AC / DC power distribution network real-time distributed voltage control method and system

Abstract

The invention discloses a real-time distributed voltage control method and system for an AC / DC power distribution network, and belongs to the technical field of smart power grids, and the method comprises the following steps: dividing the AC / DC power distribution network in a controlled region into a plurality of regions with the high aggregation degree in the regions and the low coupling among the regions as targets; establishing a voltage control model suitable for cooperative control of the AC / DC power distribution network subareas in the controlled area; substituting into the established voltage control model, respectively constructing the plurality of areas into intelligent agents, and decomposing the voltage control problem of the AC / DC power distribution network in the controlled area into a plurality of sub-problems of cooperative control of the areas; and establishing the plurality of sub-problems as a Markov cooperative game problem, solving the Markov cooperative game problem, obtaining a scheduling instruction of the controlled equipment, and realizing real-time distributed voltage control of the AC / DC power distribution network.

Description

Technical Field

[0001] The present invention belongs to the technical field of smart grids, and more specifically, relates to a real-time distributed voltage control method and system for an AC / DC power distribution network. Background Art

[0002] Distributed photovoltaic power is connected to AC and DC distribution networks at a high rate, and its volatility and uncertainty bring challenges to the safe and stable operation of AC and DC distribution networks. Due to the existence of AC branches, DC branches and converters, the operation modes and control strategies of AC and DC distribution networks are more diversified. When performing voltage control, it is necessary to consider the voltages on both the AC and DC sides, as well as the power exchange between the two. In this case, the traditional centralized control method requires a large amount of calculation, has a low solution efficiency, and relies on perfect communication equipment. Therefore, it is impossible to achieve real-time voltage control when the output of new energy changes rapidly. Summary of the invention

[0003] In order to solve the deficiencies in the prior art, the present invention provides a real-time distributed voltage control method and system for an AC / DC distribution network. By partitioning the AC / DC distribution network, a Markov game model for the partition voltage control problem is established. Based on the strategy of centralized training and decentralized execution, the model is solved through multi-agent deep reinforcement learning, thereby implementing distributed voltage control.

[0004] The present invention adopts the following technical solution. A first aspect of the present invention provides a real-time distributed voltage control method for an AC / DC power distribution network, comprising the following steps: The AC and DC distribution networks in the controlled area are divided into multiple zones with the goal of high aggregation within the zone and low coupling between zones; Establish a voltage control model suitable for coordinated control of AC and DC distribution network segments within the controlled area; Substitute the established voltage control model, construct multiple areas as intelligent agents, and decompose the voltage control problem of the AC and DC distribution network in the controlled area into multiple sub-problems of coordinated control of different areas; Multiple sub-problems are established as Markov cooperative game problems and solved to obtain the dispatching instructions of the controlled equipment and realize real-time distributed voltage control of AC and DC distribution networks.

[0005] Preferably, the dividing of the AC and DC distribution network in the controlled area into multiple areas with the goal of high aggregation within the area and low coupling between areas comprises: Based on the power flow calculation model of the AC and DC distribution network in the controlled area, a model of the sensitivity of the node voltage amplitude change of the AC and DC distribution network in the controlled area to the voltage amplitude of active power and reactive power is established; The active power-voltage influence matrix and reactive power-voltage influence are calculated based on the sensitivity of active power and reactive power to voltage amplitude respectively; The electrical distance between two points on the AC side and the electrical distance between two points on the DC side are calculated using the active-voltage influence matrix and the reactive-voltage influence matrix respectively; Based on the electrical distance, the node aggregation degree within the area and the coupling degree between areas are represented respectively. Taking into account the node aggregation degree within the area, the coupling degree between areas, the total number of areas, the area's voltage regulation capability and the interconnection status within the area, the objective function of dividing the AC and DC distribution network into areas is established. Solve the objective function of AC / DC distribution network zoning and obtain the results of AC / DC distribution network zoning in the controlled area.

[0006] Preferably, an improved particle swarm algorithm is used to solve the objective function of dividing the AC and DC distribution network into zones, including: Define the signal function; Establish the particle velocity update strategy of discrete particle swarm method based on signal function; Considering that the adjacent nodes of all nodes except the boundary nodes of the region in the AC / DC distribution network belong to the same area, a particle position update strategy is established.

[0007] Preferably, the establishment of a voltage control model suitable for coordinated control of AC and DC distribution network segments within a controlled area includes: Considering the topological structure of AC and DC distribution networks in the controlled area, a branch power flow model is established; Taking into account the VSC connection relationship, a VSC model is established; Taking into account the established branch power flow model and VSC model, taking the photovoltaic inverter, VSC and energy storage as the control objects, the objective function of the AC / DC distribution network voltage control model in the controlled area is constructed with the goal of minimizing the voltage deviation in the area; Establish constraints that match the objective function of the voltage control model of the AC / DC distribution network in the controlled area.

[0008] Preferably, taking into account the VSC connection relationship, a VSC model is established including the voltage amplitude on the VSC DC side, the injected power at the AC side bus, the injected power at the AC side C of the converter, and the active power injected into the VSC DC side; The constraints matching the objective function of the voltage control model of the AC / DC distribution network in the controlled area include: system power balance constraints, voltage upper and lower limit constraints, photovoltaic inverter power constraints, energy storage power constraints and VSC constraints.

[0009] Preferably, the voltage control model established by substitution constructs multiple areas as intelligent agents respectively, and decomposes the voltage control problem of the AC and DC distribution network in the controlled area into multiple sub-problems of coordinated control of the sub-areas, including: Establish the state space of each area, including: active power and reactive power of the area load, photovoltaic active output, and charge state of energy storage; Establish the action space for each area, including: dispatching instructions for photovoltaics, dispatching instructions for energy storage charging and discharging, and dispatching instructions for VSC active and reactive power; A reward function is established for each zone, including minimizing node voltage deviation in the presence of a penalty for exceeding the voltage limit.

[0010] Preferably, the action space of each area includes: calculating the reactive power of photovoltaic power, correcting the VSC reactive power taking into account the VSC capacity limitation, and correcting the charging and discharging power of the energy storage according to the energy storage SOC taking into account the energy storage capacity limitation.

[0011] Preferably, for the established Markov cooperative game problem, the MASAC algorithm is adopted to solve it based on the "centralized training, decentralized execution" framework.

[0012] Preferably, each agent is constructed with an Actor network, a Critic network, a target Actor network, and a target Critic network to establish the goal of the neural network training process in the MASAC algorithm; Construct loss functions and gradient update strategies for training; A centralized training method is used to train the neural network in the MASAC algorithm; After the training of each agent is completed, the parameters of the Actor network are fixed and deployed to each area. During the distributed execution process, each agent only takes actions based on the local observation state and does not rely on communication between agents.

[0013] A second aspect of the present invention provides a real-time distributed voltage control system for an AC / DC power distribution network, and the method for real-time distributed voltage control for an AC / DC power distribution network includes: The area rolling division module divides the AC and DC distribution networks in the controlled area into multiple areas according to the set time period, with the goal of high aggregation within the area and low coupling between areas; The voltage control model module is used to construct and store a voltage control model suitable for coordinated control of AC and DC distribution network segments within the controlled area; The task decomposition module is used to construct multiple areas as intelligent agents, decompose the voltage control problem of the AC and DC distribution network in the controlled area into multiple sub-problems of coordinated control of the sub-areas; and establish multiple sub-problems as Markov cooperative game problems. The solver is used to solve the Markov cooperative game problem, obtain the dispatching instructions of the controlled equipment, and realize the real-time distributed voltage control of the AC and DC distribution network.

[0014] Compared with the prior art, the beneficial effects of the present invention include at least: the present invention proposes a real-time distributed voltage control method and system for AC / DC distribution networks, which solves the technical problem that traditional methods have a long solution time and cannot meet real-time voltage control requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flow chart of a real-time distributed voltage control method for an AC / DC distribution network provided in accordance with an embodiment of the present invention; Figure 2 is a schematic diagram of a VSC model provided according to an embodiment of the present invention; Figure 3 It is a schematic diagram of the improved IEEE 33-node AC / DC distribution network partitioning result in Verification Example 1 of the present invention; Figure 4 It is a curve diagram of the load and photovoltaic output of the training set in Verification Example 1 of the present invention; Figure 5 It is a cumulative reward graph during the training process in Verification Example 1 of the present invention; Figure 6 It is a photovoltaic reactive power curve diagram of each node in Verification Example 1 of the present invention; Figure 7 It is a VSC active power curve diagram in Verification Example 1 of the present invention; Figure 8 It is a VSC reactive power curve diagram in Verification Example 1 of the present invention; Fig. 9 It is a bar chart of energy storage active output in Verification Example 1 of the present invention; Fig.10 is a voltage distribution diagram of each area after the method of the present invention is executed according to Verification Example 1 of the present invention; Fig.11 It is a diagram of the improved 162-node AC / DC distribution network topology and partitioning results according to Verification Example 2 of the present invention; Fig.12 The voltage distribution diagram of the 162-node AC / DC power distribution network is improved according to Verification Example 2 of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention more clear, the technical solution of the present invention will be described clearly and completely in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention, not all of them. Based on the spirit of the present invention, ordinary technicians in this field can make the above-mentioned invention without creative work. like Figure 1 As shown, embodiment 1 of the present invention provides a real-time distributed voltage control method for an AC / DC power distribution network, comprising the following steps: Step 1: Divide the AC and DC distribution networks in the controlled area into multiple areas with the goal of high aggregation within the area and low coupling between areas.

[0017] Preferably but not limiting, step 1 specifically comprises: Step 1.1: Based on the power flow calculation model of the AC / DC distribution network in the controlled area, a model of the sensitivity of the node voltage amplitude change of the AC / DC distribution network in the controlled area to the active power and reactive power to the voltage amplitude is established.

[0018] Based on the Jacobian matrix, the power flow calculation model of the AC and DC distribution network in the controlled area is established, which is expressed as the following formula (1): (1) Where: and are the changes of active power and reactive power injected into the node respectively; , , , are the elements of the Jacobian matrix respectively; and are the voltage phase angle and voltage amplitude changes respectively.

[0019] The power flow calculation model of the AC / DC distribution network in the controlled area is transformed into the voltage phase angle and voltage amplitude change model of the AC / DC distribution network in the controlled area, which is expressed by the following formula (2). The node voltage amplitude change of the AC / DC distribution network in the controlled area is further expressed by the following formula (3): (2) (3) Where: and They represent the sensitivity of active power, reactive power and node voltage phase angle respectively; and They respectively represent the sensitivity of active power and reactive power to voltage amplitude.

[0020] Step 1.2: The active power-voltage influence matrix and reactive power-voltage influence are characterized by the sensitivity of active power and reactive power to voltage amplitude, which can be expressed as the following formulas (4) and (5): (4) (5) Where: It represents the active power-voltage influence capability of node j on node i; It represents the reactive power-voltage influence capability of node j on node i; and are the elements of the j-th row and j-th column and the elements of the i-th row and j-th column of the active voltage sensitivity matrix respectively; and are the j-th row and j-th column element and the i-th row and j-th column element of the reactive voltage sensitivity matrix respectively; and are the voltage change of node j and the voltage change of node i when the active power injected into node j changes respectively; and They are the voltage change of node j and the voltage change of node i when reactive power changes are injected into node j respectively.

[0021] Step 1.3: Calculate the electrical distance between two points on the AC side and the electrical distance between two points on the DC side using the active-voltage influence matrix and the reactive-voltage influence matrix.

[0022] Active electrical distance between two nodes and reactive electrical distance , which can be expressed as the following formulas (6) and (7): (6) (7) Where: and are the active electrical distance and reactive electrical distance between two nodes respectively.

[0023] Since VSC can control active power and reactive power to achieve decoupling of AC distribution network and DC distribution network, the electrical distances on the AC side and DC side can be defined and partitioned separately. For the AC side, it is necessary to consider the impact of both active power and reactive power on voltage. For the DC side, only the impact of active power on voltage needs to be considered. The electrical distance between two points on the AC side and the electrical distance between two points on the DC side are expressed by the following formulas (8) and (9): (8) (9) Where: and They are the electrical distance between two points on the AC side and the electrical distance between two points on the DC side respectively.

[0024] Step 1.4: Based on the electrical distance, the node aggregation degree within the area and the coupling degree between areas are characterized respectively. Taking into account the node aggregation degree within the area, the coupling degree between areas, the total number of areas, the area's voltage regulation capability and the interconnection status within the area, the objective function of dividing the AC and DC distribution network into areas is established, which is expressed as the following formulas (10) and (11): (10) (11) (12) (13) (14) (15) (16) (17) Where: For Node The set of internal node numbers of the area to which it belongs; ISCI represents the node aggregation degree factor within the area, which is expressed as the above formula (12): for its weight; ISSI characterizes the sparsity factor between patches, that is, the degree of coupling, as expressed in the above formula (13): for its weight; SCI characterizes the number of areas. For its weight, is the actual number of sub-partitions, is the ideal number of sub-partitions, is the penalty parameter used to control the width of the function, expressed as follows: ISRCI characterizes the area's ability to participate in voltage regulation, as expressed in the above formula (16); PCI represents the connectivity status within the area and is expressed as in the above formula (17).

[0025] Step 1.5: Solve the objective function of the AC / DC distribution network area division to obtain the AC / DC distribution network area division result in the controlled area. Preferably, but not limited to, the improved particle swarm algorithm is used for solving. Specifically, the following sub-steps are included: The signal function is defined as follows: (18) (19) Where: represents the signal function, Represents the activation function.

[0026] Based on the signal function, the particle velocity update strategy of the discrete particle swarm method is established, which is expressed as the following formula (20): (20) Where: Subscript i represents the i-th particle, subscript k represents the node number, and t is the number of iterations; v is the velocity of the particle, and x is the current position of the particle; is the individual optimal value, is the optimal value for the group; is the inertia weight, and is the learning factor, and is a random number in the interval [0,1]; Represents the exclusive-or operation.

[0027] Except for the boundary nodes of the region, the adjacent nodes of the remaining nodes all belong to the same area. Based on this feature, the particle position update method is proposed as follows: (twenty one) Where: is the set of nodes connected to node k; The number of the slice to which any randomly selected node belongs.

[0028] The particle swarm algorithm is used to divide the AC distribution network and the DC distribution network into zones to obtain AC zones and DC zones.

[0029] Step 2: Establish a voltage control model suitable for coordinated control of AC and DC distribution network segments within the controlled area.

[0030] Preferably but not limiting, step 2 specifically comprises: Step 2.1: Considering the topological structure of the AC and DC distribution networks in the controlled area, establish a branch power flow model, which is expressed by the following formulas (22)-(24): (twenty two) (twenty three) (twenty four) Where: and is the injected active power and reactive power at node i, is the voltage amplitude at node i, and are the conductance and susceptance of the AC branch ij, is the voltage phase difference between nodes i and j.

[0031] Step 2.2: If Figure 2 As shown, taking into account the connection relationship of VSC (Voltage Source Converter), a VSC model is established, which includes the voltage amplitude on the VSC DC side, the injected power at the AC side bus, the injected power at the converter AC side C, and the active power injected into the VSC DC side, which is expressed by the following formulas (25)-(31): (25) (26) (27) (28) (29) (30) (31) Where: is the voltage amplitude on the DC side of the VSC, expressed as above formula (25); and are the voltage amplitude and phase angle at the connection between VSC and AC system, respectively; and are respectively the active power and reactive power flowing into the VSC from the AC bus, expressed as above formula (26); is the voltage amplitude and phase angle at the VSC AC outlet, for and The phase angle difference is expressed as above formula (27); and are the equivalent resistance and reactance reflecting the VSC losses, respectively; and are the injected power, active power and reactive power at the AC side C of the VSC, respectively, expressed as above formula (28); is the active power flowing out of the VSC DC side, expressed as above formula (29); is the voltage amplitude on the DC side of the VSC, is the DC voltage utilization rate, is the modulation ratio; is the active power loss of VSC, and is related to the AC side current It conforms to the quadratic function relationship, , , Represents the loss factor of the converter.

[0032] Step 2.3: Taking into account the branch power flow model established in step 2.1 and the VSC model established in step 2.2, the PV inverter, VSC and energy storage are taken as the control objects, and the objective function of the AC / DC distribution network voltage control model in the controlled area is constructed with the goal of minimizing the voltage deviation in the area, which is expressed as the following formula (32): (32) Where: is the objective function, is the control object, t represents the time, is the AC node voltage reference value, is the DC node voltage reference value, represents the sum of the voltage deviations of the AC side nodes, represents the sum of the DC side node voltage deviations, , , , are the controlled variables, which are the PV reactive output, the energy storage charging and discharging power, the VSC active power and the VSC reactive power.

[0033] Step 2.4: Establish constraints that match the objective function of the AC / DC distribution network voltage control model in the controlled area, including: system power balance constraints, voltage upper and lower limit constraints, PV inverter power constraints, energy storage power constraints, and VSC constraints.

[0034] The system power balance constraint is expressed as follows: (33) (34) (35) Where: For AC load, is a DC load, For photovoltaic active power output, is the power loss of the VSC.

[0035] The voltage upper and lower limit constraints are expressed as follows: (36) (37) Where: and are the upper and lower limits of the AC node voltage respectively; and are the upper and lower limits of the DC node voltage respectively.

[0036] The PV inverter power constraint is expressed as follows: (38) (39) (40) Where: is the energy storage charging and discharging power at time t, less than 0 indicates charging, and greater than 0 indicates discharging; and are the maximum charging power and maximum discharging power of energy storage respectively; is the state of charge of the energy storage at time t; and are the maximum and minimum values ​​of the energy storage charge state, respectively; is the rated capacity of the energy storage, is the time step.

[0037] The VSC power constraint is expressed as follows: (41) (42) Where: is the active power of VSC, is the reactive power of VSC, is the capacity of the VSC, It is the maximum reactive power output of VSC.

[0038] Step 3: Substitute the established voltage control model, construct multiple areas as intelligent entities respectively, and decompose the voltage control problem of the AC and DC distribution network in the controlled area into multiple sub-problems of coordinated control of different areas by dividing the AC and DC distribution networks into areas.

[0039] It is worth noting that by partitioning the AC and DC distribution networks, the voltage control problem is decomposed into multiple sub-problems, so the voltage coordinated control problem can be modeled as a Markov cooperative game. An intelligent agent is set up in each partition, and each intelligent agent takes action by locally observing the internal state of the area, and minimizes the voltage deviation through coordination.

[0040] Preferably but not limiting, step 3 specifically comprises: Step 3.1: Establish the state space of each intelligent agent, that is, the state space of each area: it contains all the state information in the area environment, including the active power and reactive power of the area load, the active output of photovoltaics, and the charge state of energy storage.

[0041] is the state space of the agent. At time t, the state of agent j is the local state of area j. ,definition . Where i represents the index of the node in slice j.

[0042] Step 3.2: Establish the action space of each agent, that is, the action space of each patch: is the action space of the agent, i.e., the dispatching instructions for controllable devices, including the dispatching instructions for photovoltaics , dispatching instructions for energy storage charging and discharging , dispatching instructions for active and reactive power of VSC converter stations , , at time t, the action of agent j , define the action space of agent j .

[0043] Photovoltaic reactive power , expressed as the following formula (43): (43) Due to the capacity limitation of energy storage, the charging and discharging power of energy storage needs to be corrected according to its SOC state, which can be expressed as the following formula (44): (44) The limitation of VSC capacity requires correction of its reactive power, which can be expressed as follows: (45) Where: is the capacity of the i-th VSC.

[0044] Step 3.3: R is the agent’s reward function, which is the feedback given by the environment after executing an action. In the voltage control problem, the goal is to minimize the node voltage deviation. Each agent shares the same reward, which is expressed as the following formula (46): (46) Where: is the reference value of the AC side voltage amplitude, is the reference value of the DC side voltage amplitude, There is a penalty for exceeding the voltage limit.

[0045] Step 4: For the established Markov cooperation game problem, the MASAC algorithm is used to solve it based on the "centralized training, decentralized execution" framework.

[0046] Preferably but not limiting, step 4 specifically comprises: Each intelligent agent is constructed with an Actor network, a Critic network, a target Actor network, and a target Critic network, and the goal of the neural network training process in the MASAC algorithm is established, which is expressed as the following formulas (47) and (48): (47) (48) Where: is the expected cumulative return of the strategy, is the action function of agent i, i.e., Actor network, for Parameters, is the temperature coefficient of entropy, is the strategy entropy, expressed as follows: The Critic network is used to collect the observed states and actions of all agents and obtain global information to evaluate the actions of the agents. The target Critic network has the same structure as the Critic network, but its parameters are not updated in real time with training. Instead, they are copied from the Critic network at a fixed frequency to improve the stability of neural network training. The Critic network is updated by minimizing the loss function, which is expressed as follows: (48) (49) Where: is the target Actor network, for Parameters; is the target Critic network, for Parameters; is the target Critic network, for Parameters; is the discount factor; is the global observation state; is the set of actions of all agents.

[0047] The Actor network of each agent takes actions based on the local observation state and does not need to interact with other agents. The target Actor network has the same structure as the Actor network, and its parameters are copied from the Actor network at a fixed frequency to improve the stability of neural network training. The Actor network of each agent is updated by gradient ascent, which is expressed as follows: (50) (51) (52) Where: is the advantage function, which is used to determine whether the increase in rewards belongs to other agents; is the baseline of the multi-agent system, which represents the average value of all actions taken by the agent in the current state. It is the experience pool of agent i, which is used to store past experience and improve the efficiency and stability of learning.

[0048] The centralized training process of the proposed method is as follows: First, initialize the parameters of the Actor network , parameters of the Critic network , and copy it to the target Actor network and the target Critic network to initialize the agent's experience pool .

[0049] The agent is trained in M ​​cycles. In each cycle, one day's load and photovoltaic data are randomly selected from the training set. One day contains T hours. At time t, the state of the environment is , the agent obtains the local observation state .

[0050] In order to explore the action space more widely, at the beginning of training, the actions of each agent are randomly generated. When the total number of rounds reaches the set value When each agent follows the Actor network Select Action , control the controllable devices.

[0051] The actions of all agents constitute the action set at time t , acting on the AC / DC distribution network, obtains the voltage deviation at time t, and gives a reward , and then the environment enters the next state , the agent obtains a local observation of the next state in the environment .

[0052] Will Save to experience pool For subsequent updates.

[0053] At each update, the agent draws from the experience pool Extract The historical experience of the group is used for network update. The parameters of the Critic network are updated by minimizing the loss function, which is expressed as the following formula (53), and the Actor network is updated by maximizing the policy gradient, which is expressed as the following formula (54): (53) (54) Where: is the learning rate.

[0054] The target actor network and target critic network updates are expressed as follows: (55) (56) After the training of each agent is completed, the parameters of the Actor network are fixed and deployed to each area. During the distributed execution process, each agent only takes actions based on the local observation state and does not rely on communication between agents, so real-time voltage control of the AC and DC distribution network can be achieved. Since the Critic network adds the global state and the actions of all agents during training, during distributed control, the actions taken by each agent based on local observations are well coordinated, which can achieve the effect of minimizing the global voltage deviation.

[0055] Embodiment 2 of the present invention provides a real-time distributed voltage control system for an AC / DC power distribution network, and the operation of the real-time distributed voltage control method for an AC / DC power distribution network includes: The area rolling division module divides the AC and DC distribution networks in the controlled area into multiple areas according to the set time period, with the goal of high aggregation within the area and low coupling between areas; The voltage control model module is used to construct and store a voltage control model suitable for coordinated control of AC and DC distribution network segments within the controlled area; The task decomposition module is used to construct multiple areas as intelligent agents, decompose the voltage control problem of the AC and DC distribution network in the controlled area into multiple sub-problems of coordinated control of the sub-areas; and establish multiple sub-problems as Markov cooperative game problems. The solver is used to solve the Markov cooperative game problem, obtain the dispatching instructions of the controlled equipment, and realize the real-time distributed voltage control of the AC and DC distribution network.

[0056] Embodiment 3 of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is loaded into the processor, the method for real-time distributed voltage control of an AC / DC distribution network according to Embodiment 1 is implemented.

[0057] Embodiment 4 of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the real-time distributed voltage control method for an AC / DC power distribution network according to Embodiment 1 is implemented.

[0058] In order to more clearly introduce the outstanding essential features of the present invention and the significant progress it brings to the prior art, two application examples of implementing the present invention are introduced below.

[0059] Simulation analysis is performed in the improved IEEE 33-bus AC / DC distribution network. Partition results Figure 3 As shown, the load curve is Figure 4 As shown in Figure 2, the reward curve during training is as follows: Figure 5 As shown in Figure 2, the optimization results of the controllable equipment are as follows: Figure 6-9 As shown, specifically, the photovoltaic reactive power of each node is as follows Figure 6 As shown, the VSC active power is Figure 7 As shown, the VSC reactive power is Figure 8 As shown, the active output of energy storage is Fig. 9 As shown, the voltage distribution of each area after executing the proposed method is as follows Fig.10 As shown, The optimization results of controllable equipment are as follows: Figure 6-9 shown. Figure 6For the reactive power output of photovoltaics, when the photovoltaic power generation power is high, the inverter is controlled to absorb reactive power to reduce the voltage level. When the photovoltaic power is low and the load power is high, the inverter is controlled to emit reactive power to increase the node voltage. Figure 7 is the active power of VSC. It can be seen that when the photovoltaic power generation power is high, the active power of VSC is less than 0, indicating that VSC is in the inverter state, and the active power flows from the DC side to the AC side, thereby transmitting electric energy to the AC distribution network. When the photovoltaic power generation power is low, the active power of VSC is greater than 0, indicating that VSC is in the rectification state, and the active power flows from the AC side to the DC side to meet the needs of the DC load. Figure 8 It is the reactive power of VSC. A positive value indicates reactive power absorption, and a negative value indicates reactive power emission. During periods of high photovoltaic output, VSC reduces the voltage level by absorbing reactive power, and during periods of low photovoltaic output, VSC raises the voltage level by emitting reactive power. Fig. 9 It is the active output of energy storage. It can be seen that during the period of high photovoltaic output, the energy storage is in a charging state, storing electric energy while reducing the voltage level, while during the period of low photovoltaic output, the energy storage is in a discharging state, releasing electric energy to increase the voltage level. After executing the proposed control method, the average voltage deviation is 0.14%, the maximum voltage rise is 0.57%, and the maximum voltage drop is 0.80%, which verifies the effectiveness of the proposed method. It takes 0.015s to generate a control decision, indicating that the proposed method can meet the requirements of real-time voltage control.

[0060] The simulation analysis is carried out in the improved 162-node AC / DC distribution network. The partition results are shown in Fig.11 , 12 After the voltage is controlled by the method of the present invention, the average voltage deviation of the system is 0.13%, the maximum voltage rise is 0.52%, the maximum voltage drop is 1.63%, and the voltage distribution over 24 hours is as follows: Figure 8 As shown in the figure, the online decision time is 0.031s. The results show that the proposed control method can effectively reduce the voltage deviation, the online decision speed is not affected by the size of the system, and it has good scalability, leaving a dispatchable margin space for the safe and stable operation of the AC / DC distribution network.

[0061] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A real-time distributed voltage control method for an AC / DC distribution network, characterized in that: The following steps are involved: The AC and DC distribution networks in the controlled area are divided into multiple zones with the goal of high aggregation within the zone and low coupling between zones; Establish a voltage control model suitable for coordinated control of AC and DC distribution network segments within the controlled area; Substitute the established voltage control model, construct multiple areas as intelligent agents, and decompose the voltage control problem of the AC and DC distribution network in the controlled area into multiple sub-problems of coordinated control of different areas; Multiple sub-problems are established as Markov cooperative game problems and solved to obtain the dispatching instructions of the controlled equipment and realize real-time distributed voltage control of AC and DC distribution networks.

2. A real-time distributed voltage control method for an AC / DC power distribution network according to claim 1, characterized in that: The goal of dividing the AC and DC distribution network in the controlled area into multiple areas is to achieve a high degree of aggregation within the area and low coupling between areas, including: Based on the power flow calculation model of the AC and DC distribution network in the controlled area, a model of the sensitivity of the node voltage amplitude change of the AC and DC distribution network in the controlled area to the voltage amplitude of active power and reactive power is established; The active power-voltage influence matrix and reactive power-voltage influence are calculated based on the sensitivity of active power and reactive power to voltage amplitude respectively; The electrical distance between two points on the AC side and the electrical distance between two points on the DC side are calculated using the active-voltage influence matrix and the reactive-voltage influence matrix respectively; Based on the electrical distance, the node aggregation degree within the area and the coupling degree between areas are represented respectively. Taking into account the node aggregation degree within the area, the coupling degree between areas, the total number of areas, the area's voltage regulation capability and the interconnection status within the area, the objective function of dividing the AC and DC distribution network into areas is established. Solve the objective function of AC / DC distribution network zoning and obtain the results of AC / DC distribution network zoning in the controlled area.

3. A real-time distributed voltage control method for an AC / DC power distribution network according to claim 2, characterized in that: The improved particle swarm algorithm is used to solve the objective function of AC / DC distribution network division, including: Define the signal function; Establish the particle velocity update strategy of discrete particle swarm method based on signal function; Considering that the adjacent nodes of all nodes except the boundary nodes of the region in the AC / DC distribution network belong to the same area, a particle position update strategy is established.

4. A real-time distributed voltage control method for an AC / DC power distribution network according to claim 1 or 3, characterized in that: The voltage control model for establishing the AC / DC distribution network segment coordinated control in the controlled area includes: Considering the topological structure of AC and DC distribution networks in the controlled area, a branch power flow model is established; Taking into account the VSC connection relationship, a VSC model is established; Taking into account the established branch power flow model and VSC model, taking the photovoltaic inverter, VSC and energy storage as the control objects, the objective function of the AC / DC distribution network voltage control model in the controlled area is constructed with the goal of minimizing the voltage deviation in the area; Establish constraints that match the objective function of the voltage control model of the AC / DC distribution network in the controlled area.

5. A real-time distributed voltage control method for an AC / DC power distribution network according to claim 4, characterized in that: Taking into account the VSC connection relationship, a VSC model is established, which includes the voltage amplitude on the VSC DC side, the injected power at the AC side bus, the injected power at the converter AC side C, and the active power injected into the VSC DC side; The constraints matching the objective function of the voltage control model of the AC / DC distribution network in the controlled area include: system power balance constraints, voltage upper and lower limit constraints, photovoltaic inverter power constraints, energy storage power constraints and VSC constraints.

6. A real-time distributed voltage control method for an AC / DC power distribution network according to claim 1 or 5, characterized in that: The voltage control model established by substitution is used to construct multiple areas as intelligent agents respectively, and the voltage control problem of the AC and DC distribution network in the controlled area is decomposed into multiple sub-problems of coordinated control of sub-areas, including: Establish the state space of each area, including: active power and reactive power of the area load, photovoltaic active output, and charge state of energy storage; Establish the action space for each area, including: dispatching instructions for photovoltaics, dispatching instructions for energy storage charging and discharging, and dispatching instructions for VSC active and reactive power; A reward function is established for each zone, including minimizing node voltage deviation in the presence of a penalty for exceeding the voltage limit.

7. A real-time distributed voltage control method for an AC / DC power distribution network according to claim 6, characterized in that: The action space of each area includes: calculating the reactive power of photovoltaics, correcting the VSC reactive power taking into account the VSC capacity limitation, and correcting the charging and discharging power of the energy storage according to the energy storage SOC taking into account the energy storage capacity limitation.

8. A real-time distributed voltage control method for an AC / DC power distribution network according to claim 1 or 7, characterized in that: For the established Markov cooperative game problem, the MASAC algorithm is used to solve it based on the "centralized training, decentralized execution" framework.

9. A real-time distributed voltage control method for an AC / DC power distribution network according to claim 8, characterized in that: Construct each agent with Actor network, Critic network, target Actor network and target Critic network, and establish the goal of the neural network training process in the MASAC algorithm; Construct loss functions and gradient update strategies for training; A centralized training method is used to train the neural network in the MASAC algorithm; After the training of each agent is completed, the parameters of the Actor network are fixed and deployed to each area. During the distributed execution process, each agent only takes actions based on the local observation state and does not rely on communication between agents.

10. A real-time distributed voltage control system for an AC / DC power distribution network, running a real-time distributed voltage control method for an AC / DC power distribution network according to any one of claims 1 to 9, characterized in that: include: The area rolling division module divides the AC and DC distribution networks in the controlled area into multiple areas according to the set time period, with the goal of high aggregation within the area and low coupling between areas; The voltage control model module is used to construct and store a voltage control model suitable for coordinated control of AC and DC distribution network segments within the controlled area; The task decomposition module is used to construct multiple areas as intelligent agents, decompose the voltage control problem of the AC and DC distribution network in the controlled area into multiple sub-problems of coordinated control of the sub-areas; and establish multiple sub-problems as Markov cooperative game problems. The solver is used to solve the Markov cooperative game problem, obtain the dispatching instructions of the controlled equipment, and realize the real-time distributed voltage control of the AC and DC distribution network.

Images