Main distribution coordinated voltage optimization control method and equipment under communication and electrical coupling

Abstract

The present invention discloses a method and device for optimizing voltage control of main distribution coordination under communication and electrical coupling. The method is based on a model prediction rolling optimization framework, takes voltage deviation minimization as the goal, and constructs a distributed voltage control model under communication and electrical coupling; decomposes the distributed voltage control model into sub-problems of adjacent partition information interaction and multiple partition parallel control optimization and reconstructs it into a partially observable Markov decision process; uses a multi-agent reinforcement learning algorithm to quickly solve the reconstructed model and determine the voltage control method. The present invention can improve the voltage regulation accuracy and real-time performance, reduce the communication bandwidth requirements, and make the voltage under main distribution coordination run more safely and stably.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system automation, and specifically relates to a main distribution coordinated voltage optimization control method and equipment under communication and electrical coupling. Background Art

[0002] As the scale of renewable energy power generation in my country has shown explosive growth, it has the characteristics of both distributed and centralized. However, with the continuous increase in the penetration rate of renewable energy, its large amount, dispersed and highly volatile characteristics have also brought severe challenges to the operation and control of the distribution network. The imbalance between power generation and load demand will change the distribution of the power flow in the distribution network, causing local absorption difficulties, while also causing terminal voltage to rise and network losses to increase. The risk of power outages and power outages caused by voltage over-limit has increased, and the perception and operation of the distribution network has become more complex. The operation and control data that needs to be transmitted has increased exponentially, which has posed a challenge to the allocation of communication resources in 5G networks.

[0003] Voltage control in active distribution networks is divided into centralized control, local control and distributed control. Centralized control requires a communication network for coordination, and the limited bandwidth and sudden communication interference of the distribution network will affect its performance and stability. Time delays will cause control signal lags and reduce dynamic performance, especially in low-inertia grids with renewable energy access. Different communication channels, such as optical fiber, wireless, etc., provide communication services with different delays, ranging from tens of milliseconds to hundreds of milliseconds. Local control is considered to be the main solution for handling rapid voltage fluctuations with communication independence and fast response. But in some cases, they may not ensure that the voltage remains within an acceptable range, and uncoordinated interactions may even aggravate stability and grid security issues. Distributed control exchanges information with adjacent intelligent agents and can make full use of the rapid response capabilities of photovoltaics.

[0004] Data-driven methods can be used to solve problems where the model is incomplete or difficult to construct mathematically, and have significant advantages in decision-making speed. Multi-agent reinforcement learning can treat each regulating device as an independent agent and find the optimal action strategy through full cooperation between them. Related papers studied the reconstruction of the multi-region coordinated voltage / reactive power control optimization problem as a partially observable Markov game and used the adjusted multi-agent deep deterministic policy gradient algorithm to solve it. However, the impact of communication transmission delay on the distributed voltage control of the distribution network was not considered, and it is not suitable for distributed voltage control under the coupling of adjustable resources. Summary of the invention

[0005] The present invention proposes a method and device for optimizing the control of main and distribution coordinated voltage under communication and electrical coupling, which can at least solve one of the technical problems in the background technology.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] A method for optimizing the coordinated voltage of a main distribution system under communication and electrical coupling comprises the following steps:

[0008] S100, based on the active distribution network topology structure, the area is divided, the network parameters of each partition are collected through the equipment, and the distribution network power flow model and voltage sensitivity matrix are constructed;

[0009] S200, constructing a communication transmission network based on variable delay;

[0010] S300, constructing a distributed voltage control model of communication and power coupling according to the distribution network power flow model, the communication transmission network and the voltage sensitivity matrix, establishing an optimization goal of minimizing the node voltage disturbance deviation and the control cost, and setting control constraints;

[0011] S400, decomposing the distributed voltage control model into sub-models for information interaction between adjacent partitions and parallel control optimization of multiple partitions, and reconstructing the sub-models using a partially observable Markov model;

[0012] S500, performing offline training on the reconstructed model using a multi-agent reinforcement learning algorithm to determine a voltage control method.

[0013] Furthermore, the method for constructing the power flow model and voltage sensitivity matrix of the distribution network in step S100 of the present invention includes:

[0014] S110, assuming There are loads and photovoltaics on each node, and the z constraint equation is constructed;

[0015] The power flow constraint equation of the active distribution network is as follows:

[0016]

[0017] in, Indicates The active power of photovoltaic power on each node, Indicates The reactive power of photovoltaic power on each node; Indicates The active power of the load on each node, Indicates The reactive power of the load on each node; and Respectively represent The node and The voltage of each node; and Respectively expressed as The node and The real and imaginary parts of the conductance and susceptance between nodes; Indicates The node and The voltage phase angle difference between nodes;

[0018] S120, solve the node voltage according to the active distribution network power flow constraint equation Photovoltaic active power The sensitivity matrix and node voltage Reactive power of photovoltaic The sensitivity matrix :

[0019]

[0020] in, , They are the partial derivative matrices of node voltage amplitude with respect to active and reactive power respectively.

[0021] Furthermore, the communication transmission network construction method in step S200 of the present invention includes:

[0022] S210, selecting a plurality of transmission channels with different constant delays;

[0023] Assume that the resource-limited communication network is a transmission server with variable delay;

[0024] use Represents multiple transmission channels with different constant delays. The specific formula is as follows:

[0025]

[0026] in, Represents bounded delay Channel;

[0027] S220, solving the transmission capacity constraint of the communication network according to the transmission channels with different constant delays;

[0028] (4)

[0029] in, Indicates channel transmission capacity.

[0030] Furthermore, the method for constructing a distributed voltage control model for coupling communication and power in step S300 of the present invention is as follows:

[0031] Build the Control variables for each regulation cycle Mathematical model, the specific mathematical model is as follows:

[0032] (5)

[0033] (6)

[0034] in, (t) indicates the The set of all transmission delays in a control cycle, Indicates The transmission delay of a control instruction, Indicates that the control instruction is The regulation cycle is passed through Channels sent to Delay on each PV node; In the The transmission delay of a control cycle is (t) is the set of control variables, Indicates The transmission delay of a control cycle is Time The control actions of each photovoltaic node, , is the set of photovoltaic nodes in the active distribution network, , Respectively, considering transmission delay The set of changes in active and reactive power of all photovoltaic nodes; represents transpose;

[0035] By The node voltage operation state of the regulation cycle is the initial value, and the transmission delay is considered. Voltage prediction model:

[0036]

[0037]

[0038] Wherein, V(t+1) is the node voltage of the t+1th regulation cycle, V(t) is the node voltage of the tth regulation cycle, is the system control matrix, is the sensitivity matrix, is the parameter uncertainty matrix;

[0039] The switching of transmission channels uses the system control matrix to build a coupling model that considers communication transmission channel scheduling and voltage control:

[0040]

[0041] in, represents the system control matrix that changes with the transmission channel switching, , Indicates the difference caused by switching the transmission channel The total number of combinations; Indicates the transmission channel scheduling instruction, Indicates the scheduling instruction constraints, which are as follows:

[0042]

[0043]

[0044] Build Area The objective function of minimizing the mid-node voltage disturbance deviation and control cost :

[0045]

[0046] in, is the reference voltage, represents the weight matrix related to the control variable, W represents the node voltage The associated weight matrix, is the set of regulation cycles;

[0047] Constructing the objective function The node voltage constraints are:

[0048]

[0049] In formula (12), represents the lower bound of the voltage, They represent the upper bounds of the voltage respectively; Indicates absolute value;

[0050] Construct the constraints of photovoltaic inverter power in the communication transmission channel scheduling and voltage control coupling model:

[0051]

[0052]

[0053]

[0054]

[0055] in, Representation Node The lower bound of the photovoltaic active power is Representation Node The upper bound of the photovoltaic active power, Representation Node The lower bound of the PV reactive power is Respectively represent nodes The upper bound of the PV reactive power is Representation Node The upper photovoltaic rated power, Represents the photovoltaic capacity factor.

[0056] Furthermore, in step S400 of the present invention, the method for reconstructing the sub-model using a partially observable Markov model includes:

[0057] Take the tth cycle as an example:

[0058] The multi-agent set is represented as ,Each agent manages a partition of the active distribution network;

[0059] State Space in Partially Observable Markov Models of Voltage Control ;

[0060]

[0061] in, Indicates that the transmission delay at the last moment is At the node The photovoltaic reactive power control command on Indicates A collection of photovoltaic nodes within an intelligent body;

[0062] Action Space in a Partially Observable Markov Model with Voltage Control ;

[0063]

[0064]

[0065] in, Indicates The action space of an agent, The transmission delay is At the node PV reactive power control instructions on the PV module;

[0066] Local Observation Aggregates in Partially Observable Markov Models with Voltage Control ;

[0067]

[0068]

[0069]

[0070] in, Indicates in the area The local observation set in Indicates in the area The operating status of the active distribution network measured in;

[0071] Reward Function in Voltage-Controlled Partially Observable Markov Model ;

[0072] .

[0073] Furthermore, the method for performing offline training of the multi-agent reinforcement learning algorithm in step S500 of the present invention includes:

[0074] 1) Input basic information of distribution network;

[0075] 2) Set training hyperparameters, randomly initialize network parameters for each distribution network partition agent, and initialize the shared experience replay pool;

[0076] 3) Set the maximum number of training rounds With a single round of training time step , set the current training round ;

[0077] 4) Initialize the current training time step ;

[0078] 5) Each distribution network area agent obtains the current status of the distribution network in its area;

[0079] 6) Each distribution network regional agent gives an action according to the distribution network status in step 5) and executes the action through the reactive power output of the distributed photovoltaic inverters in the region;

[0080] 7) Determine the reward using formula (23) , use equations (17)-(19) to determine the state and action;

[0081] 8) Each area of ​​the distribution network enters the next state , each distribution network regional intelligent agent will use local experience[ ]Save to shared experience replay pool middle;

[0082] 9) Each distribution network regional agent samples from the shared experience replay pool and uses the reverse gradient propagation algorithm to update network parameters;

[0083] 10) If ,but , and return to 5), otherwise go to the next step;

[0084] in, represents the training time step;

[0085] 11) Calculate the convergence index of each distribution network area intelligent agent , and save:

[0086]

[0087] in, represents the discount factor;

[0088] 12) If < ,but = +1, and return to step 4), otherwise proceed to the next step;

[0089] 13) Compare each round .

[0090] On the other hand, the present invention further discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the above method.

[0091] On the other hand, the present invention further discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0092] It can be seen from the above technical scheme that the present invention proposes a joint modeling method of "model prediction rolling optimization framework + partially observable Markov decision process", which decomposes the main distribution network coordinated voltage control problem into adjacent partition information interaction and multi-partition parallel optimization sub-problems, and embeds communication delays in the state space and action space to realize dynamic coupling modeling of the communication network and the electrical system. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 A fast solution model for voltage control under communication and power coupling of the present invention. DETAILED DESCRIPTION

[0094] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments.

[0095] like Figure 1 As shown, the main distribution coordinated voltage optimization control method under communication and electrical coupling described in this embodiment includes the following steps:

[0096] S100, based on the active distribution network topology structure, the area is divided, the network parameters of each partition are collected through the equipment, and the distribution network power flow model and voltage sensitivity matrix are constructed;

[0097] S200, constructing a communication transmission network based on variable delay;

[0098] S300, constructing a distributed voltage control model of communication and power coupling according to the distribution network power flow model, the communication transmission network and the voltage sensitivity matrix, establishing an optimization goal of minimizing the node voltage disturbance deviation and the control cost, and setting control constraints;

[0099] S400, decomposing the distributed voltage control model into sub-models for information interaction between adjacent partitions and parallel control optimization of multiple partitions, and reconstructing the sub-models using a partially observable Markov model;

[0100] S500, performing offline training on the reconstructed model using a multi-agent reinforcement learning algorithm to determine a voltage control method.

[0101] The following is a detailed description of each step:

[0102] S100, based on the active distribution network topology structure, the area is divided, the network parameters of each partition are collected through the equipment, and the distribution network power flow model and voltage sensitivity matrix are constructed;

[0103] Assume that the active distribution network is divided into regions, the set of load nodes is , the photovoltaic node set is , ; Among them, the area In the example, the load node set is , the branch set is , the photovoltaic node set is , ; The boundary line of the active distribution network is represented by a set of edges To indicate that, Indicates The node and The edges between nodes With Node The resistance, reactance, conductance and susceptance of the branch between are , , , ;

[0104] S110, assuming There are loads and photovoltaics on each node, and the power flow constraint equation of the active distribution network is constructed;

[0105] The power flow constraint equation of the active distribution network is as follows:

[0106]

[0107] in, Indicates The active power of photovoltaic power on each node, Indicates The reactive power of photovoltaic power on each node; Indicates The active power of the load on each node, Indicates The reactive power of the load on each node; and Respectively represent The node and The voltage of each node; and Respectively expressed as The node and The real and imaginary parts of the conductance and susceptance between nodes; Indicates The node and The voltage phase angle difference between nodes;

[0108] S120, solve the node voltage according to the active distribution network power flow constraint equation Photovoltaic active power The sensitivity matrix and node voltage Reactive power of photovoltaic The sensitivity matrix :

[0109]

[0110] in, is the Jacobian matrix, , They are the partial derivative matrices of node voltage amplitude with respect to active and reactive power respectively.

[0111] S200, constructing a communication transmission network based on variable delay;

[0112] S210, selecting a plurality of transmission channels with different constant delays;

[0113] Assume that the resource-limited communication network is a transmission server with variable delay;

[0114] use Represents multiple transmission channels with different constant delays. The specific formula is as follows:

[0115]

[0116] in, Represents bounded delay channel.

[0117] S220, solving the transmission capacity constraint of the communication network according to the transmission channels with different constant delays;

[0118] (4)

[0119] in, Indicates channel transmission capacity.

[0120] S300, constructing a distributed voltage control model of communication and power coupling according to the distribution network power flow model, the communication transmission network and the voltage sensitivity matrix, establishing an optimization goal of minimizing the node voltage disturbance deviation and the control cost, and setting control constraints;

[0121] The method for constructing a distributed voltage control model for communication and power coupling is as follows:

[0122] Build the Control variables for each regulation cycle Mathematical model, the specific mathematical model is as follows:

[0123] (5)

[0124] (6)

[0125] in, (t) indicates the The set of all transmission delays in a control cycle, Indicates The transmission delay of a control instruction, Indicates that the control instruction is The regulation cycle is passed through Channels sent to Delay on each PV node; In the The transmission delay of a control cycle is (t) is the set of control variables, Indicates The transmission delay of a control cycle is Time The control actions of each photovoltaic node, , is the set of photovoltaic nodes in the active distribution network, , Respectively, considering transmission delay The set of changes in active and reactive power of all photovoltaic nodes; represents transpose;

[0126] By The node voltage operation state of the regulation cycle is the initial value, and the transmission delay is considered. Voltage prediction model:

[0127] (7)

[0128] (8)

[0129] Wherein, V(t+1) is the node voltage of the t+1th regulation cycle, V(t) is the node voltage of the tth regulation cycle, is the system control matrix, is the sensitivity matrix, is the parameter uncertainty matrix;

[0130] The switching of transmission channels uses the system control matrix to build a coupling model that considers communication transmission channel scheduling and voltage control:

[0131]

[0132] in, represents the system control matrix that changes with the transmission channel switching, , Indicates the difference caused by switching the transmission channel The total number of combinations; Indicates the transmission channel scheduling instruction, Indicates the scheduling instruction constraints, which are as follows:

[0133]

[0134] (10)

[0135] Build Area The objective function of minimizing the mid-node voltage disturbance deviation and control cost :

[0136]

[0137] in, is the reference voltage, represents the weight matrix related to the control variable, W represents the node voltage The associated weight matrix, is the set of regulation cycles;

[0138] Constructing the objective function The node voltage constraints are:

[0139]

[0140] In formula (12), Respectively represent the upper and lower limits of voltage; Indicates absolute value;

[0141] Construct the constraints of photovoltaic inverter power in the communication transmission channel scheduling and voltage control coupling model:

[0142]

[0143]

[0144]

[0145]

[0146] in, Respectively represent nodes The upper and lower bounds of the photovoltaic active power, Respectively represent nodes The upper and lower bounds of the PV reactive power, Representation Node The upper photovoltaic rated power, Represents the photovoltaic capacity factor.

[0147] S400, decomposing the distributed voltage control model into sub-models for information interaction between adjacent partitions and parallel control optimization of multiple partitions, and reconstructing the sub-models using a partially observable Markov model;

[0148] The method of reconstructing the voltage control sub-model using the partially observable Markov model is:

[0149] Take the tth cycle as an example:

[0150] The multi-agent set is represented as ,Each agent manages a partition of the active distribution network;

[0151] State Space in Partially Observable Markov Models of Voltage Control ;

[0152]

[0153] in, Indicates that the transmission delay at the last moment is At the node The photovoltaic reactive power control command on Indicates A collection of photovoltaic nodes within an intelligent body;

[0154] Action Space in a Partially Observable Markov Model with Voltage Control ;

[0155]

[0156]

[0157] in, Indicates The action space of an agent, The transmission delay is At the node PV reactive power control instructions on the PV module;

[0158] Local Observation Aggregates in Partially Observable Markov Models with Voltage Control ;

[0159]

[0160]

[0161]

[0162] in, Indicates in the area The local observation set in Indicates in the area The operating status of the active distribution network measured in;

[0163] Reward Function in Voltage-Controlled Partially Observable Markov Model ;

[0164]

[0165] S500, performing offline training on the reconstructed model using a multi-agent reinforcement learning algorithm to determine a voltage control method;

[0166] The offline training process of the multi-agent reinforcement learning algorithm includes:

[0167] 1) Input basic information of distribution network;

[0168] 2) Set training hyperparameters, randomly initialize network parameters of each distribution network partition agent, and initialize the shared experience replay pool;

[0169] 3) Set the maximum number of training rounds With a single round of training time step , set the current training round ;

[0170] 4) Initialize the current training time step ;

[0171] 5) Each distribution network area agent obtains the current status of the distribution network in its area;

[0172] 6) Each distribution network regional agent gives an action according to the distribution network status in step 5) and executes the action through the reactive power output of the distributed photovoltaic inverters in the region;

[0173] 7) Determine the reward using formula (23) , use equations (17)-(19) to determine the state and action;

[0174] 8) Each area of ​​the distribution network enters the next state , each distribution network regional intelligent agent will use local experience[ ]Save to shared experience replay pool middle;

[0175] 9) Each distribution network regional agent samples from the shared experience replay pool and uses the reverse gradient propagation algorithm to update network parameters;

[0176] 10) If ,but , and return to 5), otherwise go to the next step;

[0177] in, represents the training time step;

[0178] 11) Calculate the convergence index of each distribution network area intelligent agent , and save:

[0179]

[0180] in, represents the discount factor;

[0181] 12) If < ,but = +1, and return to step 4), otherwise proceed to the next step;

[0182] 13) Compare each round ;

[0183] Communication networks are usually abstracted as simple time delay constraints. Time delays are usually set based on a priori given parameters rather than the real-time requirements of the control system, lacking analysis of the dynamic interaction between the network system and the physical system. The present invention constructs a communication network consisting of transmission channels with different delays, combines communication transmission delays and transmission channel switching with voltage control, and uses deep reinforcement learning algorithm training to achieve real-time response to voltage disturbances in the distribution network.

[0184] On the other hand, the present invention further discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0185] In another embodiment provided in the present application, a computer program product comprising instructions is also provided, which, when executed on a computer, enables the computer to execute any of the main-distribution coordinated voltage optimization control methods under communication and electrical coupling in the above-mentioned embodiments.

[0186] It is understandable that the system, device and storage medium provided in the embodiments of the present invention correspond to the method provided in the embodiments of the present invention, and the explanation, examples and beneficial effects of the relevant contents can refer to the corresponding parts in the above methods.

[0187] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive Solid State Disk (SSD)), etc.

[0188] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0189] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0190] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing the voltage of main distribution coordination under communication and electrical coupling, characterized in that: The following steps are involved: S100, based on the active distribution network topology structure, the area is divided, the network parameters of each partition are collected through the equipment, and the distribution network power flow model and voltage sensitivity matrix are constructed; The voltage sensitivity matrix construction method includes: S120, solve the node voltage according to the active distribution network power flow constraint equation Photovoltaic active power The sensitivity matrix and node voltage Reactive power of photovoltaic The sensitivity matrix : in, , are the partial derivative arrays of node voltage amplitude for active and reactive power respectively; S200, constructing a communication transmission network based on variable delay; S300, constructing a distributed voltage control model of communication and power coupling according to the distribution network power flow model, the communication transmission network and the voltage sensitivity matrix, establishing an optimization goal of minimizing the node voltage disturbance deviation and the control cost, and setting control constraints; Among them, the method of constructing a distributed voltage control model for communication and power coupling is as follows: Build the Control variables for each regulation cycle Mathematical model, the specific mathematical model is as follows: （5） （6） in, (t) indicates the The set of all transmission delays in a control cycle, Indicates The transmission delay of a control instruction, Indicates that the control instruction is The regulation cycle is passed through Channels sent to Delay on each PV node; In the The transmission delay of a control cycle is (t) is the set of control variables, Indicates The transmission delay of a control cycle is Time The control actions of each photovoltaic node, , is the set of photovoltaic nodes in the active distribution network, Respectively, considering transmission delay The set of changes in active and reactive power of all photovoltaic nodes; represents transpose; By The node voltage operation state of the regulation cycle is the initial value, and the transmission delay is considered. Voltage prediction model: Wherein, V(t+1) is the node voltage of the t+1th regulation cycle, V(t) is the node voltage of the tth regulation cycle, is the system control matrix, is the sensitivity matrix, is the parameter uncertainty matrix; The switching of transmission channels uses the system control matrix to build a coupling model that considers communication transmission channel scheduling and voltage control: in, represents the system control matrix that changes with the transmission channel switching, , Indicates the difference caused by switching the transmission channel The total number of combinations; Indicates the transmission channel scheduling instruction, Indicates the scheduling instruction constraints, which are as follows: Build Area The objective function of minimizing the mid-node voltage disturbance deviation and control cost : in, is the reference voltage, represents the weight matrix related to the control variable, W represents the node voltage The associated weight matrix, is the set of regulation cycles; Constructing the objective function The node voltage constraints are: In formula (12), represents the lower bound of the voltage, They represent the upper bounds of the voltage respectively; Indicates absolute value; Construct the constraints of photovoltaic inverter power in the communication transmission channel scheduling and voltage control coupling model: in, Representation Node The lower bound of the PV active power is Representation Node The upper bound of the photovoltaic active power, Representation Node The lower bound of the PV reactive power is Respectively represent nodes The upper bound of the PV reactive power is Representation Node The upper photovoltaic rated power, represents the photovoltaic capacity factor; S400, decomposing the distributed voltage control model into sub-models for information interaction between adjacent partitions and parallel control optimization of multiple partitions, and reconstructing the sub-models using a partially observable Markov model; S500, performing offline training on the reconstructed model using a multi-agent reinforcement learning algorithm to determine a voltage control method.

2. The method for optimizing the voltage of main distribution coordination under communication and electrical coupling according to claim 1 is characterized in that: The method for constructing a power distribution network flow model in step S100 includes: S110, assuming There are loads and photovoltaics on each node, and the power flow constraint equation of the active distribution network is constructed; The power flow constraint equation of the active distribution network is as follows: in, Indicates The active power of photovoltaic power on each node, Indicates The reactive power of photovoltaic power on each node; Indicates The active power of the load on each node, Indicates The reactive power of the load on each node; and Respectively represent The node and The voltage of each node; and Respectively expressed as The node and The real and imaginary parts of the conductance and susceptance between nodes; Indicates The node and The voltage phase angle difference between the nodes.

3. The method for optimizing the voltage of main distribution coordination under communication and electrical coupling according to claim 1 is characterized in that: The communication transmission network construction method in step S200 includes: S210, selecting a plurality of transmission channels with different constant delays; Assume that the resource-limited communication network is a transmission server with variable delay; use Represents multiple transmission channels with different constant delays. The specific formula is as follows: in, Represents bounded delay Channel; S220, solving the transmission capacity constraint of the communication network according to the transmission channels with different constant delays; （4） in, Indicates channel transmission capacity.

4. The method for optimizing the voltage of main distribution coordination under communication and electrical coupling according to claim 1 is characterized in that: The method for reconstructing the sub-model using a partially observable Markov model in step S400 includes: Take the tth cycle as an example: The multi-agent set is represented as ,Each agent manages a partition of the active distribution network; State Space in Partially Observable Markov Models of Voltage Control ; in, Indicates that the transmission delay at the last moment is At the node The photovoltaic reactive power control command on Indicates A collection of photovoltaic nodes within an intelligent body; Action Space in a Partially Observable Markov Model with Voltage Control ; in, Indicates The action space of an agent, The transmission delay is At the node PV reactive power control instructions on the PV module; Local Observation Aggregates in Partially Observable Markov Models with Voltage Control ; in, Indicates in the area The local observation set in Indicates in the area The operating status of the active distribution network measured in; Reward Function in Voltage-Controlled Partially Observable Markov Model ; 。 5. The method for optimizing the voltage of main distribution coordination under communication and electrical coupling according to claim 1 is characterized in that: The offline training method of the multi-agent reinforcement learning algorithm in step S500 includes: 1) Input basic information of distribution network; 2) Set training hyperparameters, randomly initialize network parameters for each distribution network partition agent, and initialize the shared experience replay pool; 3) Set the maximum number of training rounds With a single round of training time step , set the current training round ; 4) Initialize the current training time step ; 5) Each distribution network area agent obtains the current status of the distribution network in its area; 6) Each distribution network regional agent gives an action according to the distribution network status in step 5) and executes the action through the reactive power output of the distributed photovoltaic inverters in the region; 7) Determine the reward using formula (23) , use equations (17)-(19) to determine the state and action; 8) Each area of ​​the distribution network enters the next state , each distribution network regional intelligent agent will use local experience[ ]Save to shared experience replay pool middle; 9) Each distribution network regional agent samples from the shared experience replay pool and uses the reverse gradient propagation algorithm to update network parameters; 10) If ,but , and return to 5), otherwise go to the next step; in, represents the training time step; 11) Calculate the convergence index of each distribution network area intelligent agent , and save: in, represents the discount factor; 12) If < ,but = +1, and return to step 4), otherwise proceed to the next step; 13) Compare each round .

6. A computer device comprising a memory and a processor, characterized in that: The memory stores a computer program, and when the computer program is executed by the processor, the processor executes the method according to any one of claims 1 to 5.

Images