Power distribution network multi-transformer-area collaborative optimization regulation and control method and device adapting to topological change

Abstract

The embodiment of the invention provides a power distribution network multi-zone-area collaborative optimization regulation and control method and device adapting to topological changes. The method comprises the steps of performing cluster division on a power distribution network; constructing an optimization model according to a cluster division result; respectively constructing a dynamic grouping Markov decision model of each transformer area by taking each transformer area as an intelligent agent; solving the dynamic grouping Markov decision model based on the optimization model to obtain a corresponding power distribution network multi-transformer-area collaborative optimization regulation and control strategy, and regulating and controlling the power distribution network according to the regulation and control strategy; according to the method, dynamic cluster division and distributed autonomous optimization can be combined, and the SLM algorithm is adopted to enhance the station area coordination capability and reduce the cross-cluster power flow influence; according to the invention, the feature extraction network is introduced to adapt to topological change, and the generation of the regulation strategy is optimized; communication dependence is reduced through centralized training and decentralized execution, and efficient voltage regulation and control in a low-perceptibility scene are achieved in a large-scale power distribution network complex environment.

Description

Technical Field

[0001] The present application relates to the field of power grid control, and specifically to a method and device for collaborative optimization and control of multiple distribution network areas that adapt to topology changes. Background Art

[0002] With the widespread access to new energy and distributed resources, the operation of distribution networks faces many challenges. The randomness and volatility of intermittent distributed photovoltaic and other energy sources can easily lead to problems such as power flow reversal and voltage over-limit, posing a threat to the stability and reliability of the distribution network. At the same time, traditional voltage regulation equipment (such as grouped capacitors and on-load tap-changing transformers) cannot meet the real-time and high-precision requirements of modern distribution networks due to their discrete regulation methods and long response times. However, new reactive resources such as photovoltaic inverters and static VAR compensators (SVCs) have become an ideal choice for optimizing voltage control in distribution networks with their fast response and high-precision regulation.

[0003] However, the dynamic characteristics of the distribution network (such as topology changes) and the lack of multi-source data in low-perception environments make it difficult for traditional centralized optimization methods to efficiently achieve voltage regulation. Existing mathematical optimization methods rely on accurate physical models and network parameters, such as solving non-convex problems through robust optimization and linearization techniques to improve system voltage stability, but their requirements for model accuracy limit their use in actual low-perception scenarios. On the other hand, deep reinforcement learning, as a data-driven method, provides a more flexible solution through offline training and fast feedforward operations, which can deal with the problem of missing parameters to a certain extent. However, when faced with complex topologies and dynamically changing environments, the existing deep reinforcement learning algorithms have limited feature extraction capabilities of their neural networks, making it difficult to generate high-precision and highly applicable control strategies. Summary of the invention

[0004] In response to the problems in the prior art, the present application provides a method and device for collaborative optimization and control of multiple distribution areas in a distribution network that adapts to topology changes. The method and device can combine dynamic cluster division and distributed autonomous optimization, adopt the SLM algorithm to enhance the collaborative ability of the areas and reduce the impact of cross-cluster power flow; adapt to topology changes and optimize control strategy generation by introducing a feature extraction network; use centralized training and decentralized execution to reduce communication dependence, and achieve efficient voltage control in low-perception scenarios in the complex environment of large-scale distribution networks.

[0005] In order to solve at least one of the above problems, the present application provides the following technical solutions:

[0006] According to a first aspect of an embodiment of the present application, the present application provides a method for coordinated optimization and control of multiple distribution areas in a distribution network that adapts to topology changes, including:

[0007] Based on the electrical distance modularity and power balance, the distribution network is clustered using the SLM clustering algorithm.

[0008] An optimization model including voltage deviation minimization and power loss minimization objective functions is constructed according to the cluster division results, wherein the objective function of the optimization model includes the calculation of node voltage deviation within the cluster and cluster power loss, and the constraints include power balance, node voltage safety, photovoltaic reactive power adjustment range, and preset range of SOC state of energy storage equipment;

[0009] Take each area as an intelligent agent, and build a dynamic clustering Markov decision model for each area according to the observation space, action space, reward function and state transition of the intelligent agent;

[0010] Based on the optimization model, the dynamic clustering Markov decision model is solved using the multi-agent deep deterministic policy gradient MADDPG algorithm to obtain the corresponding distribution network multi-station collaborative optimization control strategy, and the distribution network is regulated according to the control strategy.

[0011] According to any implementation manner of the present application, clustering the distribution network by using an SLM clustering algorithm based on electrical distance modularity and power balance includes:

[0012] Each substation in the distribution network is regarded as an independent initial cluster, and the node connection relationship and load power data of all substations are obtained;

[0013] The predefined electrical distance modularity and power balance are used as indicators for cluster division. In each iteration, each substation is tried to be assigned to other clusters. The change in electrical distance modularity after assignment is calculated, and the assignment method that produces the maximum modularity gain is recorded. This process is repeated to gradually optimize the cluster division until the modularity no longer changes.

[0014] Several sub-networks are generated through the SLM cluster partitioning algorithm, and each sub-network is regarded as a virtual node to participate in the next round of cluster partitioning, and the modularity is continuously optimized until it reaches the global optimum.

[0015] According to any embodiment of the present application, constructing an optimization model including voltage deviation minimization and power loss minimization objective functions according to the cluster division result includes:

[0016] Based on the cluster division results, objective functions are defined with voltage deviation minimization and power loss minimization as optimization objectives, respectively. The voltage deviation objective function represents the sum of square deviations between the voltage amplitudes of all nodes in the cluster and the voltage reference value, and the power loss objective function is calculated by the branch resistance and its active power and reactive power flow in the cluster.

[0017] According to the power balance principle within the cluster, the power balance relationship of each node is calculated based on the line admittance matrix and the voltage phase angle difference between nodes to perform power balance constraints on the distribution network;

[0018] Based on the preset range, the voltage amplitude of all nodes is controlled within the safe threshold range, and the branch current is limited to not exceed the maximum allowed current;

[0019] Determine the active power output of the photovoltaic power generation system through the maximum power point tracking mode MPPT, and set the reactive power adjustment range of the photovoltaic system according to the reactive power capacity of the photovoltaic power generation system;

[0020] Based on the preset range, the SOC state of the energy storage system at the current moment is controlled to be within the safety threshold range, and the charging and discharging power range of the energy storage device is set in combination with the charging and discharging efficiency of the energy storage system.

[0021] According to any implementation of the present application, each station area is regarded as an intelligent agent, and a dynamic clustering Markov decision model for each station area is constructed according to the observation space, action space, reward function and state transition of the intelligent agent, including:

[0022] Each substation is regarded as an intelligent entity, and the observation space is defined as the real-time power flow information of all nodes in the substation according to the current cluster boundary and substation node status, including the active power, reactive power and voltage amplitude of the nodes, and the observation space is dynamically adjusted to adapt to the new cluster structure when the cluster boundary changes;

[0023] Based on the observation information of the intelligent agent, the action space is defined as the active power and reactive power regulation of the interaction between the substation and the feeder, including the power output or absorption achieved by photovoltaic and energy storage equipment, and the power of the substation nodes is redistributed after the action is executed;

[0024] Constructing a reward function of the intelligent agent according to the control target, including a reward item based on the target function and a penalty item based on the constraint condition, wherein the reward item reflects the achievement of the control target by calculating the node voltage deviation and power loss, and the penalty item punishes the violation of the energy storage device constraint condition;

[0025] Among them, from the initial state until there is no feasible solution for the power flow or the scheduling cycle ends, the agent calculates the reward function value after performing actions according to the current observation space and action space, and enters the next moment state on this basis to perform state transfer.

[0026] According to any implementation of the present application, based on the optimization model, the dynamic clustering Markov decision model is solved using a multi-agent deep deterministic policy gradient MADDPG algorithm to obtain a corresponding distribution network multi-station area collaborative optimization control strategy, including:

[0027] Build an Actor network for each agent, and output the corresponding active power and reactive power regulation based on the observation value input of the agent;

[0028] Construct a critic network for each agent, and evaluate the value of each action based on the set of observations and actions of all agents in the current state, where the output of the control critic network is based on the parameter update of the target critic network;

[0029] Set a target Actor network and a target Critic network for each agent, and regularly copy the parameters of the training network to the target network to maintain parameter stability in reinforcement learning;

[0030] In each training iteration, based on the actions generated by the Actor network and the values ​​evaluated by the Critic network, the parameters of the Critic network are optimized by gradient descent to minimize the joint regression loss, and the parameters of the Actor network are optimized by gradient ascent to maximize the action value of the agent.

[0031] According to the iterative training results, the corresponding distribution network multi-station coordinated optimization control strategy is obtained, and the power regulation amount of the station is output based on the current observation space and the optimized action value.

[0032] According to any embodiment of the present application, the method of constructing a Critic network for each agent and evaluating the value of each action based on the observation and action set of all agents in the current state, wherein the output of the Critic network is controlled based on the parameter update of the target Critic network, comprises:

[0033] Input the environmental information and attention information of the distribution network environment to the Critic network of each intelligent agent;

[0034] The attention mechanism is used to generate weights for the observation and action combination of each agent, and the key-value query method is used to calculate the correlation between the current agent and other agents, and the weighted parameters are generated to describe the contribution of different agents to value assessment.

[0035] The agent's environmental information and attention information are fused, and the fused features are input into the Critic network. Through the activation function and the fully connected network, the value output of the agent's action is generated;

[0036] Among them, according to the output of the critic network and the output of the target critic network, a joint regression loss function is defined, and the critic network is updated by minimizing the loss function.

[0037] According to a second aspect of an embodiment of the present application, the present application provides a distribution network multi-zone coordinated optimization control device that adapts to topology changes, including:

[0038] The cluster partitioning module is used to: cluster the distribution network based on the electrical distance modularity and power balance by using the SLM cluster partitioning algorithm;

[0039] An optimization model construction module is used to: construct an optimization model including voltage deviation minimization and power loss minimization objective functions according to the cluster division results, wherein the objective function of the optimization model includes the calculation of node voltage deviation and cluster power loss within the cluster, and the constraint conditions include power balance, node voltage safety, photovoltaic reactive power adjustment range and preset range of SOC state of energy storage equipment;

[0040] The agent modeling module is used to: treat each station as an agent, and construct a dynamic clustering Markov decision model for each station according to the observation space, action space, reward function and state transfer of the agent;

[0041] The optimization solution module is used to: based on the optimization model, use the multi-agent deep deterministic policy gradient MADDPG algorithm to solve the dynamic clustering Markov decision model, obtain the corresponding distribution network multi-station area collaborative optimization control strategy, and control the distribution network according to the control strategy.

[0042] According to any implementation manner of the present application, the cluster division module includes:

[0043] The cluster initialization unit is used to: regard each substation in the distribution network as an independent initial cluster, and obtain the node connection relationship and load power data of all substations;

[0044] An iterative optimization unit is used to: use the predefined electrical distance modularity and power balance as indicators for cluster division, try to allocate each substation to other clusters in each iteration, calculate the change value of the electrical distance modularity after allocation, record the allocation method that produces the maximum modularity gain, and repeat this process to gradually optimize the cluster division until the modularity no longer changes;

[0045] The subnetwork merging unit is used to generate several subnetworks through the SLM cluster partitioning algorithm, regard each subnetwork as a virtual node to participate in the next round of cluster partitioning, and continue to optimize the modularity until it reaches the global optimum.

[0046] According to any implementation of the present application, the optimization model building module includes:

[0047] An objective function definition unit is used to: define objective functions based on the cluster division result, taking voltage deviation minimization and power loss minimization as optimization objectives, respectively, wherein the voltage deviation objective function represents the sum of square deviations between the voltage amplitudes of all nodes in the cluster and the voltage reference value, and the power loss objective function is calculated by branch resistances in the cluster and the active power and reactive power flows thereof;

[0048] The power balance constraint unit is used to: calculate the power balance relationship of each node based on the line admittance matrix and the voltage phase angle difference between nodes according to the power balance principle within the cluster to perform power balance constraint on the distribution network;

[0049] A safety constraint unit is used to: control the voltage amplitude of all nodes within a safety threshold range based on a preset range, and limit the branch current to not exceed the maximum allowed current;

[0050] A photovoltaic regulation unit is used to: determine the active power output of the photovoltaic power generation system through the maximum power tracking mode MPPT, and set the reactive power regulation range of the photovoltaic system according to the reactive power capacity of the photovoltaic power generation system;

[0051] The energy storage management unit is used to: based on a preset range, control the SOC state of the energy storage system at the current moment to be within a safe threshold range, and set the charge and discharge power range of the energy storage device in combination with the charge and discharge efficiency of the energy storage system.

[0052] According to any embodiment of the present application, the agent modeling module includes:

[0053] The observation space definition unit is used to: regard each substation as an intelligent agent, and define the observation space as the real-time power flow information of all nodes in the substation according to the current cluster boundary and substation node status, including the active power, reactive power and voltage amplitude of the nodes, and dynamically adjust the observation space to adapt to the new cluster structure when the cluster boundary changes;

[0054] An action space definition unit is used to: define the action space as the active power and reactive power regulation of the interaction between the substation and the feeder based on the observation information of the intelligent agent, including the power output or absorption achieved by photovoltaic and energy storage equipment, and redistribute the power of the substation nodes after the action is executed;

[0055] A reward function construction unit, used to: construct a reward function of the intelligent agent according to the control target, including a reward item based on the target function and a penalty item based on the constraint condition, wherein the reward item reflects the achievement of the control target by calculating the node voltage deviation and power loss, and the penalty item punishes the violation of the energy storage device constraint condition;

[0056] The state transfer unit is used to: from the initial state until there is no feasible solution for the power flow or the scheduling cycle ends, calculate the reward function value after the agent performs actions according to the current observation space and action space, and enter the next moment state on this basis to perform state transfer.

[0057] According to any implementation of the present application, the optimization solution module includes:

[0058] Actor network construction unit, used to: construct an Actor network for each intelligent agent, based on the observation value input of the intelligent agent, output the corresponding active power and reactive power adjustment amount;

[0059] The critic network construction unit is used to: construct a critic network for each agent, evaluate the value of each action based on the set of observations and actions of all agents in the current state, and control the output of the critic network based on the parameter update of the target critic network;

[0060] The target network setting unit is used to set the target actor network and target critic network for each agent, and regularly copy the parameters of the training network to the target network to maintain parameter stability in reinforcement learning;

[0061] The network parameter optimization unit is used to: optimize the parameters of the Critic network by gradient descent to minimize the joint regression loss, and optimize the parameters of the Actor network by gradient ascent to maximize the action value of the agent in each training iteration based on the actions generated by the Actor network and the values ​​evaluated by the Critic network;

[0062] The control strategy generation unit is used to obtain the corresponding distribution network multi-station area collaborative optimization control strategy according to the iterative training results, and output the power regulation amount of the station area based on the current observation space and the optimized action value.

[0063] According to any implementation of the present application, the Critic network construction unit includes:

[0064] An environmental information input unit is used to: input environmental information and attention information of the distribution network environment into the Critic network of each intelligent agent;

[0065] An attention weight generation unit, used to: generate a weight for the observation and action combination of each agent through an attention mechanism, calculate the correlation between the current agent and other agents using a key-value query device, and generate weighted parameters to describe the contribution of different agents to value assessment;

[0066] The feature fusion unit is used to fuse the agent's environmental information and attention information, input the fused features into the Critic network, and generate the value output of the agent's actions through the activation function and the fully connected network;

[0067] The network updating unit is used to define a joint regression loss function according to the output of the critic network and the output of the target critic network, and update the critic network by minimizing the loss function.

[0068] According to the third aspect of the embodiments of the present application, the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method for collaborative optimization and control of multiple distribution network areas that adapt to topology changes are implemented.

[0069] According to the fourth aspect of the embodiments of the present application, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for collaborative optimization and control of multiple distribution network areas that adapts to topology changes.

[0070] According to the fifth aspect of the embodiments of the present application, the present application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the method for collaborative optimization and control of multiple distribution network areas that adapt to topology changes.

[0071] It can be seen from the above technical scheme that the present application provides a method and device for collaborative optimization and control of multiple distribution areas in a distribution network that adapts to topology changes, by clustering the distribution network; constructing an optimization model based on the cluster division results; taking each area as an intelligent body and constructing a dynamic clustering Markov decision model for each area; solving the dynamic clustering Markov decision model based on the optimization model to obtain the corresponding collaborative optimization and control strategy for multiple distribution areas, and regulating the distribution network according to the control strategy; the present application can combine dynamic clustering and distributed autonomous optimization, and use the SLM algorithm to enhance the collaborative ability of the areas and reduce the impact of cross-cluster power flow, and can combine dynamic clustering and distributed autonomous optimization, and use the SLM algorithm to enhance the collaborative ability of the areas and reduce the impact of cross-cluster power flow; adapt to topology changes by introducing a feature extraction network, and optimize the generation of control strategies; use centralized training and decentralized execution to reduce communication dependence, and achieve efficient voltage control in low-perception scenarios in complex environments of large-scale distribution networks. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0073] Figure 1 This is one of the flow charts of the method for coordinated optimization and control of multiple distribution areas in a distribution network that adapts to topology changes in an embodiment of the present application;

[0074] Figure 2 This is a second flow chart of a method for collaborative optimization and control of multiple distribution areas in a distribution network that adapts to topology changes in an embodiment of the present application;

[0075] Figure 3 This is a flow chart of the third method for collaborative optimization and control of multiple distribution areas in a distribution network that adapts to topology changes in an embodiment of the present application;

[0076] Figure 4 A schematic diagram of a feeder-multi-distribution station area collaborative structure of a distribution network multi-station area collaborative optimization control method that adapts to topology changes in an embodiment of the present application;

[0077] Figure 5 This is a fourth flow chart of a method for collaborative optimization and control of multiple distribution areas in a distribution network that adapts to topology changes in an embodiment of the present application;

[0078] Figure 6 A schematic diagram of cluster boundary changes under different topologies of a distribution network multi-zone collaborative optimization control method that adapts to topology changes in an embodiment of the present application;

[0079] Figure 7 This is a fifth flow chart of a method for collaborative optimization and control of multiple distribution areas in a distribution network that adapts to topology changes in an embodiment of the present application;

[0080] Figure 8 This is a sixth flow chart of a method for collaborative optimization and control of multiple distribution areas in a distribution network that adapts to topology changes in an embodiment of the present application;

[0081] Fig. 9 A schematic diagram of the AECN network structure of the method for collaborative optimization and control of multiple distribution areas in a distribution network that adapts to topology changes in an embodiment of the present application;

[0082] Fig.10 A structural diagram of a distribution network multi-zone collaborative optimization control device that adapts to topology changes in an embodiment of the present application;

[0083] Fig.11 It is a schematic diagram of the structure of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0084] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0085] The acquisition, storage, use, and processing of data in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0086] Taking into account the relevant problems existing in the prior art, the present application provides a method and device for collaborative optimization and control of multiple distribution areas in a distribution network that adapts to topology changes. The method and device can combine dynamic cluster division and distributed autonomous optimization, adopt the SLM algorithm to enhance the coordination capability of the areas and reduce the impact of cross-cluster power flow; adapt to topology changes and optimize control strategy generation by introducing a feature extraction network; use centralized training and decentralized execution to reduce communication dependence, and achieve efficient voltage control in low-perception scenarios in the complex environment of large-scale distribution networks.

[0087] In order to combine dynamic cluster division and distributed autonomous optimization, the SLM algorithm is used to enhance the coordination ability of the substations and reduce the impact of cross-cluster power flow; by introducing a feature extraction network to adapt to topology changes, optimize the generation of control strategies; using centralized training and decentralized execution to reduce communication dependence, and achieve efficient voltage control in low-perception scenarios in complex environments of large-scale distribution networks, this application provides an embodiment of a method for coordinated optimization and control of multiple substations in a distribution network that adapts to topology changes, see Figure 1 The distribution network multi-station area coordinated optimization control method that adapts to topology changes specifically includes the following contents:

[0088] Step S101: Based on the electrical distance modularity and the power balance, the distribution network is clustered by using the SLM clustering algorithm.

[0089] First, each substation in the distribution network is regarded as an independent initial cluster, and its node connection relationship and load power data are obtained. The SLM algorithm is used to dynamically cluster the distribution network with electrical distance modularity and power balance as indicators. Electrical distance modularity reflects the degree of tight coupling between nodes between substations, while power balance evaluates the supply and demand matching within the substation. In each iteration, the substation is tried to be assigned to other clusters, and the allocation method that maximizes the modularity gain is recorded until the modularity no longer changes. Subsequently, further optimization is performed by generating virtual subnetworks, and the global optimal clustering result is finally determined.

[0090] Step S102: construct an optimization model including voltage deviation minimization and power loss minimization objective functions according to the cluster division results, wherein the objective function of the optimization model includes the calculation of node voltage deviation within the cluster and cluster power loss, and the constraints include power balance, node voltage safety, photovoltaic reactive power adjustment range and a preset range of the SOC state of the energy storage device.

[0091] Based on the cluster division results, an optimization model is constructed with voltage deviation minimization and power loss minimization as the objective functions. The voltage deviation objective function measures the sum of squares of the deviations of all node voltages from the reference value, and the power loss objective function is calculated through the branch resistance within the cluster and the flow of active and reactive power. At the same time, the model sets a variety of constraints, including node power balance constraints, voltage safety range constraints, photovoltaic reactive power adjustment range constraints, and energy storage system SOC status and charging and discharging power limits, providing clear goals and constraints for the generation of subsequent intelligent control strategies.

[0092] Step S103: Take each area as an intelligent agent, and construct a dynamic clustering Markov decision model for each area according to the observation space, action space, reward function and state transition of the intelligent agent.

[0093] Each substation is abstracted as an intelligent agent and described by a dynamic clustering Markov decision model. The observation space includes real-time power flow information of all nodes in the substation, such as active power, reactive power and voltage amplitude, and can dynamically adapt to changes in cluster boundaries. The action space is defined as the active and reactive power regulation achieved by photovoltaic and energy storage equipment. The reward function combines the objective function and constraints, including reward items for achieving the goal and penalty items for violating the constraints. The agent generates the next state based on the current observation and action until the end of the regulation cycle or the power flow cannot be solved, completing the state transfer and dynamic modeling.

[0094] Step S104: Based on the optimization model, the dynamic clustering Markov decision model is solved using the multi-agent deep deterministic policy gradient MADDPG algorithm to obtain the corresponding distribution network multi-station area collaborative optimization control strategy, and the distribution network is controlled according to the control strategy.

[0095] The dynamic clustering Markov decision model is solved by the MADDPG algorithm. An Actor network is constructed for each agent to generate action outputs based on the observation space; at the same time, a Critic network is constructed to evaluate the action value and stabilize the training through the target network. During the training process, the Critic network is optimized by gradient descent to minimize the regression loss, and the Actor network is optimized by gradient ascent to maximize the strategy value. After the training is completed, the generated control strategy is used for real-time power regulation of the substation equipment, covering photovoltaic reactive power output and energy storage equipment charging and discharging, to achieve multi-station coordinated optimization and control of the distribution network.

[0096] From the above description, it can be seen that the collaborative optimization and control method for multiple distribution network areas that adapt to topology changes provided in the embodiment of the present application can combine dynamic cluster division and distributed autonomous optimization, use the SLM algorithm to enhance the area coordination capability and reduce the impact of cross-cluster power flow; adapt to topology changes by introducing a feature extraction network, and optimize the generation of control strategies; use centralized training and decentralized execution to reduce communication dependence, and achieve efficient voltage control in low-perception scenarios in the complex environment of large-scale distribution networks.

[0097] In order to improve the coordination capability within the cluster, reduce the impact of cross-cluster power flow, dynamically adapt to topology changes, optimize modularity to achieve global optimal division, and reduce power loss and control complexity, in one embodiment of the coordinated optimization control method of multiple distribution areas in the distribution network that adapts to topology changes in the present application, see Figure 2 The method of clustering the distribution network based on the electrical distance modularity and power balance by using the SLM cluster partitioning algorithm includes:

[0098] Step S101A: Consider each substation in the distribution network as an independent initial cluster, and obtain the node connection relationship and load power data of all substations;

[0099] Step S101B: Use the predefined electrical distance modularity and power balance as indicators for cluster division, try to allocate each substation to other clusters in each iteration, calculate the change value of the electrical distance modularity after allocation, record the allocation method that produces the maximum modularity gain, and repeat this process to gradually optimize the cluster division until the modularity no longer changes;

[0100] Step S101C: Generate several sub-networks through the SLM cluster partitioning algorithm, regard each sub-network as a virtual node to participate in the next round of cluster partitioning, and continue to optimize the modularity until it reaches the global optimum.

[0101] For example, the cluster division index is the basis for clustering distribution substations. The electrical coupling degree within the divided clusters is high and the inter-cluster decoupling is approximately high. This paper is guided by the voltage regulation of the distribution network. At the same time, in order to reduce the power loss across clusters, the two cluster division indexes of electrical distance modularity and power balance are mainly considered.

[0102] Electrical distance modularity index: Modularity is used to measure the strength index of community network structure. The larger the modularity value, the better the cluster division result. The modularity function ρ is defined as:

[0103]

[0104] Where: W ij is the adjacency matrix in the modularity index; K i is the sum of the edge weights connected to node i; s is the sum of all network weights; C i represents the cluster to which the station i belongs, λ is a 0-1 function, and λ=1 indicates that nodes i and j are divided into one cluster.

[0105] Power balance index: In order to achieve full absorption of DG and reduce the power transmission of the cluster, a power balance index e is defined to describe the balance of power supply and demand within the cluster.

[0106] Considering the modularity index and power balance index, the cluster comprehensive partition index λ is expressed as:

[0107] λ=a1ρ+a2e (2)

[0108] Where: a1 and a2 are the weight coefficients of the two cluster division indicators.

[0109] The process of clustering the distribution network based on the SLM cluster partitioning algorithm is as follows: First, each substation in the distribution network is regarded as a cluster. Then, each substation is assigned to the cluster where the nodes of other substations are located in turn, and the difference in modularity before and after the assignment is calculated, and the position of the node with the largest difference is recorded. Repeat this process until the modularity no longer changes. Next, run the LMH algorithm to generate several subnetworks, regard each subnetwork as a node, and re-participate in the cluster allocation iteration process until the modularity no longer increases, that is, the optimal partitioning method is found.

[0110] In order to provide the goal of minimizing voltage deviation and power loss, build an optimization model for the control strategy, combine constraints, improve the power grid operation efficiency and control stability, in one embodiment of the coordinated optimization control method of multiple distribution areas in the distribution network that adapts to topology changes in this application, see Figure 3 , the optimization model including voltage deviation minimization and power loss minimization objective functions is constructed according to the cluster division result, including:

[0111] Step S102A: Based on the cluster division result, define objective functions with voltage deviation minimization and power loss minimization as optimization objectives, respectively, wherein the voltage deviation objective function represents the sum of square deviations between the voltage amplitudes of all nodes in the cluster and the voltage reference value, and the power loss objective function is calculated by the branch resistance in the cluster and the active power and reactive power flow;

[0112] Step S102B: According to the power balance principle within the cluster, the power balance relationship of each node is calculated based on the line admittance matrix and the voltage phase angle difference between the nodes to perform power balance constraints on the distribution network;

[0113] Step S102C: Based on the preset range, the voltage amplitudes of all nodes are controlled to be within the safety threshold range, and the branch current is limited to not exceed the maximum allowable current;

[0114] Step S102D: determining the active power output of the photovoltaic power generation system through the maximum power point tracking mode MPPT, and setting the reactive power adjustment range of the photovoltaic system according to the reactive power capacity of the photovoltaic power generation system;

[0115] Step S102E: Based on the preset range, the SOC state of the energy storage system at the current moment is controlled to be within the safety threshold range, and the charge and discharge power range of the energy storage device is set in combination with the charge and discharge efficiency of the energy storage system.

[0116] For example, the connection relationship between the low-voltage distribution station area and the feeder is as follows: Figure 4 As shown, the active power and reactive power to the outside world at each moment are interacted with the feeder layer within the substation.

[0117] The substation includes PV, energy storage system (ESS) and DC load and other equipment. The power balance equation of substation i is:

[0118]

[0119] Where: and are the active and reactive power flowing into the station area i from DG respectively; and are the charging and discharging power of energy storage respectively; and are the active and reactive power of load in area i respectively; and are the active and reactive power flowing into the substation respectively.

[0120] In order to fully tap the regulation capacity of adjustable resources in the distribution substation area, this application takes the voltage deviation and active network loss of the distribution network as the objective function, considers the constraints such as power balance, regulation resource output, node voltage safety, etc., and constructs a multi-distribution substation area collaborative optimization model. By optimizing the PV reactive output and ESS charging and discharging power in the substation area, the economy and safety of the distribution network operation are improved. Based on the cluster division results in the previous section, cluster C is constructed. m The objective function in is as follows:

[0121]

[0122]

[0123] Where: C is the number of the cluster, Ω clu is the set of all clusters; C m represents the node set of cluster m; T represents the total time period of the scheduling cycle; U i,t is the voltage amplitude of node i at time t, U ref is the voltage reference value; is the active network loss of cluster m at time t; Ω br,m is the branch set in cluster m, r ij is the resistance of branch ij, P ij,t and Q ij,t are the active power and reactive power flowing into branch ij respectively.

[0124] For constraints, the settings for this application are as follows:

[0125] Distribution network power balance constraints:

[0126]

[0127] Where: G ij and B ij are the real and imaginary parts of the admittance of line ij, θ ij is the voltage phase angle difference between nodes i and j.

[0128] Runtime safety constraints:

[0129]

[0130] Where: U max and U min are the safety upper and lower limits of the node voltage; is the maximum current allowed to flow through branch ij.

[0131] DG regulation performance constraints:

[0132] The photovoltaic active power adopts the maximum power point tracking (MPPT) mode. This paper considers the full absorption of distributed photovoltaics and only uses its reactive capacity for regulation. The reactive power regulation range of Hengshi in the power control mode is:

[0133]

[0134] Where: S PV,max is the maximum capacity of photovoltaic, P PV is the photovoltaic active power output, Q PV Provide reactive power for photovoltaic power generation.

[0135] The current state of charge (SOC) of the ESS is related to the SOC state and charge and discharge power at the previous moment. The ESS charge and discharge power constraints are:

[0136]

[0137] E i.t ≤E max (11)

[0138]

[0139] Where: E i.t represents the amount of electricity stored in the ESS in the station area i at time t, η ch and η dis are the ESS charging and discharging efficiencies, E max is the upper limit of storage power of ESS; It is the discharge status indicator of ESS. Indicates that the ESS is in the discharging state; and Represent the maximum value of ESS charging and discharging power respectively.

[0140] In order to adapt to the changes in cluster boundaries, adaptive control is achieved through the reward function to enhance the collaborative optimization capability of the distribution area. In one embodiment of the collaborative optimization control method of multiple distribution areas of the distribution network that adapts to topology changes in the present application, see Figure 5 , each station area is regarded as an intelligent agent, and a dynamic clustering Markov decision model for each station area is constructed according to the observation space, action space, reward function and state transfer of the intelligent agent, including:

[0141] Step S103A: each substation is regarded as an intelligent agent, and the observation space is defined as the real-time power flow information of all nodes in the substation according to the current cluster boundary and substation node status, including the active power, reactive power and voltage amplitude of the nodes, and the observation space is dynamically adjusted to adapt to the new cluster structure when the cluster boundary changes;

[0142] Step S103B: Based on the observation information of the intelligent agent, the action space is defined as the active power and reactive power regulation of the interaction between the substation and the feeder, including the power output or absorption achieved by photovoltaic and energy storage equipment, and the power of the substation nodes is redistributed after the action is executed;

[0143] Step S103C: constructing a reward function of the intelligent agent according to the control target, including a reward item based on the target function and a penalty item based on the constraint condition, wherein the reward item reflects the achievement of the control target by calculating the node voltage deviation and power loss, and the penalty item punishes the violation of the energy storage device constraint condition;

[0144] Step S103D: From the initial state until there is no feasible solution for the power flow or the scheduling cycle ends, the agent calculates the reward function value after performing actions according to the current observation space and action space, and enters the next moment state on this basis to perform state transfer.

[0145] For example, the coordinated optimization of multiple distribution areas in the distribution network makes decisions based on the operating status at the current moment and the previous moment, which can be regarded as a sequential decision-making problem. Since the electrical coupling between clusters is weak after cluster division, the clusters are autonomously controlled. Therefore, this paper models multiple distribution areas based on the multi-agent Markov decision model, with each distribution area as an agent and the distribution network as the training environment. The main components of the model are as follows:

[0146] 1) Agent observation space O i,t : The observation space represents the observation quantity on which the intelligent agent i makes decisions, which is the flow information of all nodes in the same cluster:

[0147]

[0148] Where: and is the active power and reactive power of the distribution network node connected to the substation i at time t; Ω m,SA is the set of all the area nodes in cluster m.

[0149] Considering the topology Figure 6 After the changes shown in the figure, the observation quantity of the agent also changes accordingly with the cluster boundary. Taking the three area agents i1, i2 and i3 in topology A as an example, their observation spaces are:

[0150]

[0151] When the topological structure changes from A to B, the observation spaces of the three intelligent agents in the station area become:

[0152]

[0153] 2) Agent action space A i,t :The action space represents the intelligent agent i in the area based on the observation O i,t The decision amount is:

[0154]

[0155] Where: and They are the active power and reactive power interacting between substation i and the feeder layer, which are realized by the photovoltaic and energy storage inverters on the substation side respectively. Indicates the active power generated by the area. Indicates the absorbed active power, Same reason.

[0156] Action A i,t After execution, the power of the distribution network node connected to the substation i is:

[0157]

[0158] 3) Reward function R i,t : The reward function is used to measure the agent i's set of observations-actions (O i,t ,A i,t ) value, the reward includes the objective function reward and the constraint reward. The larger the reward function, the greater the current (O i,t ,A i,t ) is closer to the optimal target, the inter-station relationship is a collaborative relationship, and the reward function is expressed as a negative objective function and penalty term:

[0159]

[0160] Where: is the energy storage capacity constraint violated in the action decision process of agent j, λ is the weight coefficient for balancing the target reward and constraint reward, and are all ESS penalty function coefficients.

[0161] 4) State transfer: Agent j changes state according to the current observation O i,t Execute action A j,t After that, we get the timely reward function R j,t And enter the next state, get the observation O at the next moment i,t This process is a state transfer, and the whole process starts from the initial moment and transfers the state along the time section until there is no feasible solution for the power flow or a scheduling cycle ends.

[0162] In order to combine centralized training with decentralized execution, reduce communication requirements, adapt to large-scale scenarios, and improve control efficiency and accuracy, in one embodiment of the coordinated optimization control method for multiple distribution areas of the distribution network that adapts to topology changes in the present application, see Figure 7 Based on the optimization model, the dynamic clustering Markov decision model is solved by using the multi-agent deep deterministic policy gradient MADDPG algorithm to obtain the corresponding distribution network multi-station area collaborative optimization control strategy, including:

[0163] Step S104A: construct an Actor network for each agent, and output corresponding active power and reactive power adjustment values ​​based on the observation value input of the agent;

[0164] Step S104B: construct a critic network for each agent, and evaluate the value of each action based on the observation and action set of all agents in the current state, wherein the output of the control critic network is based on the parameter update of the target critic network;

[0165] Step S104C: Set a target Actor network and a target Critic network for each agent, and regularly copy the parameters of the training network to the target network to maintain parameter stability in reinforcement learning;

[0166] Step S104D: In each iteration of training, based on the actions generated by the Actor network and the values ​​evaluated by the Critic network, the parameters of the Critic network are optimized by the gradient descent method to minimize the joint regression loss, and the parameters of the Actor network are optimized by the gradient ascent method to maximize the action value of the agent;

[0167] Step S104E: Obtain the corresponding distribution network multi-area coordinated optimization control strategy according to the iterative training result, and output the power regulation amount of the area based on the current observation space and the optimized action value.

[0168] For example, the action of the intelligent agent in the substation is the active or reactive power of DG and ESS, which are both continuously adjustable. Therefore, the MADDPG algorithm of the continuous action behavior space in multi-agent reinforcement learning is selected for solution. For each intelligent agent i in the substation, the MADDPG algorithm needs to maintain four neural networks: i,t Execute action A j,t Actor Network π i,t (O i,t ξ i ) and the target Actor network; the Critic network Q that evaluates the value of the action i,t (S t ,A t ; η i ) and the target Critic network, where St and A t It is represented as the observation-action set of agent i and other agents in the same cluster at time t, ξ i and η i The target network does not participate in network training, and the network parameters of the Actor and Critic participating in the training are regularly copied to the corresponding target network to stabilize the network training process.

[0169] The output of the Actor network is action A i,t , expressed as:

[0170]

[0171] Where: σ is Gaussian noise, and are the upper and lower limits of the action, that is, the upper and lower limits of the interactive power between the substation and the distribution network; Indicates taking a random number between the upper and lower limits of the action.

[0172] The output of the critic network is for S t and A t The value evaluation is based on the gradient descent to update the network parameters η i , Actor network training uses gradient ascent to maximize the current (O i,t ,A i,t ) value Q j (O i,t ,A i,t ).

[0173] In order to optimize value assessment by using environment and attention information, enhance feature extraction capability, dynamically adapt to complex environments, and optimize the stability and accuracy of the Critic network through joint regression, in one embodiment of the coordinated optimization and control method of multiple distribution areas in the distribution network that adapts to topology changes in the present application, see Figure 8 , the Critic network is constructed for each agent, and the value of each action is evaluated based on the observation and action set of all agents in the current state, wherein the output of the Critic network is controlled based on the parameter update of the target Critic network, including:

[0174] Step S104B1: inputting environmental information and attention information of the distribution network environment into the Critic network of each intelligent agent;

[0175] Step S104B2: Generate weights for the observation and action combination of each agent through the attention mechanism, calculate the correlation between the current agent and other agents using the key-value query method, and generate weighted parameters to describe the contribution of different agents to value assessment;

[0176] Step S104B3: Fuse the agent's environmental information and attention information, input the fused features into the Critic network, and generate the value output of the agent's action through the activation function and the fully connected network;

[0177] Step S104B4: wherein, a joint regression loss function is defined according to the output of the critic network and the output of the target critic network, and the critic network is updated by minimizing the loss function.

[0178] Among them, the critic network of the traditional MADDPG algorithm is composed of a fully connected neural network, and the number of neurons in its input layer is (S t ,A t ), once the Critic network is built, the input dimension cannot be changed, so it is limited to scenarios with a fixed number of agents under a fixed topology. When the topology changes, the cluster boundaries also expand or shrink, which brings about the problem of dynamic changes in the agent observation space as described in Section 3.2. To address this problem, this paper proposes an attention encoder critic network (AECN) based on the attention encoder, which maps the dynamic observation input to a fixed dimensional space through encoding, ensuring the dynamic scalability of the network. The network structure is shown in the figure below. Fig. 9 shown.

[0179] For example, for any agent i, the input of its AECN can be divided into two parts: one part is the environmental information h i,t =g i (O i,t ,A i,t ), which represents the observation-action combination of the distribution network environment (O i,t ,A i,t ) encoding, and the other part is the attention information h of the observation-action combination of other agents in the cluster j,t :

[0180]

[0181] Where: g j (O j,t ,A j,t ) represents the encoding of agent j, W k and W q For the mapping g j (O j,t ,A j,t ) to key value and map g i (O i,t ,A i,t ) query parameters, e ij,tCalculate the correlation coefficient of the observation-action combination of agents i and j. ij,t To ij,t The normalized attention weight coefficient represents the importance of the actions of other agents observed by agent i to the task.

[0182] The Critic network of agent i concatenates the two inputs and obtains the output Q i,t (S t ,A t ) is expressed as:

[0183]

[0184] Where: F i is a fully connected neural network, V is the function on g j (O j,t ,A j,t ), and h is the activation function.

[0185] The Critic network update is realized based on minimizing the joint regression function, and the loss function is:

[0186]

[0187] y i,t =R i,t +γQ i '(S t ',A t ') (26)

[0188] Where: E represents expectation, Q i '(S t ',A t ') is the output of the target Critic network.

[0189] It can be seen from the above embodiments that the present application provides a new solution for voltage regulation in low-perception and dynamic topology environments of distribution networks by combining dynamic cluster division with distributed autonomous optimization. In addition, by adopting the SLM algorithm for cluster division, the coordination ability of substations within the cluster is enhanced, and the impact of power flow across clusters on overall regulation is reduced.

[0190] The introduction of feature extraction network in the reinforcement learning model enables the control strategy to adapt to the frequent adjustment of the topology structure, and optimizes the control strategy generation process between stations. The centralized training and decentralized execution method reduces the dependence on the whole network communication, which helps to maintain the control function under the condition of data loss.

[0191] This application can support large-scale distribution network scenarios, meet the requirements of real-time, adaptability and accuracy, and provide a basis for the coordinated optimization of multiple areas in complex operating environments.

[0192] In order to combine dynamic cluster division and distributed autonomous optimization, the SLM algorithm is used to enhance the coordination ability of the substations and reduce the impact of cross-cluster power flow; by introducing a feature extraction network to adapt to topology changes, optimize the generation of control strategies; use centralized training and decentralized execution to reduce communication dependence, and achieve efficient voltage control in low-perception scenarios in complex environments of large-scale distribution networks. This application provides an embodiment of a distribution network multi-station area collaborative optimization control device that adapts to topology changes and is used to implement all or part of the content of the distribution network multi-station area collaborative optimization control method that adapts to topology changes, see Fig.10 The distribution network multi-station area coordinated optimization control device that adapts to topology changes specifically includes the following contents:

[0193] The cluster division module 1101 is used to: divide the distribution network into clusters by using the SLM cluster division algorithm based on the electrical distance modularity and the power balance degree;

[0194] The optimization model construction module 1102 is used to: construct an optimization model including voltage deviation minimization and power loss minimization objective functions according to the cluster division results, wherein the objective function of the optimization model includes the calculation of node voltage deviation and cluster power loss within the cluster, and the constraint conditions include power balance, node voltage safety, photovoltaic reactive power adjustment range and preset range of SOC state of energy storage equipment;

[0195] The agent modeling module 1103 is used to: regard each station area as an agent, and construct a dynamic clustering Markov decision model for each station area according to the observation space, action space, reward function and setting state transfer of the agent;

[0196] The optimization solution module 1104 is used to: based on the optimization model, use the multi-agent deep deterministic policy gradient MADDPG algorithm to solve the dynamic clustering Markov decision model, obtain the corresponding distribution network multi-station area collaborative optimization control strategy, and control the distribution network according to the control strategy.

[0197] According to any implementation manner of the present application, the cluster division module includes:

[0198] The cluster initialization unit is used to: regard each substation in the distribution network as an independent initial cluster, and obtain the node connection relationship and load power data of all substations;

[0199] An iterative optimization unit is used to: use the predefined electrical distance modularity and power balance as indicators for cluster division, try to allocate each substation to other clusters in each iteration, calculate the change value of the electrical distance modularity after allocation, record the allocation method that produces the maximum modularity gain, and repeat this process to gradually optimize the cluster division until the modularity no longer changes;

[0200] The subnetwork merging unit is used to generate several subnetworks through the SLM cluster partitioning algorithm, regard each subnetwork as a virtual node to participate in the next round of cluster partitioning, and continue to optimize the modularity until it reaches the global optimum.

[0201] According to any implementation of the present application, the optimization model building module includes:

[0202] An objective function definition unit is used to: define objective functions based on the cluster division result, taking voltage deviation minimization and power loss minimization as optimization objectives, respectively, wherein the voltage deviation objective function represents the sum of square deviations between the voltage amplitudes of all nodes in the cluster and the voltage reference value, and the power loss objective function is calculated by branch resistances in the cluster and the active power and reactive power flows thereof;

[0203] The power balance constraint unit is used to: calculate the power balance relationship of each node based on the line admittance matrix and the voltage phase angle difference between nodes according to the power balance principle within the cluster to perform power balance constraint on the distribution network;

[0204] A safety constraint unit is used to: control the voltage amplitude of all nodes within a safety threshold range based on a preset range, and limit the branch current to not exceed the maximum allowed current;

[0205] A photovoltaic regulation unit is used to: determine the active power output of the photovoltaic power generation system through the maximum power tracking mode MPPT, and set the reactive power regulation range of the photovoltaic system according to the reactive power capacity of the photovoltaic power generation system;

[0206] The energy storage management unit is used to: based on a preset range, control the SOC state of the energy storage system at the current moment to be within a safe threshold range, and set the charge and discharge power range of the energy storage device in combination with the charge and discharge efficiency of the energy storage system.

[0207] According to any embodiment of the present application, the agent modeling module includes:

[0208] The observation space definition unit is used to: regard each substation as an intelligent agent, and define the observation space as the real-time power flow information of all nodes in the substation according to the current cluster boundary and substation node status, including the active power, reactive power and voltage amplitude of the nodes, and dynamically adjust the observation space to adapt to the new cluster structure when the cluster boundary changes;

[0209] An action space definition unit is used to: define the action space as the active power and reactive power regulation of the interaction between the substation and the feeder based on the observation information of the intelligent agent, including the power output or absorption achieved by photovoltaic and energy storage equipment, and redistribute the power of the substation nodes after the action is executed;

[0210] A reward function construction unit, used to: construct a reward function of the intelligent agent according to the control target, including a reward item based on the target function and a penalty item based on the constraint condition, wherein the reward item reflects the achievement of the control target by calculating the node voltage deviation and power loss, and the penalty item punishes the violation of the energy storage device constraint condition;

[0211] The state transfer unit is used to: from the initial state until there is no feasible solution for the power flow or the scheduling cycle ends, calculate the reward function value after the agent performs actions according to the current observation space and action space, and enter the next moment state on this basis to perform state transfer.

[0212] According to any implementation of the present application, the optimization solution module includes:

[0213] Actor network construction unit, used to: construct an Actor network for each intelligent agent, based on the observation value input of the intelligent agent, output the corresponding active power and reactive power adjustment amount;

[0214] The critic network construction unit is used to: construct a critic network for each agent, evaluate the value of each action based on the set of observations and actions of all agents in the current state, and control the output of the critic network based on the parameter update of the target critic network;

[0215] The target network setting unit is used to set the target actor network and target critic network for each agent, and regularly copy the parameters of the training network to the target network to maintain parameter stability in reinforcement learning;

[0216] The network parameter optimization unit is used to: optimize the parameters of the Critic network by gradient descent to minimize the joint regression loss, and optimize the parameters of the Actor network by gradient ascent to maximize the action value of the agent in each training iteration based on the actions generated by the Actor network and the values ​​evaluated by the Critic network;

[0217] The control strategy generation unit is used to obtain the corresponding distribution network multi-station area collaborative optimization control strategy according to the iterative training results, and output the power regulation amount of the station area based on the current observation space and the optimized action value.

[0218] According to any implementation of the present application, the Critic network construction unit includes:

[0219] An environmental information input unit is used to: input environmental information and attention information of the distribution network environment into the Critic network of each intelligent agent;

[0220] An attention weight generation unit, used to: generate a weight for the observation and action combination of each agent through an attention mechanism, calculate the correlation between the current agent and other agents using a key-value query device, and generate weighted parameters to describe the contribution of different agents to value assessment;

[0221] The feature fusion unit is used to fuse the agent's environmental information and attention information, input the fused features into the Critic network, and generate the value output of the agent's actions through the activation function and the fully connected network;

[0222] The network updating unit is used to define a joint regression loss function according to the output of the critic network and the output of the target critic network, and update the critic network by minimizing the loss function.

[0223] From the above description, it can be seen that the distribution network multi-area collaborative optimization and control device that adapts to topology changes provided in the embodiment of the present application can combine dynamic cluster division and distributed autonomous optimization, adopt the SLM algorithm to enhance the area coordination capability and reduce the impact of cross-cluster power flow; adapt to topology changes by introducing a feature extraction network, and optimize the control strategy generation; use centralized training and decentralized execution to reduce communication dependence, and realize efficient voltage control in low-perception scenarios in the complex environment of large-scale distribution networks.

[0224] From the hardware level, in order to combine dynamic cluster division and distributed autonomous optimization, the SLM algorithm is used to enhance the coordination capability of the substation area and reduce the impact of cross-cluster power flow; by introducing a feature extraction network to adapt to topology changes, optimize the generation of control strategies; using centralized training and decentralized execution to reduce communication dependence, and achieve efficient voltage control in low-perception scenarios in complex environments of large-scale distribution networks. This application provides an embodiment of an electronic device for implementing all or part of the content of the method for coordinated optimization and control of multiple distribution areas of a distribution network that adapts to topology changes, and the electronic device specifically includes the following content:

[0225] Processor, memory, communication interface and bus; wherein the processor, memory and communication interface communicate with each other through the bus; the communication interface is used to realize information transmission between the collaborative optimization and control device of multiple distribution areas in the distribution network that adapts to topology changes and related equipment such as core business systems, user terminals and related databases; the logic controller can be a desktop computer, a tablet computer and a mobile terminal, etc., but the present embodiment is not limited thereto. In the present embodiment, the logic controller can be implemented with reference to the embodiment of the collaborative optimization and control method of multiple distribution areas in the distribution network that adapts to topology changes and the embodiment of the collaborative optimization and control device of multiple distribution areas in the distribution network that adapts to topology changes in the embodiment, and the contents thereof are incorporated herein, and the repeated parts are not repeated.

[0226] It is understandable that the user terminal may include a smart phone, a tablet electronic device, a network set-top box, a portable computer, a desktop computer, a personal digital assistant (PDA), a vehicle-mounted device, a smart wearable device, etc. Among them, the smart wearable device may include smart glasses, a smart watch, a smart bracelet, etc.

[0227] In practical applications, part of the method for collaborative optimization and control of multiple distribution network areas that adapt to topology changes can be executed on the electronic device side as described above, or all operations can be completed in the client device. The specific selection can be based on the processing capability of the client device and the limitations of the user's usage scenario. This application does not limit this. If all operations are completed in the client device, the client device may also include a processor.

[0228] The client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission with the server. The server may include a server on the task scheduling center side, and other implementation scenarios may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, or a server cluster consisting of multiple servers, or a server structure of a distributed device.

[0229] Fig.11 FIG. 9 is a schematic block diagram of the system structure of the electronic device 9600 according to an embodiment of the present application. Fig.11 As shown, the electronic device 9600 may include a central processor 9100 and a memory 9140; the memory 9140 is coupled to the central processor 9100. It is worth noting that Fig.11 is exemplary; other types of structures may also be used to supplement or replace this structure to implement telecommunication functions or other functions.

[0230] In one embodiment, the function of the coordinated optimization control method for multiple distribution network areas adapting to topology changes can be integrated into the central processor 9100. The central processor 9100 can be configured to perform the following control:

[0231] Step S101: clustering the distribution network using the SLM clustering algorithm based on electrical distance modularity and power balance;

[0232] Step S102: constructing an optimization model including voltage deviation minimization and power loss minimization objective functions according to the cluster division results, wherein the objective function of the optimization model includes calculation of node voltage deviation within the cluster and cluster power loss, and the constraint conditions include power balance, node voltage safety, photovoltaic reactive power adjustment range and preset range of SOC state of energy storage equipment;

[0233] Step S103: taking each area as an intelligent agent, constructing a dynamic clustering Markov decision model for each area according to the observation space, action space, reward function and setting state transition of the intelligent agent;

[0234] Step S104: Based on the optimization model, the dynamic clustering Markov decision model is solved using the multi-agent deep deterministic policy gradient MADDPG algorithm to obtain the corresponding distribution network multi-station area collaborative optimization control strategy, and the distribution network is controlled according to the control strategy.

[0235] From the above description, it can be seen that the electronic device provided in the embodiment of the present application can combine dynamic cluster division and distributed autonomous optimization, adopt the SLM algorithm to enhance the coordination capability of the substation area and reduce the impact of cross-cluster power flow; adapt to topology changes by introducing a feature extraction network, and optimize the generation of control strategies; use centralized training and decentralized execution to reduce communication dependence, and realize efficient voltage control in low-perception scenarios in the complex environment of large-scale distribution networks.

[0236] In another embodiment, the device for collaborative optimization and control of multiple distribution areas in a distribution network that adapts to topology changes can be configured separately from the central processor 9100. For example, the device for collaborative optimization and control of multiple distribution areas in a distribution network that adapts to topology changes can be configured as a chip connected to the central processor 9100, and the function of the method for collaborative optimization and control of multiple distribution areas in a distribution network that adapts to topology changes can be implemented through the control of the central processor.

[0237] like Fig.11 As shown, the electronic device 9600 may also include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily have to include Fig.11In addition, the electronic device 9600 may also include Fig.11 For components not shown, reference may be made to the prior art.

[0238] like Fig.11 As shown, the central processing unit 9100 is sometimes also referred to as a controller or an operation control, and may include a microprocessor or other processor device and / or logic device. The central processing unit 9100 receives input and controls the operation of various components of the electronic device 9600.

[0239] The memory 9140 may be, for example, one or more of a cache, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory or other suitable devices. The above-mentioned information related to the failure may be stored, and a program for executing the relevant information may also be stored. The CPU 9100 may execute the program stored in the memory 9140 to implement information storage or processing, etc.

[0240] The input unit 9120 provides input to the central processing unit 9100. The input unit 9120 is, for example, a key or a touch input device. The power supply 9170 is used to provide power to the electronic device 9600. The display 9160 is used to display display objects such as images and texts. The display may be, for example, an LCD display, but is not limited thereto.

[0241] The memory 9140 may be a solid-state memory, such as a read-only memory (ROM), a random access memory (RAM), a SIM card, etc. It may also be a memory that saves information even when the power is off, can be selectively erased, and is provided with more data, examples of which are sometimes referred to as EPROMs, etc. The memory 9140 may also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application / function storage unit 9142, which is used to store application programs and function programs or processes for executing the operation of the electronic device 9600 through the central processor 9100.

[0242] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various drivers for communication functions of the electronic device and / or for executing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0243] The communication module 9110 is a transmitter / receiver that sends and receives signals via the antenna 9111. The communication module (transmitter / receiver) 9110 is coupled to the central processor 9100 to provide input signals and receive output signals, which may be the same as the case of a conventional mobile communication terminal.

[0244] Based on different communication technologies, multiple communication modules 9110 may be provided in the same electronic device, such as a cellular network module, a Bluetooth module and / or a wireless LAN module, etc. The communication module (transmitter / receiver) 9110 is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide an audio output via the speaker 9131 and receive an audio input from the microphone 9132, thereby realizing a common telecommunication function. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor 9130 is also coupled to the central processor 9100, so that recording can be performed on the local machine through the microphone 9132, and the sound stored on the local machine can be played through the speaker 9131.

[0245] The embodiments of the present application also provide a computer-readable storage medium capable of implementing all the steps in the method for collaborative optimization and control of multiple distribution network areas that adapt to topology changes in the above-mentioned embodiments, where the execution subject is a server or a client. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, all the steps of the method for collaborative optimization and control of multiple distribution network areas that adapt to topology changes in the above-mentioned embodiments are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0246] Step S101: clustering the distribution network using the SLM clustering algorithm based on electrical distance modularity and power balance;

[0247] Step S102: constructing an optimization model including voltage deviation minimization and power loss minimization objective functions according to the cluster division results, wherein the objective function of the optimization model includes calculation of node voltage deviation within the cluster and cluster power loss, and the constraint conditions include power balance, node voltage safety, photovoltaic reactive power adjustment range and preset range of SOC state of energy storage equipment;

[0248] Step S103: taking each area as an intelligent agent, constructing a dynamic clustering Markov decision model for each area according to the observation space, action space, reward function and setting state transition of the intelligent agent;

[0249] Step S104: Based on the optimization model, the dynamic clustering Markov decision model is solved using the multi-agent deep deterministic policy gradient MADDPG algorithm to obtain the corresponding distribution network multi-station area collaborative optimization control strategy, and the distribution network is controlled according to the control strategy.

[0250] From the above description, it can be seen that the computer-readable storage medium provided in the embodiment of the present application can combine dynamic cluster division and distributed autonomous optimization, adopt the SLM algorithm to enhance the coordination capability of the substation area and reduce the impact of cross-cluster power flow; adapt to topology changes by introducing a feature extraction network, and optimize the generation of control strategies; use centralized training and decentralized execution to reduce communication dependence, and realize efficient voltage control in low-perception scenarios in the complex environment of large-scale distribution networks.

[0251] The embodiments of the present application also provide a computer program product capable of implementing all steps of the method for collaborative optimization and control of multiple distribution network areas adapting to topology changes in the above embodiments, where the execution subject is a server or a client. When the computer program / instruction is executed by a processor, the steps of the method for collaborative optimization and control of multiple distribution network areas adapting to topology changes are implemented. For example, the computer program / instruction implements the following steps:

[0252] Step S101: clustering the distribution network using the SLM clustering algorithm based on electrical distance modularity and power balance;

[0253] Step S102: constructing an optimization model including voltage deviation minimization and power loss minimization objective functions according to the cluster division results, wherein the objective function of the optimization model includes calculation of node voltage deviation within the cluster and cluster power loss, and the constraint conditions include power balance, node voltage safety, photovoltaic reactive power adjustment range and preset range of SOC state of energy storage equipment;

[0254] Step S103: taking each area as an intelligent agent, constructing a dynamic clustering Markov decision model for each area according to the observation space, action space, reward function and setting state transition of the intelligent agent;

[0255] Step S104: Based on the optimization model, the dynamic clustering Markov decision model is solved using the multi-agent deep deterministic policy gradient MADDPG algorithm to obtain the corresponding distribution network multi-station area collaborative optimization control strategy, and the distribution network is controlled according to the control strategy.

[0256] From the above description, it can be seen that the computer program product provided in the embodiment of the present application can combine dynamic cluster division and distributed autonomous optimization, adopt the SLM algorithm to enhance the coordination capability of the substation area and reduce the impact of cross-cluster power flow; adapt to topology changes by introducing a feature extraction network, and optimize the generation of control strategies; use centralized training and decentralized execution to reduce communication dependence, and realize efficient voltage control in low-perception scenarios in the complex environment of large-scale distribution networks.

[0257] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0258] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (apparatus), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0259] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0260] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0261] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for coordinated optimization and control of multiple distribution areas in a distribution network that adapts to topology changes, characterized in that: The method comprises: Based on the electrical distance modularity and power balance, the distribution network is clustered using the SLM clustering algorithm. An optimization model including voltage deviation minimization and power loss minimization objective functions is constructed according to the cluster division results, wherein the objective function of the optimization model includes the calculation of node voltage deviation within the cluster and cluster power loss, and the constraints include power balance, node voltage safety, photovoltaic reactive power adjustment range, and preset range of SOC state of energy storage equipment; Take each area as an intelligent agent, and build a dynamic clustering Markov decision model for each area according to the observation space, action space, reward function and state transition of the intelligent agent; Based on the optimization model, the dynamic clustering Markov decision model is solved using the multi-agent deep deterministic policy gradient MADDPG algorithm to obtain the corresponding distribution network multi-station collaborative optimization control strategy, and the distribution network is regulated according to the control strategy.

2. The method for coordinated optimization and control of multiple distribution network areas that adapts to topology changes according to claim 1 is characterized in that: The method of clustering the distribution network based on the electrical distance modularity and the power balance by using the SLM cluster partitioning algorithm includes: Each substation in the distribution network is regarded as an independent initial cluster, and the node connection relationship and load power data of all substations are obtained; The predefined electrical distance modularity and power balance are used as indicators for cluster division. In each iteration, each substation is tried to be assigned to other clusters. The change in electrical distance modularity after assignment is calculated, and the assignment method that produces the maximum modularity gain is recorded. This process is repeated to gradually optimize the cluster division until the modularity no longer changes. Several sub-networks are generated through the SLM cluster partitioning algorithm, and each sub-network is regarded as a virtual node to participate in the next round of cluster partitioning, and the modularity is continuously optimized until it reaches the global optimum.

3. The method for coordinated optimization and control of multiple distribution network areas adapting to topology changes according to claim 1 is characterized in that: The optimization model including voltage deviation minimization and power loss minimization objective functions is constructed according to the cluster division result, including: Based on the cluster division results, objective functions are defined with voltage deviation minimization and power loss minimization as optimization objectives, respectively. The voltage deviation objective function represents the sum of square deviations between the voltage amplitudes of all nodes in the cluster and the voltage reference value, and the power loss objective function is calculated by the branch resistance and its active power and reactive power flow in the cluster. According to the power balance principle within the cluster, the power balance relationship of each node is calculated based on the line admittance matrix and the voltage phase angle difference between nodes to perform power balance constraints on the distribution network; Based on the preset range, the voltage amplitude of all nodes is controlled within the safe threshold range, and the branch current is limited to not exceed the maximum allowed current; Determine the active power output of the photovoltaic power generation system through the maximum power point tracking mode MPPT, and set the reactive power adjustment range of the photovoltaic system according to the reactive power capacity of the photovoltaic power generation system; Based on the preset range, the SOC state of the energy storage system at the current moment is controlled to be within the safety threshold range, and the charging and discharging power range of the energy storage device is set in combination with the charging and discharging efficiency of the energy storage system.

4. The method for coordinated optimization and control of multiple distribution network areas adapting to topology changes according to claim 1 is characterized in that: Each station area is regarded as an intelligent agent, and a dynamic clustering Markov decision model for each station area is constructed according to the observation space, action space, reward function and state transition of the intelligent agent, including: Each substation is regarded as an intelligent entity, and the observation space is defined as the real-time power flow information of all nodes in the substation according to the current cluster boundary and substation node status, including the active power, reactive power and voltage amplitude of the nodes, and the observation space is dynamically adjusted to adapt to the new cluster structure when the cluster boundary changes; Based on the observation information of the intelligent agent, the action space is defined as the active power and reactive power regulation of the interaction between the substation and the feeder, including the power output or absorption achieved by photovoltaic and energy storage equipment, and the power of the substation nodes is redistributed after the action is executed; Constructing a reward function of the intelligent agent according to the control target, including a reward item based on the target function and a penalty item based on the constraint condition, wherein the reward item reflects the achievement of the control target by calculating the node voltage deviation and power loss, and the penalty item punishes the violation of the energy storage device constraint condition; Among them, from the initial state until there is no feasible solution for the power flow or the scheduling cycle ends, the agent calculates the reward function value after performing actions according to the current observation space and action space, and enters the next moment state on this basis to perform state transfer.

5. The method for coordinated optimization and control of multiple distribution network areas adapting to topology changes according to claim 1 is characterized in that: Based on the optimization model, the dynamic clustering Markov decision model is solved using a multi-agent deep deterministic policy gradient MADDPG algorithm to obtain a corresponding distribution network multi-station area collaborative optimization control strategy, including: Build an Actor network for each agent, and output the corresponding active power and reactive power regulation based on the observation value input of the agent; Construct a critic network for each agent, and evaluate the value of each action based on the set of observations and actions of all agents in the current state, where the output of the control critic network is based on the parameter update of the target critic network; Set a target Actor network and a target Critic network for each agent, and regularly copy the parameters of the training network to the target network to maintain parameter stability in reinforcement learning; In each training iteration, based on the actions generated by the Actor network and the values ​​evaluated by the Critic network, the parameters of the Critic network are optimized by gradient descent to minimize the joint regression loss, and the parameters of the Actor network are optimized by gradient ascent to maximize the action value of the agent. According to the iterative training results, the corresponding distribution network multi-station coordinated optimization control strategy is obtained, and the power regulation amount of the station is output based on the current observation space and the optimized action value.

6. The method for coordinated optimization and control of multiple distribution network areas adapting to topology changes according to claim 5 is characterized in that: The critic network is constructed for each agent, and the value of each action is evaluated based on the observation and action set of all agents in the current state, wherein the output of the critic network is controlled based on the parameter update of the target critic network, including: Input the environmental information and attention information of the distribution network environment to the Critic network of each intelligent agent; The attention mechanism is used to generate weights for the observation and action combination of each agent, and the key-value query method is used to calculate the correlation between the current agent and other agents, and the weighted parameters are generated to describe the contribution of different agents to value assessment. The agent's environmental information and attention information are fused, and the fused features are input into the Critic network. Through the activation function and the fully connected network, the value output of the agent's action is generated; Among them, according to the output of the critic network and the output of the target critic network, a joint regression loss function is defined, and the critic network is updated by minimizing the loss function.

7. A distribution network multi-zone coordinated optimization control device that adapts to topology changes, characterized in that: The device comprises: The cluster partitioning module is used to: cluster the distribution network based on the electrical distance modularity and power balance by using the SLM cluster partitioning algorithm; An optimization model construction module is used to: construct an optimization model including voltage deviation minimization and power loss minimization objective functions according to the cluster division results, wherein the objective function of the optimization model includes the calculation of node voltage deviation and cluster power loss within the cluster, and the constraint conditions include power balance, node voltage safety, photovoltaic reactive power adjustment range and preset range of SOC state of energy storage equipment; The agent modeling module is used to: treat each station as an agent, and construct a dynamic clustering Markov decision model for each station according to the observation space, action space, reward function and state transfer of the agent; The optimization solution module is used to: based on the optimization model, use the multi-agent deep deterministic policy gradient MADDPG algorithm to solve the dynamic clustering Markov decision model, obtain the corresponding distribution network multi-station area collaborative optimization control strategy, and control the distribution network according to the control strategy.

8. The distribution network multi-zone coordinated optimization control device that adapts to topology changes according to claim 7 is characterized in that: The cluster division module comprises: The cluster initialization unit is used to: regard each substation in the distribution network as an independent initial cluster, and obtain the node connection relationship and load power data of all substations; An iterative optimization unit is used to: use the predefined electrical distance modularity and power balance as indicators for cluster division, try to allocate each substation to other clusters in each iteration, calculate the change value of the electrical distance modularity after allocation, record the allocation method that produces the maximum modularity gain, and repeat this process to gradually optimize the cluster division until the modularity no longer changes; The subnetwork merging unit is used to generate several subnetworks through the SLM cluster partitioning algorithm, regard each subnetwork as a virtual node to participate in the next round of cluster partitioning, and continue to optimize the modularity until it reaches the global optimum.

9. The distribution network multi-zone coordinated optimization control device that adapts to topology changes according to claim 7 is characterized in that: The optimization model building module includes: An objective function definition unit is used to: define objective functions based on the cluster division result, taking voltage deviation minimization and power loss minimization as optimization objectives, respectively, wherein the voltage deviation objective function represents the sum of square deviations between the voltage amplitudes of all nodes in the cluster and the voltage reference value, and the power loss objective function is calculated by branch resistances in the cluster and the active power and reactive power flows thereof; The power balance constraint unit is used to: calculate the power balance relationship of each node based on the line admittance matrix and the voltage phase angle difference between nodes according to the power balance principle within the cluster to perform power balance constraint on the distribution network; A safety constraint unit is used to: control the voltage amplitude of all nodes within a safety threshold range based on a preset range, and limit the branch current to not exceed the maximum allowed current; A photovoltaic regulation unit is used to: determine the active power output of the photovoltaic power generation system through the maximum power tracking mode MPPT, and set the reactive power regulation range of the photovoltaic system according to the reactive power capacity of the photovoltaic power generation system; The energy storage management unit is used to: based on a preset range, control the SOC state of the energy storage system at the current moment to be within a safe threshold range, and set the charge and discharge power range of the energy storage device in combination with the charge and discharge efficiency of the energy storage system.

10. The distribution network multi-zone coordinated optimization control device that adapts to topology changes according to claim 7 is characterized in that: The agent modeling module includes: The observation space definition unit is used to: regard each substation as an intelligent agent, and define the observation space as the real-time power flow information of all nodes in the substation according to the current cluster boundary and substation node status, including the active power, reactive power and voltage amplitude of the nodes, and dynamically adjust the observation space to adapt to the new cluster structure when the cluster boundary changes; An action space definition unit is used to: define the action space as the active power and reactive power regulation of the interaction between the substation and the feeder based on the observation information of the intelligent agent, including the power output or absorption achieved by photovoltaic and energy storage equipment, and redistribute the power of the substation nodes after the action is executed; A reward function construction unit, used to: construct a reward function of the intelligent agent according to the control target, including a reward item based on the target function and a penalty item based on the constraint condition, wherein the reward item reflects the achievement of the control target by calculating the node voltage deviation and power loss, and the penalty item punishes the violation of the energy storage device constraint condition; The state transfer unit is used to: from the initial state until there is no feasible solution for the power flow or the scheduling cycle ends, calculate the reward function value after the agent performs actions according to the current observation space and action space, and enter the next moment state on this basis to perform state transfer.

11. The distribution network multi-zone coordinated optimization control device that adapts to topology changes according to claim 7 is characterized in that: The optimization solution module includes: Actor network construction unit, used to: construct an Actor network for each intelligent agent, based on the observation value input of the intelligent agent, output the corresponding active power and reactive power adjustment amount; The critic network construction unit is used to: construct a critic network for each agent, evaluate the value of each action based on the set of observations and actions of all agents in the current state, and control the output of the critic network based on the parameter update of the target critic network; The target network setting unit is used to set the target actor network and target critic network for each agent, and regularly copy the parameters of the training network to the target network to maintain parameter stability in reinforcement learning; The network parameter optimization unit is used to: optimize the parameters of the Critic network by gradient descent to minimize the joint regression loss, and optimize the parameters of the Actor network by gradient ascent to maximize the action value of the agent in each training iteration based on the actions generated by the Actor network and the values ​​evaluated by the Critic network; The control strategy generation unit is used to obtain the corresponding distribution network multi-station area collaborative optimization control strategy according to the iterative training results, and output the power regulation amount of the station area based on the current observation space and the optimized action value.

12. The distribution network multi-zone coordinated optimization control device that adapts to topology changes according to claim 11 is characterized in that: The Critic network construction unit includes: An environmental information input unit is used to: input environmental information and attention information of the distribution network environment into the Critic network of each intelligent agent; An attention weight generation unit, used to: generate weights for the observation and action combination of each agent through an attention mechanism, calculate the correlation between the current agent and other agents using a key-value query device, and generate weighted parameters to describe the contribution of different agents to value assessment; The feature fusion unit is used to fuse the agent's environmental information and attention information, input the fused features into the Critic network, and generate the value output of the agent's actions through the activation function and the fully connected network; The network updating unit is used to define a joint regression loss function according to the output of the critic network and the output of the target critic network, and update the critic network by minimizing the loss function.

13. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method for collaborative optimization and control of multiple distribution network areas that adapt to topology changes as described in any one of claims 1 to 6 are implemented.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for coordinated optimization and control of multiple distribution network areas that adapt to topology changes as described in any one of claims 1 to 6 are implemented.

15. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps of the method for coordinated optimization and control of multiple distribution network areas that adapt to topology changes as described in any one of claims 1 to 6 are implemented.

Images