Power distribution network active and reactive cooperation voltage control method and system

Abstract

The invention relates to the technical field of electric power regulation and control, in particular to an active and reactive cooperative voltage control method and system for a power distribution network, and the method comprises the steps: an active power distribution network agent is used for outputting an award coefficient and a penalty coefficient of active load income to each distributed photovoltaic agent; and outputting capacitive reactive power injected into a connection node by each static reactive power compensator to regulate and control reactive power equipment. And each distributed photovoltaic intelligent agent is used for acquiring a power reverse transmission quota value, an award coefficient and a penalty coefficient of an overvoltage responsibility value active load income, outputting a load lifting amount and a load time shifting amount at the current moment, and performing active load response. According to the method, power grid global optimization and user individual benefits are taken into consideration, the voltage control efficiency of the high-photovoltaic-permeability power distribution network can be effectively improved, and the benefits of the distributed photovoltaic main body in the power distribution network are redistributed.

Description

Technical Field

[0001] The present invention relates to the technical field of power regulation and control, and in particular to a method and system for coordinated voltage control of active and reactive power in a distribution network. Background Art

[0002] Distributed Photovoltaic (DPV) is widely connected to the Active Distribution Network (ADN) due to its high utilization efficiency and low negative impact on the environment. With the large-scale promotion and application of distributed photovoltaics, the state of the distribution network has changed to a complex and changeable state with unclear regularity, which has brought new challenges to the voltage stability of the active distribution network. In addition, as the proportion of photovoltaic power generation increases, the identity of distributed photovoltaic users has changed from traditional electricity users to producers and consumers with dual identities of supplying and consuming electricity. Their own operating conditions have also changed their income and the voltage level of the distribution network. In this context, when the net output of distributed photovoltaic users is too high, the voltage safety state of the distribution network can be effectively improved by guiding the active load of distributed photovoltaic producers and consumers.

[0003] Current active distribution network voltage control issues are often described as reactive power optimization problems, aiming to achieve voltage stability by regulating reactive power. However, the widespread integration of distributed photovoltaics has altered the power flow characteristics of the distribution network. During periods of abundant sunlight, when the generated power of distributed photovoltaics exceeds local load demand, power backflow occurs. This causes voltages at some nodes in the distribution network to rise significantly, leading to voltage overshooting. If voltage overshooting issues are not effectively addressed, power systems will be forced to abandon photovoltaic power generation to ensure safe and stable grid operation, resulting in a significant waste of photovoltaic power generation resources and reduced clean energy utilization efficiency.

[0004] Furthermore, existing distributed photovoltaic power generation strategies fail to fully consider the power generation capacity, access location, and overall system topology of distributed photovoltaic users, making it difficult to accurately guide their electricity consumption. This makes it difficult to achieve fair distribution of benefits among distributed photovoltaic prosumers when participating in active load response. This unfairness may not only weaken users' enthusiasm for load response but also hinder the healthy and sustainable development of the distributed photovoltaic industry, and is detrimental to the optimal allocation of power system resources and efficient operation.

[0005] Moreover, current voltage optimization is mostly reactive power optimization, while few consider active power coordination in the distribution network. In particular, there is a lack of research on adjusting active load by considering the flexible active load response of distributed photovoltaic producers and consumers. Summary of the Invention

[0006] To this end, the technical problem to be solved by the present invention is to overcome the problem that the voltage control mode based on reactive power optimization in the existing technology is difficult to cope with the voltage limit and light abandonment problems caused by power backflow, and the existing distributed photovoltaic power generation strategy is difficult to effectively guide users' electricity consumption behavior, affecting the fairness of the benefits of distributed photovoltaic producers and consumers in active load response.

[0007] To solve the above technical problems, the present invention provides a method for coordinated voltage control of active and reactive power in a distribution network, comprising: The TD3 algorithm is used to construct an active distribution network agent and multiple distributed photovoltaic agents; In the active distribution network agent, the observation state is constructed using the voltage of each node, and the action state is constructed using the capacitive reactive power injected by each static VAR compensator into its connected node and the reward coefficient and penalty coefficient of the active load benefit; the reward function is constructed using the total cost of distribution network operation. In each distributed photovoltaic agent, the observation state of the current distributed photovoltaic agent is constructed based on the current distributed photovoltaic user's power reverse transmission quota value, overvoltage responsibility value, and total time-shifted load. The active load of the current distributed photovoltaic user is divided into fixed load, scalable load, and time-shiftable load, and the action state is constructed based on the current distributed photovoltaic user's actual load scalability and actual load time-shifting amount. The reward function is constructed based on the current distributed photovoltaic user's income. The active distribution network agent obtains the voltage of each node at every moment and outputs the reward coefficient and penalty coefficient of the active load benefit at the current moment to each distributed photovoltaic agent. It also outputs the capacitive reactive power injected by each static VAR compensator into its connected node at the current moment and regulates each static VAR compensator. Each distributed photovoltaic intelligent agent obtains the current power reverse transmission quota value and overvoltage responsibility value, as well as the reward coefficient and penalty coefficient of the active load benefit sent by the active distribution network intelligent agent, outputs the actual load increase and actual load time shift at the current moment, and performs active load response.

[0008] Preferably, in the active distribution network agent, the reward function is constructed based on the total cost of distribution network operation, and the formula is expressed as: ; ; in, is the reward function of the active distribution network agent at time t; is the marginal network loss coefficient, is the sum of the losses of the lines in the active distribution network at time t; is the voltage over-limit additional cost coefficient, The node voltage exceeds the limit at time t; is the active load response cost corresponding to the reward coefficient and penalty coefficient of the current active load benefit, is the reactive expenditure under the current active load response; is the set of nodes in the active distribution network, is the voltage of the jth node at time t, and are the upper and lower limits of the active distribution network voltage, is a linear rectification function.

[0009] Preferably, the active load of the current distributed photovoltaic user is divided into fixed load, scalable load and time-shiftable load, and the formula is expressed as follows: ; ; ; ; ; in, is the original total active load of the i-th distributed photovoltaic user at time t, is the fixed load of the i-th distributed photovoltaic user at time t, is the load that can be increased by the i-th distributed photovoltaic user at time t, The time-shiftable load of the i-th distributed photovoltaic user at time t; is the total active load after the active load response of the i-th distributed photovoltaic user at time t, is the actual load increase of the i-th distributed photovoltaic user at time t, is the actual load time shift of the i-th distributed photovoltaic user at time t; Represents a time collection.

[0010] Preferably, the reward function is constructed based on the current income of distributed photovoltaic users, and the formula is expressed as: ; in, is the reward function of the i-th distributed photovoltaic user at time t, is the photovoltaic grid-connected electricity price coefficient of the current distributed photovoltaic users, The response cost when distributed photovoltaic users respond to active loads, is the original total active load of the i-th distributed photovoltaic user at time t; and are the reward coefficient and penalty coefficient of active load income respectively; is the total active load response of the i-th distributed photovoltaic user at time t, is the overvoltage responsibility value of the i-th distributed photovoltaic user at time t, is the power reverse transmission quota value of the i-th distributed photovoltaic user at time t.

[0011] Preferably, the active distribution network calculates the power reverse transmission quota value and overvoltage responsibility value of each distributed photovoltaic user, including: Obtain the active power-flow influence matrix and the active power-voltage influence matrix, and calculate the line flow influence score matrix and the node voltage influence score matrix of the distributed photovoltaic users on the active distribution network respectively; take the average of the line flow influence score matrix and the node voltage influence score matrix to obtain the comprehensive influence score matrix of the distributed photovoltaic users on the active distribution network; Based on the size of the elements in the comprehensive influence score matrix, a power reverse transmission quota is allocated to each distributed photovoltaic user, and the power reverse transmission quota value of the distributed photovoltaic user at each moment is obtained; When the power reversed by a distributed photovoltaic user exceeds its corresponding power reverse quota value and causes the active distribution network voltage to exceed the limit, the voltage over-limit severity index at the current moment is calculated, and the overvoltage responsibility value of each distributed photovoltaic user at the current moment is calculated.

[0012] Preferably, the comprehensive influence score matrix of distributed photovoltaic users on the active distribution network is expressed as: ; in, The comprehensive influence score matrix of distributed photovoltaic users on the active distribution network; is the score matrix of the influence of distributed photovoltaic users on the line flow of the active distribution network, The score of the influence of the nth distributed photovoltaic user on the line flow of the active distribution network; is the score matrix of the influence of distributed photovoltaic users on the node voltage of the active distribution network, It is the influence score of the nth distributed photovoltaic user on the node voltage of the active distribution network.

[0013] Preferably, the grey correlation analysis method is used to calculate the line flow influence score and node voltage influence score of the i-th distributed photovoltaic user on the active distribution network.

[0014] Preferably, the grey correlation analysis method is used to calculate the line flow influence score of the i-th distributed photovoltaic user on the active distribution network, including: Calculate the normalized active power-power flow influence matrix ; Construct an ideal active power-flow influence matrix based on the maximum influence of distributed photovoltaic users on each branch of the active distribution network after access ; The influence of distributed photovoltaic users on the branch line flow is calculated based on the correlation between the active power-flow influence and the ideal active power-flow influence. The formula is: ; in, is the influence of the line flow of the i-th distributed photovoltaic user on the l-th branch, is the number of distributed photovoltaic users, is the number of branches; is the difference between the actual influence and ideal influence of the i-th distributed photovoltaic user on the l-th branch, is a matrix The value of the element in row l and column i, is a matrix The element value of row l and column i; is a matrix With the matrix The minimum element difference of ; is a matrix With the matrix The maximum value of the element difference; is the resolution coefficient; The weight coefficient of the influence of distributed photovoltaic users on the branch line flow is calculated using the following formula: ; in, is the weight coefficient of the influence of the i-th distributed photovoltaic user on the line flow of the l-th branch, is a matrix No. The value of the element in the row i-th column; Calculate the influence score of each distributed photovoltaic user on the line flow of the active distribution network. The formula is expressed as: ; in, is the influence score of the i-th distributed photovoltaic user on the line flow of the active distribution network.

[0015] Preferably, the voltage over-limit severity index at the current moment is calculated using the formula: ; in, is the voltage over-limit severity index at time t, is the set of nodes whose voltage exceeds the limit in the active distribution network at time t; is the voltage of the jth node at time t, is the upper limit of the active distribution network voltage, Indicates the rated voltage of the active distribution network; The Shapley value method is used to calculate the overvoltage responsibility value of each distributed photovoltaic user at the current moment. The formula is: ; in, is the overvoltage responsibility value of the i-th distributed photovoltaic user at time t, and S is any alliance including the i-th distributed photovoltaic user connected to node j; is the voltage over-limit severity indicator when only the reverse power of the nodes in alliance S is greater than the power reverse quota; is the voltage over-limit severity indicator after removing the i-th distributed photovoltaic user connected to node j from alliance S; M is the set of all distributed photovoltaic entities whose reverse power is greater than the power reverse quota at time t; |S| and |M| represent the number of elements in S and M, respectively.

[0016] The present invention also provides a distribution network active and reactive coordinated voltage control system, comprising: The active distribution network agent is used to construct the observation state based on the voltage of each node, and the action state based on the capacitive reactive power injected by each static VAR compensator into its connected node, as well as the reward coefficient and penalty coefficient of the active load benefit. The reward function is constructed based on the total cost of distribution network operation. The active distribution network agent obtains the voltage of each node at each moment, outputs the reward coefficient and penalty coefficient of the active load benefit at the current moment to each distributed photovoltaic agent, and outputs the capacitive reactive power injected by each static VAR compensator into its connected node at the current moment, and regulates each static VAR compensator. Multiple distributed photovoltaic intelligent agents are used to construct the observation state of the current distributed photovoltaic intelligent agent based on the power reverse transmission quota value, overvoltage responsibility value and total time-shifted load of the current distributed photovoltaic user; the active load of the current distributed photovoltaic user is divided into fixed load, increaseable load and time-shiftable load, and the action state is constructed based on the actual load increase and actual load time-shift of the current distributed photovoltaic user; the reward function is constructed based on the current distributed photovoltaic user's income; each distributed photovoltaic intelligent agent obtains the power reverse transmission quota value and overvoltage responsibility value at the current moment, as well as the reward coefficient and penalty coefficient of the active load income sent by the active distribution network intelligent agent, outputs the actual load increase and actual load time-shift at the current moment, and responds to the active load.

[0017] The above technical solution of the present invention has the following beneficial effects compared with the prior art: The present invention discloses a method for coordinated voltage control of active and reactive power in a distribution network. The method constructs an active distribution network intelligent body and a distributed photovoltaic intelligent body. The active distribution network intelligent body outputs the capacitive reactive power injected into its connected nodes by each static VAR compensator to regulate the reactive equipment. The method outputs a reward and penalty coefficient for active load benefits to guide distributed photovoltaic users in the distribution network to perform flexible active load response. In each distributed photovoltaic intelligent body, its corresponding power reverse transmission quota value and overvoltage responsibility value are introduced, and the reward and penalty coefficient for active load benefits output by the active distribution network intelligent body is received. A profit strategy is constructed that comprehensively considers the power generation, access location, and topology of the entire system of distributed photovoltaic users. The strategy guides distributed photovoltaic users to reduce their adverse effects on the distribution network by actively increasing their active load or changing their electricity consumption time. The method also allows users to obtain fair and reasonable photovoltaic power generation benefits when responding to the distribution network strategy. The present invention takes into account both the overall optimization of the power grid and the individual interests of users, can effectively improve the voltage control efficiency of distribution networks with high photovoltaic penetration rates, and redistributes the benefits of distributed photovoltaic entities in the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments of the present invention in conjunction with the accompanying drawings, wherein: Figure 1 This is a framework diagram of a method for coordinated voltage control of active and reactive power in a distribution network according to the present invention; Figure 2 It is a flow chart of a method for coordinated voltage control of active and reactive power in a distribution network according to the present invention. DETAILED DESCRIPTION

[0019] The present invention will be further described below with reference to the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it. However, the embodiments are not intended to limit the present invention.

[0020] The first embodiment of the present invention provides a method for coordinated active and reactive voltage control in a distribution network. By allocating voltage over-limit issues to participating users across the distribution network, and leveraging the cooperative and game-playing relationship between the active distribution network and distributed photovoltaic users, an effective active load revenue strategy is customized to guide the active demand response of distributed photovoltaic users. This method further explores the collaborative optimization capabilities of active and reactive power when a high proportion of distributed photovoltaic power is connected to the grid. In this process, the distribution network is treated as an intelligent agent, and each user with distributed photovoltaic power is also treated as an intelligent agent, forming a multi-agent collaborative control framework.

[0021] Preferably, the present invention adopts a hierarchical control architecture to realize multi-agent collaborative optimization, the upper layer is an active distribution network agent, and the lower layer is multiple distributed photovoltaic user agents. Figure 1Each agent is built based on the TD3 (Twin Delayed Deep Deterministic Policy Gradient) algorithm. The TD3 algorithm, with its dual network, target policy smoothing regularization, and delayed update strategy, enables more stable output of the coordinated voltage control strategy.

[0022] Specifically, the TD3 algorithm perceives the state of the agent's environment and selects an action to respond based on the perceived results. This process is recorded in an experience replay pool, storing the interaction between the active distribution network agent and the distributed photovoltaic user agent as samples for training the agent. This includes the action strategy and reward-penalty strategy of the static VAR compensator of the active distribution network, the active load response strategy of the distributed photovoltaic agent, and the state space before and after coordinated control and the reward value after executing the strategy. After multiple training iterations, the agent will obtain a set of strategies that can achieve the best response to different environments, which is used to continuously obtain the maximum reward. Through the reinforcement learning mechanism, this method realizes dynamic strategy optimization for the active and reactive power coordination problem.

[0023] The active distribution network uses the real-time status of the current distribution network as observation input, controls the incentive strategy parameters considering voltage over-limit responsibility, and adjusts the output of reactive equipment, aiming to minimize the overall cost.

[0024] The active distribution network agent constructs the observation state based on the voltage of each node. The observation state of the active distribution network agent at time t is defined as , specifically including the voltage vector of each node measured by the active distribution network , recorded as , where t represents the operating time of the distribution network, and N is the number of observable nodes in the active distribution network. In practice, nodes connected to distributed photovoltaics and some key nodes can be used as observable nodes. At each time step, the active distribution network agent observes the initial state of the distribution network as observation input, and records its own actions and the new distribution network state after the distributed photovoltaic agent takes action as post-action observations into the experience replay pool.

[0025] The active distribution network agent constructs its action state based on the capacitive reactive power injected by each static VAR compensator into its connected nodes, the reward coefficient and penalty coefficient of the active load income. In the real-time action phase, the active distribution network agent's action state is defined by using the adjustable range of the static VAR compensator (SVC), the reactive power regulation device of the distribution network, and the reward-penalty coefficient of the active load income. , where m is the number of static VAR compensators, is the capacitive reactive power injected into the connection node by the mth static VAR compensator at time t, and are the reward coefficient and penalty coefficient of active load income at time t respectively.

[0026] The active distribution network intelligent agent constructs a reward function based on the total cost of distribution network operation. The present invention constructs the total cost of active distribution network operation based on distribution network loss cost, voltage over-limit cost, active load cost, and static VAR compensator operation cost, and designs a multi-objective weighted function with the goal of minimizing the total cost of active distribution network operation. Among them, the distribution network loss cost and voltage over-limit cost serve as indicators of safe operation of the distribution network, guiding the distribution network to improve towards a safe and stable operating state; the active load cost and static VAR compensator operation cost guide the active distribution network to transfer excessive voltage regulation tasks to distributed photovoltaic prosumers, and coordinate the economy and rationality of operating costs.

[0027] Specifically, in the Markov decision process, the real-time reward function of the active distribution network agent at each iteration is: ; ; in, is the reward function of the active distribution network agent at time t; is the marginal network loss coefficient, is the sum of the losses of the lines in the active distribution network at time t; is the voltage over-limit additional cost coefficient, The node voltage exceeds the limit at time t; is the active load response cost corresponding to the reward coefficient and penalty coefficient of the current active load benefit, is the reactive expenditure under the current active load response; is the set of nodes in the active distribution network, is the voltage of the jth node at time t, and are the upper and lower limits of the active distribution network voltage, is a linear rectification function.

[0028] From the perspective of the distribution network, when the active distribution network transmits the reward-penalty coefficient of active load benefits, it can dynamically adjust the reward-penalty coefficient in the early stage of training to improve the participation of distributed photovoltaic users with a more aggressive strategy. When the interaction between the active distribution network and distributed photovoltaic prosumers tends to be stable, the reward-penalty coefficient is adjusted to optimize the strategy expenditure, thereby improving the safety and economy of the active distribution network operation.

[0029] Multiple distributed photovoltaic user agents in the lower layer monitor the status of their own distributed photovoltaic power generation in real time and optimize their power consumption strategies based on the reward-penalty coefficients for active load revenue published by the distribution network to maximize individual benefits for distributed photovoltaic users. Based on the current "self-generation for own use, with surplus power fed to the grid" operating model, this invention uses a demand response mechanism to guide user active load adjustments, proactively increasing local consumption during peak photovoltaic output periods.

[0030] Each distributed photovoltaic agent constructs the observation state of the current distributed photovoltaic agent based on the power reverse quota value, overvoltage responsibility value and total time-shifted load of the current distributed photovoltaic user. The observation state of the nth distributed photovoltaic user agent at time t is defined as ,in is the current power reverse transmission quota value of distributed photovoltaic users, is the overvoltage responsibility value of the current distributed photovoltaic user, The total time-shifted load of the current distributed PV user is calculated by the active distribution network and then transmitted to each distributed PV user.

[0031] The current active load of distributed photovoltaic users is divided into fixed load, scalable load, and time-shiftable load. Before constructing the action state of the distributed photovoltaic intelligent body, the active load behavior of distributed photovoltaic users is first modeled, and the current active load of distributed photovoltaic users is divided into fixed load, scalable load, and time-shiftable load. Among them, fixed load refers to the user's daily fixed electricity consumption that does not change; time-shiftable load represents that part of the electricity consumption during the period can be transferred to other times of the same day for peak shaving and valley filling; scalable load represents the distributed photovoltaic user's realization of the mismatch between their photovoltaic power generation and electricity load, and chooses to increase electricity consumption.

[0032] The active load of the current distributed photovoltaic users is divided into fixed load, ramp-up load and time-shiftable load. The formula is: ; ; ; ; ; in, is the original total active load of the i-th distributed photovoltaic user at time t, is the fixed load of the i-th distributed photovoltaic user at time t, is the load that can be increased by the i-th distributed photovoltaic user at time t, The time-shiftable load of the i-th distributed photovoltaic user at time t; is the total active load after the active load response of the i-th distributed photovoltaic user at time t, is the actual load increase of the i-th distributed photovoltaic user at time t, is the actual load time shift of the i-th distributed photovoltaic user at time t; Represents a time collection.

[0033] Each distributed photovoltaic intelligent agent builds an action state based on the actual load increase and actual load time shift of the current distributed photovoltaic user ,in represents the action state of the nth distributed photovoltaic agent, is the actual load time shift of the current distributed photovoltaic users, that is, ; is the actual load increase of the current distributed photovoltaic users, that is, .

[0034] Each distributed PV agent constructs a reward function based on the current revenue of distributed PV users. To achieve a game of bargaining between the distribution network and distributed PV users, the distributed PV agent must constantly prioritize its own revenue and continuously adjust its response to the active load revenue strategy. By constructing a convex reward function and a linear penalty function, a composite mechanism combining convex and linear functions is formed. This structure can leverage convex rewards to stimulate the active load response potential of distributed PV prosumers when penalties are unavoidable when the distributed PV output is excessive.

[0035] Specifically, the reward function is constructed based on the current income of distributed photovoltaic users, and the formula is expressed as: ; in, is the reward function of the i-th distributed photovoltaic user at time t, is the photovoltaic grid-connected electricity price coefficient of the current distributed photovoltaic users, The response cost when distributed photovoltaic users respond to active loads, is the original total active load of the i-th distributed photovoltaic user at time t; and are the reward coefficient and penalty coefficient of active load income respectively; is the total active load response of the i-th distributed photovoltaic user at time t, is the overvoltage responsibility value of the i-th distributed photovoltaic user at time t, is the power reverse transmission quota value of the i-th distributed photovoltaic user at time t.

[0036] Specifically, the income of distributed photovoltaic prosumers is composed of photovoltaic grid-connected electricity price income. , actively cooperate to reduce the reward for net backfeeding , Penalties for exceeding the photovoltaic output quota , Actively cooperate with costs In addition to being determined by the responses of distributed PV users themselves, rewards and penalties are also related to the real-time reward and penalty coefficients transmitted by the active distribution network, forming a collaborative control model. Under this revenue model, the profits of distributed PV users will not only be related to the amount of power generated by "surplus power grid connection", but also coupled with the safe operation of the distribution network, favoring users who can absorb their own excess photovoltaic power.

[0037] like Figure 2 As shown, the present invention proposes a method for coordinated voltage control of active and reactive power in a distribution network, and the specific steps are as follows: The active distribution network agent obtains the voltage of each node at every moment and outputs the reward coefficient and penalty coefficient of the active load benefit at the current moment to each distributed photovoltaic agent. It also outputs the capacitive reactive power injected by each static VAR compensator into its connected node at the current moment and regulates each static VAR compensator. Each distributed photovoltaic intelligent agent obtains the current power reverse transmission quota value and overvoltage responsibility value, as well as the reward coefficient and penalty coefficient of the active load benefit sent by the active distribution network intelligent agent, outputs the actual load increase and actual load time shift at the current moment, and performs active load response.

[0038] Specifically, in the multi-agent collaborative control model proposed by the present invention based on active and reactive coordinated voltage control of the distribution network, the multi-agent offline training process is as follows: The historical data of the interaction between the active distribution network and the photovoltaic agent during the operation of the active distribution network is used as offline learning data. Data collection is divided into a time step every 15 minutes, and 96 time steps are divided into 24 hours per day. The active distribution network operation data is used as the offline training data set. Each active distribution network agent or distributed photovoltaic agent has its own two actor networks and four critic networks based on the TD3 algorithm, which serve as the agent's strategy network; the experience replay pool of the distributed photovoltaic agent only includes its own observation state. , action status ,award And the observed state after the action ,Will The experience replay pool stored in the distributed photovoltaic intelligent agent is consistent with the information acquisition of distributed photovoltaic producers and consumers during the actual operation of the distribution network; the experience replay pool of the active distribution network intelligent agent includes all observation states, action states, rewards and observation states after the action in the entire multi-agent collaborative model , and put it into the experience replay pool of the active distribution network agent. At each time step, the active distribution network agent and the distributed photovoltaic agent store their offline learning data into the corresponding replay pool until the experience replay pool reaches the upper limit.

[0039] The active distribution network agent and the distributed photovoltaic producer and seller agent achieve strategy convergence through iterative interaction. The active distribution network agent generates the action state of the incentive strategy and reactive power regulation command according to the real-time status. ; Distributed photovoltaic intelligent body builds observation state after receiving parameters , adjust the load to optimize local benefits; the system updates the state and calculates the reward, and updates the policy network through the TD3 algorithm; repeat the above training process until the upper limit of training rounds is reached, and at the same time, the distribution network and distributed photovoltaics reach a balanced state of voltage stability and optimal cost.

[0040] In the strategy update process of the TD3 algorithm, the two Critic networks are updated by obtaining the action state under the observation state s in the target Actor network. ; It is the random noise added when the TD3 algorithm is running, which satisfies the normal distribution At the same time, the clipping function clip is used to cut the noise, and the clipping range is ±x to ensure the generalization of the action after adding noise; random noise is added in the process of calculating the action state to avoid falling into poor action too early. Then, the target value is calculated through two target Critic networks ,in, represents the reward value at time t, represents the weight coefficient; is the smaller value of the two target Critic network estimates, represents the i-th target Critic network, and Represent the observation state and action state at time t+1 respectively, Represents the parameters of the i-th target Critic network. Then the difference between the evaluation value and the target value is calculated by gradient descent method. Minimize to update the parameters of the two Critic networks; represents the i-th Critic network, and They represent the observation state and action state at time t, respectively. Represents the parameters of the i-th Critic network. After the Critic network is updated m steps, the Actor network is updated and the action state when the observation state s is obtained using the Actor network. , ,in Represents the Actor network, Represents the parameters of the Actor network. Finally, the observation state and action state are obtained through the Critic network. The assessed value and through Update the Actor network, such as .

[0041] This active distribution network directly controls the reactive power of the reactive compensation equipment and indirectly guides the active power of the distributed photovoltaic users' active load response, together forming an active and reactive coordinated voltage control method. The two are coupled through a deep reinforcement learning algorithm, voltage over-limit risk, and photovoltaic output quota.

[0042] Under normal circumstances, electricity flows from the grid to users, who consume active power. However, when users have their own power generation equipment, such as distributed photovoltaic systems or small wind turbines, and their power generation exceeds their own electricity needs, the excess active power may be fed back into the distribution network. This is known as active power reverse transmission. When a high proportion of distributed photovoltaic power is connected to the distribution network, the high midday photovoltaic power generation exceeds user demand, which can cause overvoltage problems in the distribution network without energy storage. In this interactive energy consumption model, a mechanism should exist to distribute the responsibility for overvoltage to each distributed photovoltaic user.

[0043] The power reverse transmission quota value in this invention refers to the amount of active power that distributed photovoltaic users are allowed to transmit back to the distribution network in the power system. The overvoltage responsibility value refers to the overvoltage responsibility that each distributed photovoltaic user should bear when the distribution network transmits overvoltage.

[0044] Preferably, the active distribution network calculates the power reverse transmission quota value and overvoltage responsibility value of each distributed photovoltaic user, including: S1: Use the line parameters of the active distribution network to obtain the node-susceptance matrix and the line-susceptance matrix ; Assume that the node model of the distribution network contains N nodes and L lines, with a total of n DPV nodes, is the reactance between nodes k and j.

[0045] Construct the relationship between node active power, line flow and voltage phase angle: ; ; in, and The active power injected into the node and the active power flowing through each line are respectively; Indicates the voltage phase angle.

[0046] The relationship between node injection power and line power flow is: ; Get the power transfer distribution factor matrix .

[0047] S2: The power-voltage sensitivity matrix can be further calculated from the Jacobian determinant of the power system flow calculation method, which can effectively reflect the impact of distributed photovoltaic output on the node voltage amplitude.

[0048] ; Among them, the previous formula is the tidal differential equation in polar coordinate form; 、 、 、 They are the active power, reactive power, voltage and phase angle of the node respectively; To obtain partial derivative operations; is the Jacobian matrix, is the inverse matrix of the Jacobian matrix; the power-voltage sensitivity matrix is the inverse Jacobian matrix The lower left block matrix of , that is, the effect of injected active power on the voltage amplitude.

[0049] S3: When the i-th node is a distributed photovoltaic user, the power transmission distribution factor matrix and the power-voltage sensitivity matrix can be used to obtain the active power-flow impact matrix of the i-th node as a distributed photovoltaic user on each branch and the active power-voltage impact matrix on each node.

[0050] ; ; in, represents the active power-power flow impact matrix of the i-th distributed photovoltaic user, , n represents the number of distributed photovoltaic users, represents the active power-power flow influence matrix; represents the active power-voltage impact matrix of the i-th distributed photovoltaic user, Represents the active power-voltage influence matrix.

[0051] S4: Based on the active power-flow influence matrix and the active power-voltage influence matrix, calculate the line power flow influence score matrix and the node voltage influence score matrix of the distributed PV user on the active distribution network. Take the average of the line power flow influence score matrix and the node voltage influence score matrix to obtain the comprehensive influence score matrix of the distributed PV user on the active distribution network.

[0052] Preferably, the grey correlation analysis method is used to calculate the line flow influence score and node voltage influence score of the i-th distributed photovoltaic user on the active distribution network.

[0053] S41: Use the grey correlation analysis method to calculate the line flow influence score of the i-th distributed photovoltaic user on the active distribution network, including: S41-1: Calculate the normalized active power-flow influence matrix : ; in, Represents the active power-power influence matrix The element value in the lth row and i-th column is the influence of the i-th distributed photovoltaic user on the line flow of the l-th branch; represents the element value of the lth row and ith column of the normalized active power-flow influence matrix; represents the normalized active power-power flow impact matrix of the nth distributed photovoltaic user.

[0054] S41-2: Construct an ideal active power-power influence matrix based on the maximum influence of distributed photovoltaic users on each branch of the active distribution network after access .

[0055] S41-3: The influence of distributed photovoltaic users on the branch line flow is calculated based on the correlation between the active power-flow influence and the ideal active power-flow influence. The formula is: ; in, is the influence of the line flow of the i-th distributed photovoltaic user on the l-th branch, is the number of distributed photovoltaic users, is the number of branches; is the difference between the actual influence and ideal influence of the i-th distributed photovoltaic user on the l-th branch, is a matrix The value of the element in row l and column i, is a matrix The element value of row l and column i; is a matrix With the matrix The minimum element difference of ; is a matrix With the matrix The maximum value of the element difference; is the resolution coefficient.

[0056] S41-4: Calculate the weight coefficient of the influence of distributed photovoltaic users on the branch line flow. The formula is: ; in, is the weight coefficient of the influence of the i-th distributed photovoltaic user on the line flow of the l-th branch, is a matrix No. The value of the element in the row i-th column.

[0057] S41-5: Calculate the influence score of each distributed photovoltaic user on the line flow of the active distribution network. The formula is expressed as: ; in, is the influence score of the i-th distributed photovoltaic user on the line flow of the active distribution network.

[0058] S41-6: Constructing a scoring matrix for the influence of distributed photovoltaic users on the line flow of active distribution networks .

[0059] S42: Use the grey correlation analysis method to calculate the node voltage influence score of the i-th distributed photovoltaic user on the active distribution network, including: S42-1: Calculate the normalized active power-voltage influence matrix : ; in, Represents the original active-voltage influence matrix The element value in the lth row and i-th column is the influence of the i-th distributed photovoltaic user on the node voltage of the l-th branch; represents the element value of the lth row and ith column of the normalized active-voltage influence matrix; represents the normalized active-voltage impact matrix of the nth distributed photovoltaic user.

[0060] S42-2: Construct an ideal active power-voltage influence matrix based on the maximum influence of distributed photovoltaic users on each node of the active distribution network after access .

[0061] S42-3: The influence of distributed photovoltaic users on node voltage is calculated based on the correlation between active power-voltage influence and ideal active power-voltage influence. The formula is: ; in, is the influence of the i-th distributed photovoltaic user on the node voltage of the j-th node, is the number of distributed photovoltaic users, is the number of nodes; is the difference between the actual influence and ideal influence of the i-th distributed photovoltaic user on the j-th node, is a matrix The value of the element in row j and column i, is a matrix The value of the element in row j and column i; is a matrix With the matrix The minimum element difference of ; is a matrix With the matrix The maximum value of the element difference; is the resolution coefficient.

[0062] S42-4: Calculate the weight coefficient of the influence of distributed photovoltaic users on the node voltage. The formula is: ; in, is the weight coefficient of the influence of the i-th distributed photovoltaic user on the node voltage of the j-th node, is a matrix No. The value of the element in the row i-th column.

[0063] S42-5: Calculate the influence score of each distributed photovoltaic user on the node voltage of the active distribution network. The formula is: ; in, is the influence score of the i-th distributed photovoltaic user on the node voltage of the active distribution network.

[0064] S42-6: Constructing a scoring matrix for the influence of distributed photovoltaic users on node voltage of active distribution networks .

[0065] S43: Take the average of the line flow influence score matrix and the node voltage influence score matrix to obtain the comprehensive influence score matrix of distributed photovoltaic users on the active distribution network. The formula is: ; in, The comprehensive influence score matrix of distributed photovoltaic users on the active distribution network; is the score matrix of the influence of distributed photovoltaic users on the line flow of the active distribution network, The score of the influence of the nth distributed photovoltaic user on the line flow of the active distribution network; is the score matrix of the influence of distributed photovoltaic users on the node voltage of the active distribution network, It is the influence score of the nth distributed photovoltaic user on the node voltage of the active distribution network.

[0066] The grey correlation analysis method used in the present invention can optimize the proportion of quota allocation by adjusting the resolution coefficient. Compared with the approximate ideal solution and the traditional average allocation method, it can more finely divide the photovoltaic output quota.

[0067] S5: Allocate power reverse transmission quota to each distributed photovoltaic user based on the size of the element value in the comprehensive influence score matrix, that is, based on the size of the element value in the matrix M, and obtain the power reverse transmission quota value of the distributed photovoltaic user at each moment. ,in Indicates the power reverse transmission quota value of the nth distributed photovoltaic user.

[0068] S6: When the power reversed by the distributed photovoltaic user exceeds its corresponding power reverse quota value and causes the active distribution network voltage to exceed the limit, calculate the voltage limit severity index at the current moment and calculate the overvoltage responsibility value of each distributed photovoltaic user at the current moment.

[0069] This embodiment defines the voltage over-limit severity index at the current moment, and the formula is: ; in, is the voltage over-limit severity index at time t, is the set of nodes whose voltage exceeds the limit in the active distribution network at time t; is the voltage of the jth node at time t, is the upper limit of the active distribution network voltage, Indicates the rated voltage of the active distribution network.

[0070] The Shapley value method is used to calculate the overvoltage responsibility value of each distributed photovoltaic user at the current moment. The formula is: ; in, is the overvoltage responsibility value of the i-th distributed photovoltaic user at time t, and S is any alliance including the i-th distributed photovoltaic user connected to node j; is the voltage over-limit severity indicator when only the reverse power of the nodes in alliance S is greater than the power reverse quota; is the voltage over-limit severity indicator after removing the i-th distributed photovoltaic user connected to node j from alliance S; M is the set of all distributed photovoltaic entities whose reverse power is greater than the power reverse quota at time t; |S| and |M| represent the number of elements in S and M, respectively.

[0071] After calculating each node containing distributed photovoltaic users, the power reverse transmission quota value and overvoltage responsibility value of the distributed photovoltaic users are passed to the distributed photovoltaic users. The distributed photovoltaic users will adjust their electricity consumption behavior according to the demand response status of their own active load and the reward and punishment coefficient of the active load, striving to obtain the maximum benefit.

[0072] In summary, the present invention provides a method for coordinated voltage control of active and reactive power in a distribution network. The method constructs an active distribution network intelligent body and a distributed photovoltaic intelligent body. The active distribution network intelligent body outputs the capacitive reactive power injected into its connected node by each static VAR compensator to regulate the reactive equipment. The active load benefit reward and penalty coefficient is output to guide the distributed photovoltaic users in the distribution network to perform flexible active load response. In each distributed photovoltaic intelligent body, its corresponding power reverse transmission quota value and overvoltage responsibility value are introduced, and the active load benefit reward and penalty coefficient output by the active distribution network intelligent body is received. An active load benefit strategy is constructed that comprehensively considers the power generation, access location and topology of the entire system of distributed photovoltaic users. The strategy guides distributed photovoltaic users to reduce their adverse effects on the distribution network by actively increasing their active load or changing their electricity consumption time, and enables users to obtain fair and reasonable photovoltaic power generation benefits when responding to the distribution network strategy. The present invention takes into account both the overall optimization of the power grid and the individual interests of users, can effectively improve the voltage control efficiency of the distribution network with high photovoltaic penetration, and redistributes the benefits of distributed photovoltaic entities in the distribution network.

[0073] Based on the distribution network active and reactive coordinated voltage control method described in Example 1, this embodiment provides a distribution network active and reactive coordinated voltage control system, including: The active distribution network agent is used to construct the observation state based on the voltage of each node, and the action state based on the capacitive reactive power injected by each static VAR compensator into its connected node, as well as the reward coefficient and penalty coefficient of the active load benefit. The reward function is constructed based on the total cost of distribution network operation. The active distribution network agent obtains the voltage of each node at each moment, outputs the reward coefficient and penalty coefficient of the active load benefit at the current moment to each distributed photovoltaic agent, and outputs the capacitive reactive power injected by each static VAR compensator into its connected node at the current moment, and regulates each static VAR compensator. Multiple distributed photovoltaic intelligent agents are used to construct the observation state of the current distributed photovoltaic intelligent agent based on the power reverse transmission quota value, overvoltage responsibility value and total time-shifted load of the current distributed photovoltaic user; the active load of the current distributed photovoltaic user is divided into fixed load, increaseable load and time-shiftable load, and the action state is constructed based on the actual load increase and actual load time-shift of the current distributed photovoltaic user; the reward function is constructed based on the current distributed photovoltaic user's income; each distributed photovoltaic intelligent agent obtains the power reverse transmission quota value and overvoltage responsibility value at the current moment, as well as the reward coefficient and penalty coefficient of the active load income sent by the active distribution network intelligent agent, outputs the actual load increase and actual load time-shift at the current moment, and responds to the active load.

[0074] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0075] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. 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 produce 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.

[0076] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work 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 The function specified in one or more boxes.

[0077] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device 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.

[0078] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications arising therefrom remain within the scope of protection of the present invention.

Claims

1. A method for coordinated voltage control of active and reactive power in a distribution network, characterized in that: include: The TD3 algorithm is used to construct an active distribution network agent and multiple distributed photovoltaic agents; In the active distribution network agent, the observation state is constructed by the voltage of each node, and the action state is constructed by the capacitive reactive power injected by each static VAR compensator into its connected node, the reward coefficient and penalty coefficient of the active load benefit; Construct a reward function based on the total cost of distribution network operation; In each distributed photovoltaic agent, the observation state of the current distributed photovoltaic agent is constructed based on the current distributed photovoltaic user's power reverse transmission quota value, overvoltage responsibility value, and total time-shifted load. The active load of the current distributed photovoltaic user is divided into fixed load, scalable load, and time-shiftable load, and the action state is constructed based on the current distributed photovoltaic user's actual load scalability and actual load time-shifting amount. The reward function is constructed based on the current distributed photovoltaic user's income. The active distribution network agent obtains the voltage of each node at every moment and outputs the reward coefficient and penalty coefficient of the active load benefit at the current moment to each distributed photovoltaic agent. It also outputs the capacitive reactive power injected by each static VAR compensator into its connected node at the current moment and regulates each static VAR compensator. Each distributed photovoltaic intelligent agent obtains the current power reverse transmission quota value and overvoltage responsibility value, as well as the reward coefficient and penalty coefficient of the active load benefit sent by the active distribution network intelligent agent, outputs the actual load increase and actual load time shift at the current moment, and performs active load response.

2. A method for controlling active and reactive power coordinated voltage in a distribution network according to claim 1, characterized in that: In the active distribution network agent, the reward function is constructed based on the total cost of distribution network operation, and the formula is expressed as: ； ； in, is the reward function of the active distribution network agent at time t; is the marginal network loss coefficient, is the sum of the losses of the lines in the active distribution network at time t; is the voltage over-limit additional cost coefficient, The node voltage exceeds the limit at time t; is the active load response cost corresponding to the reward coefficient and penalty coefficient of the current active load benefit, is the reactive expenditure under the current active load response; is the set of nodes in the active distribution network, is the voltage of the jth node at time t, and are the upper and lower limits of the active distribution network voltage, is a linear rectification function.

3. A method for coordinated voltage control of active and reactive power in a distribution network according to claim 1, characterized in that: The active load of the current distributed photovoltaic users is divided into fixed load, ramp-up load and time-shiftable load. The formula is: ； ； ； ； ； in, is the original total active load of the i-th distributed photovoltaic user at time t, is the fixed load of the i-th distributed photovoltaic user at time t, is the load that can be increased by the i-th distributed photovoltaic user at time t, The time-shiftable load of the i-th distributed photovoltaic user at time t; is the total active load after the active load response of the i-th distributed photovoltaic user at time t, is the actual load increase of the i-th distributed photovoltaic user at time t, is the actual load time shift of the i-th distributed photovoltaic user at time t; Represents a time collection.

4. A method for coordinated voltage control of active and reactive power in a distribution network according to claim 1, characterized in that: The reward function is constructed based on the current income of distributed photovoltaic users. The formula is expressed as: ； in, is the reward function of the i-th distributed photovoltaic user at time t, is the photovoltaic grid-connected electricity price coefficient of the current distributed photovoltaic users, The response cost when distributed photovoltaic users respond to active loads, is the original total active load of the i-th distributed photovoltaic user at time t; and are the reward coefficient and penalty coefficient of active load income respectively; is the total active load response of the i-th distributed photovoltaic user at time t, is the overvoltage responsibility value of the i-th distributed photovoltaic user at time t, is the power reverse transmission quota value of the i-th distributed photovoltaic user at time t.

5. The method for coordinated voltage control of active and reactive power in a distribution network according to claim 1, characterized in that: The active distribution network calculates the power reverse transmission quota and overvoltage responsibility value of each distributed photovoltaic user, including: Obtain the active power-flow influence matrix and the active power-voltage influence matrix, and calculate the line flow influence score matrix and the node voltage influence score matrix of the distributed photovoltaic users on the active distribution network respectively; take the average of the line flow influence score matrix and the node voltage influence score matrix to obtain the comprehensive influence score matrix of the distributed photovoltaic users on the active distribution network; Based on the size of the elements in the comprehensive influence score matrix, a power reverse transmission quota is allocated to each distributed photovoltaic user, and the power reverse transmission quota value of the distributed photovoltaic user at each moment is obtained; When the power reversed by a distributed photovoltaic user exceeds its corresponding power reverse quota value and causes the active distribution network voltage to exceed the limit, the voltage over-limit severity index at the current moment is calculated, and the overvoltage responsibility value of each distributed photovoltaic user at the current moment is calculated.

6. A distribution network active and reactive coordinated voltage control method according to claim 5, characterized in that: The comprehensive influence score matrix of distributed photovoltaic users on the active distribution network is expressed as: ； in, The comprehensive influence score matrix of distributed photovoltaic users on the active distribution network; is the score matrix of the influence of distributed photovoltaic users on the line flow of the active distribution network, The score of the influence of the nth distributed photovoltaic user on the line flow of the active distribution network; is the score matrix of the influence of distributed photovoltaic users on the node voltage of the active distribution network, It is the influence score of the nth distributed photovoltaic user on the node voltage of the active distribution network.

7. A method for coordinated voltage control of active and reactive power in a distribution network according to claim 6, characterized in that: The grey correlation analysis method is used to calculate the line flow influence score and node voltage influence score of the i-th distributed photovoltaic user on the active distribution network.

8. A method for coordinated voltage control of active and reactive power in a distribution network according to claim 7, characterized in that: The grey correlation analysis method is used to calculate the line flow influence score of the i-th distributed photovoltaic user on the active distribution network, including: Calculate the normalized active power-power flow influence matrix ; Construct an ideal active power-flow influence matrix based on the maximum influence of distributed photovoltaic users on each branch of the active distribution network after access ; The influence of distributed photovoltaic users on the branch line flow is calculated based on the correlation between the active power-flow influence and the ideal active power-flow influence. The formula is: ； in, is the influence of the line flow of the i-th distributed photovoltaic user on the l-th branch, is the number of distributed photovoltaic users, is the number of branches; is the difference between the actual influence and ideal influence of the i-th distributed photovoltaic user on the l-th branch, is a matrix The value of the element in row l and column i, is a matrix The element value of row l and column i; is a matrix With the matrix The minimum element difference of ; is a matrix With the matrix The maximum value of the element difference; is the resolution coefficient; The weight coefficient of the influence of distributed photovoltaic users on the branch line flow is calculated using the following formula: ； in, is the weight coefficient of the influence of the i-th distributed photovoltaic user on the line flow of the l-th branch, is a matrix No. The value of the element in the row i-th column; Calculate the influence score of each distributed photovoltaic user on the line flow of the active distribution network. The formula is expressed as: ； in, is the influence score of the i-th distributed photovoltaic user on the line flow of the active distribution network.

9. A method for controlling active and reactive power coordinated voltage in a distribution network according to claim 5, characterized in that ,Calculate the voltage over-limit severity index at the current moment, the formula is: ； in, is the voltage over-limit severity index at time t, is the set of nodes whose voltage exceeds the limit in the active distribution network at time t; is the voltage of the jth node at time t, is the upper limit of the active distribution network voltage, Indicates the rated voltage of the active distribution network; The Shapley value method is used to calculate the overvoltage responsibility value of each distributed photovoltaic user at the current moment. The formula is: ； in, is the overvoltage responsibility value of the i-th distributed photovoltaic user at time t, and S is any alliance including the i-th distributed photovoltaic user connected to node j; is the voltage over-limit severity indicator when only the reverse power of the nodes in alliance S is greater than the power reverse quota; is the voltage over-limit severity indicator after removing the i-th distributed photovoltaic user connected to node j from alliance S; M is the set of all distributed photovoltaic entities whose reverse power is greater than the power reverse quota at time t; |S| and |M| represent the number of elements in S and M, respectively.

10. A distribution network active and reactive coordinated voltage control system, characterized in that: include: The active distribution network agent is used to construct the observation state based on the voltage of each node, and the action state based on the capacitive reactive power injected by each static VAR compensator into its connected node, the reward coefficient and penalty coefficient of the active load benefit; Construct a reward function based on the total cost of distribution network operation; The active distribution network agent obtains the voltage of each node at every moment and outputs the reward coefficient and penalty coefficient of the active load benefit at the current moment to each distributed photovoltaic agent. It also outputs the capacitive reactive power injected by each static VAR compensator into its connected node at the current moment and regulates each static VAR compensator. Multiple distributed photovoltaic intelligent agents are used to construct the observation state of the current distributed photovoltaic intelligent agent based on the power reverse transmission quota value, overvoltage responsibility value and total time-shifted load of the current distributed photovoltaic user; the active load of the current distributed photovoltaic user is divided into fixed load, increaseable load and time-shiftable load, and the action state is constructed based on the actual load increase and actual load time-shift of the current distributed photovoltaic user; the reward function is constructed based on the current distributed photovoltaic user's income; each distributed photovoltaic intelligent agent obtains the power reverse transmission quota value and overvoltage responsibility value at the current moment, as well as the reward coefficient and penalty coefficient of the active load income sent by the active distribution network intelligent agent, outputs the actual load increase and actual load time-shift at the current moment, and responds to the active load.

Images