based on fast alternating direction multiplier method

Abstract

The application discloses a kind of wisdom park collaborative regulation methods and systems based on fast alternating direction multiplier method, it is related to wisdom park distributed collaborative regulation technical field.The method includes the following steps: according to the strength of the relationship between agents Partition;According to the partition result, with generator as node, the minimum operating cost of generator set is taken as target, and a distributed optimization control model is established;The fast alternating direction multiplier method is used to iteratively calculate the distributed optimization control model;Add adaptive penalty parameter in the process of iterative calculation, and obtain the optimal solution by combining the accelerated gradient method;According to the optimal solution, each partition of the wisdom park to be regulated is collaboratively regulated.The application reasonably partitions the multi-agent system, and optimizes the multi-agent system after partitioning and parallel distributed collaborative regulation, solves the problem that the existing distributed power supply unit structure is complex, resulting in low collaborative regulation efficiency.

Description

Technical Field

[0001] This invention relates to the field of distributed collaborative control technology for smart parks, and in particular to a collaborative control method and system for smart parks based on the fast alternating direction multiplier method. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] With the continuous development of society and the increasing emphasis on environmental protection, distributed renewable energy sources, represented by photovoltaics and wind power, have received widespread attention in the power system. However, due to the uncertainty of the spatiotemporal distribution of distributed renewable energy output, the regulation requirements for power systems with a high proportion of renewable energy are constantly increasing.

[0004] Traditional centralized power system control methods include convex optimization methods such as quadratic programming and Lagrange relaxation. With the development of computer communication technology, artificial intelligence has also provided an advanced method for finding optimal solutions in power system control. However, such algorithms require coordination by a centralized control center. When applied to smart park systems with multiple agents, the control center's centralized controller needs to collect information from all distributed power sources, energy storage devices, and flexible loads, integrating and processing this information to obtain the optimal allocation scheme. Due to the massive scale of distributed information in multi-agent systems, centralized processing requires exceptionally powerful computing capabilities from the control center, which is difficult to achieve in practical systems. Furthermore, centralized control for multi-agent systems also presents numerous challenges related to agility, system reliability, and information privacy.

[0005] To address the aforementioned problems, distributed collaborative regulation is typically employed to optimize multi-agent systems. Different scholars have proposed various optimization algorithms for distributed collaborative regulation: these include source-load collaborative fully distributed optimization regulation strategies based on consensus algorithms, coordinated power controller framework algorithms based on distributed economic scheduling methods, distributed energy management methods based on the multiplier alternating direction method, and FCM (fuzzy C-means, FCM) clustering algorithms based on initial cluster center selection, among others. These distributed algorithms do not require the collection of global data; they only need local information and weak communication with neighbors to achieve the control effect of centralized regulation.

[0006] However, the complex and diverse combinations of distributed power supply units pose significant challenges to real-time control and substantially increase the complexity of system operation. Therefore, how to rationally partition the multi-agent system and optimize its parallel distributed collaborative control after partitioning has become an urgent technical problem to be solved. Summary of the Invention

[0007] To address the shortcomings of existing technologies, the present invention aims to provide a smart park collaborative control method and system based on the fast alternating direction multiplier method. The method first divides the system into partitions according to the correlation between intelligent agents, and then uses the fast alternating direction multiplier method for distributed optimization scheduling, ultimately achieving rapid collaborative control of the smart park.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solution:

[0009] The first aspect of this invention provides a collaborative control method for smart parks based on the fast alternating direction multiplier method. The smart park to be controlled is viewed as a multi-agent system containing multiple agents. In the collaborative control of this multi-agent system, all agents must satisfy consistency, meaning that all agents eventually converge to the same state. Specifically, the method includes the following steps:

[0010] Partitioning based on the strength of relationships between agents;

[0011] Based on the partitioning results, a distributed optimization control model is established with generators as nodes and the goal of minimizing generator set operating costs.

[0012] The distributed optimization control model is iteratively calculated using the fast alternating direction multiplier method; an adaptive penalty parameter is added during the iterative calculation process, and the optimal solution is obtained by combining it with the accelerated gradient method.

[0013] Based on the optimal solution, each zone of the smart park to be regulated is coordinated and regulated.

[0014] Furthermore, the partitioning is based on the strength of the relationships between agents. When the relationship between a single agent and the other agents is weak, a single agent partitioning method is used. When there are multiple agents that are closely connected, a multi-agent partitioning method is used.

[0015] Furthermore, the single-agent partitioning method divides a single agent with weak connections to other agents into a region; the multi-agent partitioning method divides multiple closely connected agents into a region, and the presence of the same agent in two regions is called region coupling.

[0016] Furthermore, when using a multi-agent partitioning approach, if the states of the regional coupling parts between regions are consistent during the collaborative regulation of each partition of the smart park to be regulated based on the optimal solution, then the consistency condition is satisfied.

[0017] Furthermore, in the single-agent partitioning method, the consistency condition is satisfied when the agent's power generation is equal to its load.

[0018] Furthermore, the distributed optimization control model uses global power balance as a constraint.

[0019] Furthermore, the distributed optimization control model is iteratively calculated using the fast alternating direction multiplier method; the specific steps for obtaining the optimal solution by incorporating an adaptive penalty parameter and combining it with the accelerated gradient method during the iterative calculation are as follows:

[0020] Collect node data to obtain initialization parameters, global initial power generation, global initial load, and initial penalty parameters;

[0021] Based on the node data, the optimal node power generation is solved by using the fast alternating direction multiplier method.

[0022] The adaptive penalty parameter is determined based on the optimal node power generation, and the penalty parameter is updated based on the determination result;

[0023] Calculate the dual residuals based on the updated penalty parameters;

[0024] The initialization parameters, global initial power generation, and global initial load are updated using the accelerated gradient method until the dual residuals meet the preset accuracy requirements, thus obtaining the optimal solution.

[0025] A second aspect of the present invention provides a smart park collaborative control system based on the fast alternating direction multiplier method, comprising:

[0026] The partitioning module is configured to treat the smart park to be regulated as a multi-agent system containing multiple agents; and to partition the park according to the strength of the relationships between the agents.

[0027] The model building module is configured to establish a distributed optimization control model based on the partitioning results, with generators as nodes and the goal of minimizing generator set operating costs.

[0028] The model solving module is configured to perform iterative calculations on the distributed optimization control model using the fast alternating direction multiplier method; an adaptive penalty parameter is added during the iterative calculation process, and the optimal solution is obtained by combining it with the accelerated gradient method;

[0029] The coordinated control module is configured to perform coordinated control on each zone of the smart park to be controlled based on the optimal solution.

[0030] A third aspect of the present invention provides a medium on which a program is stored, which, when executed by a processor, implements the steps of the smart park collaborative control method based on the fast alternating direction multiplier method described in the first aspect of the present invention.

[0031] The fourth aspect of the present invention provides an apparatus including a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the smart park collaborative control method based on the fast alternating direction multiplier method as described in the first aspect of the present invention.

[0032] The above one or more technical solutions have the following beneficial effects:

[0033] This invention discloses a smart park collaborative control method and system based on the fast alternating direction multiplier method. Two partitioning methods are proposed based on the different numbers of agents, allowing for the selection of different methods to partition the multi-agent system as needed. Furthermore, a distributed optimization scheduling method based on A2DM2 is proposed for the distributed optimization control model constructed based on the above partitioning methods. Compared with the standard ADMM, this method can achieve parallel distributed optimization computation.

[0034] In solving the distributed optimization control model, the A2DM2 algorithm of this invention utilizes an accelerated gradient method, which greatly speeds up the convergence speed of the algorithm and obtains the optimal solution more quickly. It also employs an adaptive penalty parameter, comparing the original residual with the dual residual to determine whether the penalty parameter needs to be updated. The penalty parameter is updated synchronously during iteration according to the adaptive rule, further accelerating the convergence speed.

[0035] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0036] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0037] Figure 1 This is a partitioning structure diagram of an undecoupled multi-agent system when using a multi-agent partitioning method in Embodiment 1 of the present invention;

[0038] Figure 2 This is a partition structure diagram of the decoupled multi-agent system when using the multi-agent partitioning method in Embodiment 1 of the present invention;

[0039] Figure 3 The flowchart for solving the distributed optimization control model using the A2D2M algorithm is shown. Detailed Implementation

[0040] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0041] It should be noted that, in the embodiments of the present invention, data related to generator sets in smart parks is involved. When the above embodiments of the present invention are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0042] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0043] Example 1:

[0044] Embodiment 1 of this invention provides a collaborative control method for smart parks based on the fast alternating direction multiplier method, treating the smart park to be controlled as a multi-agent system containing multiple agents. A multi-agent system refers to a computing system in an environment composed of multiple agents with independent autonomous capabilities interacting through certain relationships. If each agent in the multi-agent system is considered a node, and information transmission between any two agents is connected by directed edges, then the topology of the multi-agent system can be represented using a corresponding directed graph.

[0045] Let G denote the network topology of the system. G represents a set, specifically G = (V, E, A), where V = {V1, V2, ..., V...} n} represents the set of n nodes in graph G. It is a non-empty set of nodes, also known as the node set. Let V be the set of edges, representing the set of unordered pairs of vertices in V, called the edge set; A is a symmetric n×n non-negative weighted adjacency matrix, where the adjacency elements a in the matrix are... ij It is a non-negative value.

[0046] The edges of G can be e ij =(V i V j The graph is represented as follows: the adjacent elements corresponding to the edges of the graph are positive, that is:

[0047]

[0048] Let N i The set of neighboring vertices of vertex i is defined as:

[0049] N i ={V j ∈V / (V iV j )∈E} (2)

[0050] Let d be the sum of the number of edges connected to vertex i. i d i It can be represented as the degree of vertex i, d i =|N i Let the edges of graph G be bidirectional and have equal weights. We can further define the adjacency element a in the adjacency matrix A. ij for:

[0051]

[0052] In a graph G, if any two nodes can be connected by a directed path from the edge set, then G is called a strongly connected graph; if G is an undirected graph and there is a path connecting any two nodes, then G is called a connected graph. The mathematical description of G being a strongly connected graph is as follows:

[0053] V i →V j :(V i V k1 )→(V k1 V k2 )→···→(V kl V j (4)

[0054] Among them, (V) m V n ) represents a directed path from node m to n.

[0055] Based on the graph theory described above, the connections between intelligent agents can be intuitively determined. In this embodiment, a directed path indicates that electrical energy can only be transmitted in this direction, while an undirected path indicates that electrical energy can be transmitted bidirectionally.

[0056] In the coordinated control of multi-agent systems, all agents must achieve consistency, meaning that all agents eventually converge to the same state. Consistency refers to the consensus reached by the agents during mutual interaction and cooperation, thereby achieving a consistent goal. In this invention, the goal refers to achieving optimal coordinated control. For example, when the system achieves optimal control, the total power generation of each agent should be equal to the total load, thus satisfying consistency. Consistency in a multi-agent system is one of the key conditions for ensuring its ability to complete coordinated control tasks.

[0057] Let x i Let i represent the state of node i. It can be considered that the nodes in the system reach consensus when the state values ​​of all nodes are equal.

[0058] x1=x2=···=x n (5)

[0059] The consensus algorithm for a linear system is defined as follows:

[0060]

[0061] The matrix form can be defined as:

[0062]

[0063] Among them, L n Let x be the n×n Laplace matrix of graph G; x = [x1, x2, ..., xn] n ] T .

[0064] In a multi-agent system, information exchange occurs between different agents; therefore, the state characteristics of an agent can be represented by the following equation:

[0065]

[0066] Since matrix A is a row random matrix, the above equation can be simplified to:

[0067]

[0068] The smart park collaborative control method based on the fast alternating direction multiplier method specifically includes the following steps:

[0069] Step 1: Divide the data into partitions based on the strength of the relationships between agents.

[0070] Step 2: Based on the partitioning results, establish a distributed optimization control model with generators as nodes and the goal of minimizing generator set operating costs.

[0071] Step 3: The distributed optimization control model is iteratively calculated using the fast alternating direction multiplier method; an adaptive penalty parameter is added during the iterative calculation process, and the optimal solution is obtained by combining it with the accelerated gradient method.

[0072] Step 4: Based on the optimal solution, perform coordinated regulation on each zone of the smart park to be regulated.

[0073] In step 1, in this invention, power equipment with variable output, such as distributed power sources, energy storage devices, and flexible loads, is considered as intelligent agents. The system is partitioned according to the strength of the relationships between these intelligent agents. A single-agent partitioning approach is used when the relationships between a single intelligent agent and other intelligent agents are weak; a multi-agent partitioning approach is used when multiple intelligent agents are closely connected.

[0074] The connections between agents are used as the basis for association. If an agent is only connected to another agent, the connection between the two agents is considered weak; the more agents an agent is connected to, the stronger the connection between that agent and other agents is considered.

[0075] (1) Single agent partitioning method

[0076] The single-agent partitioning method divides agents with weak connections to other agents into a single region. In this embodiment, the consistency condition is met when the agent's power generation equals its load.

[0077] In one specific implementation, assume there are N agents in the multi-agent system, and let each agent be considered a partition. As shown in Section 1.1, if V i Let i represent the i-th agent. Then the entire multi-agent system can be represented as:

[0078] V = {V1, V2, ..., V} N}(10)

[0079] Let B i，ij Let the boundary state be represented as the boundary state between the i-th agent and the j-th agent. Two scenarios will occur: (1) When agents i and j are not associated, their boundary states should both be 0; (2) When agents i and j are associated, their boundary states should be equal. That is:

[0080]

[0081] In a multi-agent system, the system is considered to satisfy the consistency condition when the boundary states of any two agents are equal.

[0082] B i，ij =B j，ij (i≠j) (12)

[0083] In this invention, boundary states refer to power generation and load. When an agent can receive / transmit electrical energy to another agent, the two agents are considered to be related. Conversely, if they cannot, they are not related. This partitioning method can be applied to multi-agent systems where the relationships between a single agent and the others are weak. Another partitioning method is given below, suitable for multi-agent systems where multiple agents are closely connected.

[0084] (2) Multi-agent partitioning method

[0085] The multi-agent partitioning approach divides multiple closely connected agents into regions. The presence of identical agents in two regions is called region coupling. When using this approach, during the coordinated control of each region within the smart park to be regulated based on the optimal solution, if the states of the region coupling portions between regions are consistent, then the consistency condition is satisfied.

[0086] In one specific implementation, multiple closely connected agents are divided into a region, such as... Figure 1 As shown, nodes 1 to 15 represent the number of multi-agents in the system; ①②③ represent the regions to be formed after decoupling, i.e., the closely connected multi-agent parts of the multi-agent system.

[0087] The structure diagram of the decoupled multi-agent system is as follows: Figure 2 As shown. (Through) Figure 2 It can be seen that there is a common coupling part between region ① and region ② after decoupling, namely agent V4 and agent V5; there is also a common coupling part between region ② and region ③, namely agent V7 and agent V8.

[0088] Therefore, when performing distributed optimal solutions, regions ①②③ can be used as distributed computing entities. When the computation ends, if the states of agents V4 and V5 in region ① are consistent with the states of agents V4 and V5 in region ②, then regions ① and ② are considered to satisfy the consistency condition. The same applies to regions ② and ③.

[0089] The above situation can be expressed by the following formula:

[0090] B m，j =B n，j (13)

[0091] Among them, B m，j V represents the agent V in the m-th partition after the calculation is completed. j state, B n，j Similarly; m and n are two coupled partitions, and agent V j This refers to the coupled portion of partition m and n.

[0092] In step 2, for a smart park that includes multiple agents such as distributed power sources, energy storage devices, and flexible loads, it can be assumed that for each generator node... Cost function of generator i It is continuous and differentiable. According to the superposition principle, the cost function of multiple generators is also continuous and differentiable. For ease of analysis, the cost function is assumed to be a quadratic function.

[0093] Assume P i L is the power generation of node i. j For node j, due to power balance, the sum of its power generation should equal the total load P. * As shown in equation (14)

[0094]

[0095] Among them, v represents the uncertainty of the output of the distributed power source in the simulation system. i∈{0,1} are the start / stop state variables of generator i. i =1 indicates that generator i is running, otherwise it is shut down. Based on consistency, the system is considered to be in consistency when the power generation in the smart park equals the load consumption.

[0096] This invention employs single-agent partitioning for optimization, but it can also be applied to multi-agent partitioning optimization. When applying multi-agent partitioning, each partitioned region can be treated as a new agent. Then, a distributed optimization and control model is constructed and solved based on the newly partitioned individual agents, thereby achieving control optimization.

[0097] To minimize operating costs, a distributed optimization control model was established with global power balance as a constraint and the goal of minimizing generator unit operating costs, as detailed below:

[0098]

[0099] in, P i and δ represents the upper and lower limits of the power of generator i, respectively. R represents the determined reserve capacity, which is equal to the product of the total load and the reserve percentage δ. As can be seen from (15), the objective function is convex and contains only linear constraints (equality constraints and inequality constraints). The Alternating Direction Method of Multipliers (ADMM) can be used for distributed solution calculation.

[0100] In step 3, for ease of representation, let

[0101]

[0102] Among them, S i Indicates that generator node i is in v i Power generation under the specified conditions; L * This represents the load consumed by all nodes.

[0103] Based on equations (15) and (16), the augmented Lagrange function can be listed as follows:

[0104]

[0105] Where λ is the augmented Lagrange multiplier vector; ρ is the penalty parameter.

[0106] The standard ADMM iteration steps can be listed according to equation (17):

[0107]

[0108] Where k is the number of iterations.

[0109] When applying the standard ADMM algorithm, the solution order between different subproblems is fixed. The solution result of the previous subproblem needs to be substituted into the solution of the next subproblem. After the solution is completed, the global coordinator propagates and updates λ based on the solution result. The standard ADMM algorithm solves the problem serially, with a convergence speed of O(1 / k), which is slow and not conducive to distributed optimization. Based on the above problems, this embodiment proposes the Accelerated Alternating Direction Method of Multipliers (A2DM2) for solving the problem.

[0110] The distributed optimization control model is iteratively calculated using the fast alternating direction multiplier method. Adaptive penalty parameters are added during the iterative calculation process, and the optimal solution is obtained by combining this with the accelerated gradient method. The specific steps are as follows: Figure 3 As shown:

[0111] Collect node data to obtain the initialization parameter λ and the global initial power generation P. * Global initial load L * The initial penalty parameter ρ is also specified; in this embodiment, the initialization parameter λ is set to 0. When there is prior knowledge of the actual problem, the initialization parameter can also be set to a positive number. The initial penalty parameter is set to a small positive number within [0,1], which can be set directly or selected based on past problem data.

[0112] Based on the node data, the optimal node power generation is solved by using the fast alternating direction multiplier method.

[0113] In one specific implementation, extracting the terms other than the objective function from equation (17) yields:

[0114]

[0115] Let u k =λ k / ρ, constant term If it can be omitted, the iteration steps can be updated to:

[0116]

[0117] In the standard ADMM algorithm, u k Updates require the participation of a global coordinator. To avoid using a global coordinator for updates, the algorithm is converted to synchronous calculation, selecting the average of the previous power generation calculation result and the total load calculation result as the reference value for the next iteration calculation, i.e.:

[0118]

[0119] Using the A2DM2 algorithm, the original residual and dual residual are obtained as follows: and

[0120] The adaptive penalty parameter is determined based on the optimal node power generation, and the penalty parameter is updated based on the determination result.

[0121] When using the standard ADMM algorithm to solve distributed optimization control models, its computational efficiency is significantly affected by the penalty parameter. An inappropriate penalty parameter selection can lead to a substantial increase in the number of iterations. Therefore, based on the residual balance principle, this paper proposes an adaptive penalty parameter to synchronously update the penalty parameter during iteration, thereby accelerating the convergence speed.

[0122] In one specific implementation, the adaptive penalty parameter ρ is calculated as follows:

[0123]

[0124] Where η>1, τ incr >1, τ decr >1. , where η is the residual judgment factor, generally set to 10; τ incr τ is the multiplication factor, usually set to 2; decr This is the multiplication factor, typically set to 2. The above parameters can also be selected according to the actual system requirements.

[0125] The dual residuals are calculated based on the updated penalty parameters. Increasing the value of ρ enhances the minimization of the norm terms in equations (17) and (19), promoting the equalization of power generation with the global load and accelerating the convergence of the original residuals. Decreasing the value of ρ suppresses the oscillation of the objective function and accelerates the convergence of the global load.

[0126] The initialization parameters, global initial power generation, and global initial load are updated using the accelerated gradient method until the dual residuals meet the preset accuracy requirements, thus obtaining the optimal solution.

[0127] To improve the convergence of the algorithm and achieve a convergence speed of O(1 / k²), the accelerated gradient method is used to update the Lagrange multiplier λ and the reference value.

[0128]

[0129] Where α is a slack variable used to accelerate the solution of λ and

[0130] The A2DM2 algorithm uses dual residual d k+1 The convergence criterion is that the point approaches zero, i.e.:

[0131]

[0132] At this point, the convergence accuracy is ε, and the optimal solution is output.

[0133] In step 4, the output of each distributed power source when the overall result is optimal, i.e., the cost is lowest, can be obtained by solving the problem. The distributed power sources can then be regulated based on the optimal solution.

[0134] Example 2:

[0135] Embodiment 2 of the present invention provides a smart park collaborative control system based on the fast alternating direction multiplier method, comprising:

[0136] The partitioning module is configured to treat the smart park to be regulated as a multi-agent system containing multiple agents; and to partition the park according to the strength of the relationships between the agents.

[0137] The model building module is configured to establish a distributed optimization control model based on the partitioning results, with generators as nodes and the goal of minimizing generator set operating costs.

[0138] The model solving module is configured to perform iterative calculations on the distributed optimization control model using the fast alternating direction multiplier method; an adaptive penalty parameter is added during the iterative calculation process, and the optimal solution is obtained by combining it with the accelerated gradient method;

[0139] The coordinated control module is configured to perform coordinated control on each zone of the smart park to be controlled based on the optimal solution.

[0140] Example 3:

[0141] Embodiment 3 of the present invention provides a medium on which a program is stored. When the program is executed by a processor, it implements the steps in the smart park collaborative control method based on the fast alternating direction multiplier method as described in Embodiment 1 of the present invention.

[0142] Example 4:

[0143] Embodiment 4 of the present invention provides a device, including a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the smart park collaborative control method based on the fast alternating direction multiplier method described in Embodiment 1 of the present invention.

[0144] The steps and methods involved in Embodiments 2, 3, and 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0145] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0146] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A collaborative control method for smart parks based on the fast alternating direction multiplier method, which treats the smart park to be controlled as a multi-agent system containing multiple agents. In the collaborative control of the multi-agent system, all agents must satisfy consistency, meaning that all agents eventually converge to the same state; its characteristic is that... The method specifically includes the following steps: The partitioning is based on the strength of the relationships between agents; this includes partitioning based on the strength of the relationships between agents. When a single agent has a weak relationship with other agents, a single-agent partitioning method is used; when multiple agents have a strong relationship, a multi-agent partitioning method is used. The single-agent partitioning method divides a single agent with a weak relationship with other agents into one region. The multi-agent partitioning method divides multiple agents with a strong relationship into one region. The presence of the same agent in two regions is called region coupling. When using a multi-agent partitioning approach, if the states of the regional coupling parts between regions are consistent during the coordinated regulation of each partition of the smart park to be regulated according to the optimal solution, then the consistency condition is satisfied. In the single-agent partitioning method, the consistency condition is satisfied when the agent's power generation is equal to its load. Based on the partitioning results, a distributed optimization control model is established with generators as nodes and the goal of minimizing generator set operating costs. The distributed optimization control model is iteratively calculated using the fast alternating direction multiplier method; an adaptive penalty parameter is added during the iterative calculation process, and the optimal solution is obtained by combining it with the accelerated gradient method. Based on the optimal solution, each zone of the smart park to be regulated is coordinated and regulated.

2. The smart park collaborative control method based on the fast alternating direction multiplier method as described in claim 1, characterized in that, The distributed optimization control model takes global power balance as a constraint.

3. The smart park collaborative control method based on the fast alternating direction multiplier method as described in claim 1, characterized in that, The distributed optimization control model is iteratively calculated using the fast alternating direction multiplier method. The specific steps for obtaining the optimal solution during the iterative calculation, incorporating an adaptive penalty parameter and combining it with the accelerated gradient method, are as follows: Collect node data to obtain initialization parameters, global initial power generation, global initial load, and initial penalty parameters; Based on the node data, the optimal node power generation is solved by using the fast alternating direction multiplier method. The adaptive penalty parameter is determined based on the optimal node power generation, and the penalty parameter is updated based on the determination result; Calculate the dual residuals based on the updated penalty parameters; The initialization parameters, global initial power generation, and global initial load are updated using the accelerated gradient method until the dual residuals meet the preset accuracy requirements, thus obtaining the optimal solution.

4. A smart park collaborative control system based on the fast alternating direction multiplier method, characterized in that, The smart park collaborative control method based on the fast alternating direction multiplier method as described in any one of claims 1-3 includes: The partitioning module is configured to treat the smart park to be regulated as a multi-agent system containing multiple agents; and to partition the park according to the strength of the relationships between the agents. The model building module is configured to establish a distributed optimization control model based on the partitioning results, with generators as nodes and the goal of minimizing generator set operating costs. The model solving module is configured to perform iterative calculations on the distributed optimization control model using the fast alternating direction multiplier method; an adaptive penalty parameter is added during the iterative calculation process, and the optimal solution is obtained by combining it with the accelerated gradient method; The coordinated control module is configured to perform coordinated control on each zone of the smart park to be controlled based on the optimal solution.

5. A computer-readable storage medium, characterized in that, The device stores multiple instructions, which are adapted to be loaded and executed by the processor of the terminal device. The instructions are based on the fast alternating direction multiplier method for collaborative control of smart parks, as described in any one of claims 1-3.

6. A terminal device, characterized in that, The invention includes a processor and a computer-readable storage medium, wherein the processor implements various instructions; and the computer-readable storage medium stores multiple instructions adapted to be loaded by the processor and executed as described in any one of claims 1-3, namely, the smart park collaborative control method based on the fast alternating direction multiplier method.

Images