Construction method and device of distributed photovoltaic consumption capacity model, equipment and medium

Abstract

The invention discloses a distributed photovoltaic consumption capacity model construction method and device, equipment and a medium, and the method comprises the steps: determining a target function of a consumption capacity optimization model according to the maximum photovoltaic consumption capacity of a power distribution network photovoltaic power generation system and the active loss of the system; determining constraint conditions of the consumption capability optimization model according to system power flow distribution related parameters of the power distribution network photovoltaic power generation system; determining a photovoltaic consumption decision agent according to the consumption capability optimization model; and determining a final distributed photovoltaic consumption capacity model according to the photovoltaic consumption decision agent and a double-delay depth deterministic strategy gradient algorithm. The maximum boundary capacity value of multiple photovoltaic simultaneous grid connection is rapidly determined through constraint conditions, processing of continuous action space problems and fine adjustment actions are achieved through a double-delay algorithm, the algorithm training stability is improved, the problems of excessive estimation and excessive conservative are reduced, the method can adapt to the dynamic environment, and the method is suitable for large-scale popularization and application. And the accuracy of the final model is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power distribution networks, and in particular to a method, device, equipment and medium for constructing a distributed photovoltaic absorption capacity model. Background Art

[0002] In the field of distributed photovoltaic absorption capacity assessment methods, when the distribution network is connected to high-penetration photovoltaics, the distribution network changes from a simple radial network to a multi-terminal active network, which is bound to affect the voltage safety of the distribution network. Not only will the voltage of the grid node exceed the limit, but there will also be problems such as branch current overload, voltage deviation, and high-order harmonics. In this context, it is very important to study the maximum photovoltaic capacity that can be connected to the distribution network for the safe operation of the distribution network.

[0003] At present, the core of the research on the evaluation of photovoltaic absorption capacity of distribution networks at home and abroad is how to effectively calculate and optimize the absorption capacity of distributed photovoltaic power generation systems. The impact of high-proportion distributed power access on the performance of distribution networks is generally quantified as the calculation of power grid hosting capability or absorption capacity. At present, the mainstream research methods for evaluating the photovoltaic absorption capacity of distribution networks include trial methods, analytical methods, and intelligent optimization algorithms. The existing technology is usually based on the calculation method of photovoltaic absorption capacity based on statistical models or optimization algorithms.

[0004] The existing photovoltaic absorption capacity calculation methods based on statistical models or optimization algorithms have low calculation accuracy. Traditional methods may be limited by the simplification or assumptions of the model and cannot accurately reflect the actual photovoltaic absorption situation. The calculation complexity is high. Some existing methods may require a long calculation time when dealing with large-scale systems and are not efficient enough. The processing ability of complex data correlation is limited. Traditional methods may find it difficult to capture the complex correlation between photovoltaic power generation system data, resulting in the accuracy and reliability of the calculation results being affected. Summary of the invention

[0005] The present invention provides a method, device, equipment and medium for constructing a distributed photovoltaic absorption capacity model to improve the accuracy of the distributed photovoltaic absorption capacity model.

[0006] According to a first aspect of the present invention, a method for constructing a distributed photovoltaic consumption capacity model is provided, comprising:

[0007] According to the maximum photovoltaic absorption capacity of the photovoltaic power generation system of the distribution network and the active power loss of the system, the objective function of the absorption capacity optimization model is determined;

[0008] Determining the constraint conditions of the absorption capacity optimization model according to the system power flow distribution related parameters of the photovoltaic power generation system of the distribution network;

[0009] Determine a photovoltaic consumption decision-making intelligent agent according to the consumption capacity optimization model;

[0010] According to the photovoltaic absorption decision-making intelligent agent and the double-delay deep deterministic policy gradient algorithm, the final distributed photovoltaic absorption capacity model is determined.

[0011] According to a second aspect of the present invention, there is provided a device for constructing a distributed photovoltaic consumption capacity model, comprising:

[0012] A function determination module is used to determine the objective function of the absorption capacity optimization model according to the maximum absorption capacity of the photovoltaic power generation system of the distribution network and the active power loss of the system;

[0013] A condition determination module, used to determine the constraint conditions of the absorption capacity optimization model according to the system power flow distribution related parameters of the photovoltaic power generation system of the distribution network;

[0014] An intelligent agent determination module is used to determine a photovoltaic consumption decision-making intelligent agent according to the consumption capacity optimization model;

[0015] The model determination module is used to determine the final distributed photovoltaic absorption capacity model according to the photovoltaic absorption decision-making agent and the double-delay deep deterministic policy gradient algorithm.

[0016] According to a third aspect of the present invention, there is provided an electronic device, the electronic device comprising:

[0017] at least one processor; and

[0018] a memory communicatively connected to the at least one processor; wherein,

[0019] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the method for constructing a distributed photovoltaic accommodation capacity model described in any embodiment of the present invention.

[0020] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the method for constructing a distributed photovoltaic absorption capacity model described in any embodiment of the present invention when executed.

[0021] According to a fifth aspect of the present invention, an embodiment of the present invention further provides a computer program product, the computer program product comprising a computer program, which, when executed by a processor, implements a method for constructing a distributed photovoltaic accommodation capacity model according to any embodiment of the present invention.

[0022] The technical solution of the embodiment of the present invention determines the objective function of the absorption capacity optimization model according to the maximum absorption capacity of photovoltaic power generation system of distribution network and the system active power loss; determines the constraint conditions of the absorption capacity optimization model according to the system flow distribution related parameters of photovoltaic power generation system of distribution network; determines the photovoltaic absorption decision-making agent according to the absorption capacity optimization model; determines the final distributed photovoltaic absorption capacity model according to the photovoltaic absorption decision-making agent and the double-delay deep deterministic policy gradient algorithm. By setting multiple constraints, the maximum boundary capacity value of multiple photovoltaics connected to the grid at the same time can be quickly determined, reducing the calculation time of the model. Through the double-delay deep deterministic policy gradient algorithm, the processing of continuous action space problems and fine-tuning actions are realized, the algorithm training stability is improved, the problems of over-estimation and over-conservatism are reduced, and it can adapt to dynamic environments, thereby improving the accuracy of the final model and providing strong support for the planning and operation of the power system.

[0023] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 is a flow chart of a method for constructing a distributed photovoltaic consumption capacity model provided in accordance with the first embodiment of the present invention;

[0026] Figure 2 is a flow chart of a method for constructing a distributed photovoltaic consumption capacity model provided in accordance with the second embodiment of the present invention;

[0027] Figure 3 An example diagram of an intelligent agent decision-making method for constructing a distributed photovoltaic consumption capacity model is provided according to Embodiment 2 of the present invention;

[0028] Figure 4 It is an example diagram of an algorithm architecture of a method for constructing a distributed photovoltaic consumption capacity model according to a second embodiment of the present invention;

[0029] Figure 5 Schematic diagram of a distributed photovoltaic consumption capacity model construction device according to the third embodiment of the present invention

[0030] Figure 6It is a schematic diagram of the structure of an electronic device implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0031] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0032] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0033] Embodiment 1

[0034] Figure 1 A flowchart of a method for constructing a distributed photovoltaic absorption capacity model is provided for the first embodiment of the present invention. This embodiment is applicable to the case of distributed photovoltaic absorption capacity calculation. The method can be executed by a distributed photovoltaic absorption capacity model construction device. The distributed photovoltaic absorption capacity model construction device can be implemented in the form of hardware and / or software. The distributed photovoltaic absorption capacity model construction device can be configured in an electronic device. Figure 1 As shown, the method includes:

[0035] S110. Determine the objective function of the absorption capacity optimization model according to the maximum absorption capacity of the photovoltaic power generation system of the distribution network and the active power loss of the system.

[0036] In this embodiment, the photovoltaic power generation system of the distribution network can be understood as a system that connects the photovoltaic power generation system to the distribution network and transmits and distributes electricity through the distribution network. The maximum photovoltaic absorption capacity can be understood as the maximum photovoltaic power generation that can be accommodated in the power system of the distribution network. The system active loss can be understood as the energy loss caused by factors such as resistance and reactance in the power system. The absorption capacity optimization model can be understood as a model for determining the maximum absorption capacity of the photovoltaic power generation system of the distribution network. The objective function can be understood as a mathematical expression used to measure the model performance or solution value in the absorption capacity optimization model.

[0037] Specifically, as the number of photovoltaic grid-connected nodes increases, it is necessary to ensure that the established optimization model can determine the maximum boundary capacity value of multiple photovoltaic grid-connected nodes at the same time. The processor can maximize the capacity of photovoltaic access in the distribution network in the construction of the photovoltaic absorption capacity optimization model. However, as the photovoltaic grid-connected capacity increases, the active power loss of the power grid will increase. Therefore, it is not enough to only pursue the maximization of photovoltaic grid-connected capacity, but also to maximize the difference between photovoltaic absorption capacity and active power network loss to avoid unlimited increase in photovoltaic grid-connected capacity. Therefore, the processor can take the difference between the maximum photovoltaic absorption capacity and the system active power loss of the entire system as the actual photovoltaic absorption capacity, and take the maximization of the actual photovoltaic absorption capacity as the objective function.

[0038]

[0039] Where: N PV Indicates the total number of distributed photovoltaics connected to the system; represents the active power generated by the i-th distributed photovoltaic; N represents the total number of nodes in the system; P ij , Q ij Respectively represent the active power and reactive power flowing between node i and node j; U i is the node voltage of node i; l ij is the line length between node i and node j; r and x are the resistance and reactance per unit line length, respectively.

[0040] It is understandable that with the increase in the number of photovoltaic grid-connected nodes, for the case of multiple photovoltaic grid-connected at the same time, the established optimization model needs to be able to determine the maximum boundary capacity value of multiple photovoltaic grid-connected at the same time. The established photovoltaic absorption capacity optimization model, the objective function construction first considers the maximum capacity of photovoltaic access to the distribution network, and proposes a series of indicator constraints such as voltage quality, short-circuit capacity and line flow constraints under random scenarios, and establishes a multi-constraint distributed photovoltaic grid-connected capacity optimization model based on scenarios and random planning with the maximum actual photovoltaic absorption capacity of the system as the objective function. This optimization model not only considers the maximum photovoltaic absorption capacity, but also considers the active network loss of the line in the objective function, in order to minimize the impact of photovoltaic access on the distribution network.

[0041] S120. Determine the constraint conditions of the absorption capacity optimization model according to the system flow distribution related parameters of the photovoltaic power generation system of the distribution network.

[0042] In this embodiment, the system power flow distribution related parameters can be understood as the voltage from the power source to the load, the magnitude and direction of the current and the power distribution parameters. Constraints can be understood as restrictions on decision variables in the optimization problem.

[0043] Specifically, the processor can determine a series of system flow distribution related parameters and indicators such as voltage quality, short-circuit capacity, line flow constraints, etc. brought about by the actual photovoltaic integration into the distribution network, and determine the corresponding constraints of each indicator and parameter under the premise of ensuring the safe and stable operation of the distribution network through relevant standards and other means.

[0044] S130. Determine the photovoltaic absorption decision-making intelligent entity according to the absorption capacity optimization model.

[0045] In this embodiment, the photovoltaic consumption decision-making agent can be understood as a software system used to perceive environmental changes in the form of an agent and make photovoltaic consumption decisions based on learned knowledge and algorithms. For example, the agent can use algorithms such as neural networks.

[0046] Specifically, the processor can transform the constraints in the absorption capacity optimization model through the Markov decision process into the five elements required for the decision-making process, and use the transformed state elements to represent the changes in the environment and input them into the photovoltaic absorption decision-making agent, and use the transformed action elements to represent the actions taken by the photovoltaic absorption decision-making agent based on the state. The processor can also transform the objective function into a reward function element based on the action feedback taken by the photovoltaic absorption decision-making agent, the transition probability between the previous and next states of the photovoltaic absorption decision-making agent, and the attenuation function between the previous and next rewards. The five elements constitute the photovoltaic absorption decision-making agent and the overall decision-making process.

[0047] S140. Determine the final distributed photovoltaic absorption capacity model based on the photovoltaic absorption decision-making agent and the double-delay deep deterministic policy gradient algorithm.

[0048] In this embodiment, the double-delayed deep deterministic policy gradient algorithm can be understood as an algorithm for training the photovoltaic consumption decision-making agent through a dual network and a delayed update mechanism. The final distributed photovoltaic consumption capacity model can be understood as the final model after training.

[0049] Specifically, the processor can initialize the dual strategy network through the double-delayed deep deterministic policy gradient algorithm, train the photovoltaic absorption decision-making agent through the dual strategy network, and update the weights of the dual strategy network through the delayed update mechanism until the training end conditions are met, thereby obtaining the trained photovoltaic absorption decision-making agent. The processor can configure it in the final distributed photovoltaic absorption capacity model to determine the maximum photovoltaic capacity through the final distributed photovoltaic absorption capacity model.

[0050] The technical solution of the embodiment of the present invention determines the objective function of the absorption capacity optimization model according to the maximum absorption capacity of photovoltaic power generation system of distribution network and the system active power loss; determines the constraint conditions of the absorption capacity optimization model according to the system flow distribution related parameters of photovoltaic power generation system of distribution network; determines the photovoltaic absorption decision-making agent according to the absorption capacity optimization model; determines the final distributed photovoltaic absorption capacity model according to the photovoltaic absorption decision-making agent and the double-delay deep deterministic policy gradient algorithm. By setting multiple constraints, the maximum boundary capacity value of multiple photovoltaics connected to the grid at the same time can be quickly determined, reducing the calculation time of the model. Through the double-delay deep deterministic policy gradient algorithm, the processing of continuous action space problems and fine-tuning actions are realized, the algorithm training stability is improved, the problems of over-estimation and over-conservatism are reduced, and it can adapt to dynamic environments, thereby improving the accuracy of the final model and providing strong support for the planning and operation of the power system.

[0051] Embodiment 2

[0052] Figure 2 This is a flow chart of a method for constructing a distributed photovoltaic consumption capacity model provided in the second embodiment of the present invention. This embodiment is a further refinement of the above embodiment. Figure 2 As shown, the method includes:

[0053] S201. Determine the objective function of the absorption capacity optimization model according to the maximum absorption capacity of the photovoltaic power generation system of the distribution network and the active power loss of the system.

[0054] S202. Determine power flow constraint sub-conditions according to the distributed photovoltaic output injected into each node in each scenario, the power consumed by the load, the node attribute information of each node, and the distribution network attribute information of the distribution network.

[0055] In this embodiment, the scene can be understood as the operating status and environmental background of power generation, transmission, substation, distribution and power consumption. The node can be understood as the node in the power system, which refers to the connection point of equipment such as loads, generators, substations, etc., and is also the intersection of energy flows. Distributed photovoltaic output can be understood as the electric power output of the distributed photovoltaic power generation system per unit time, which can include active and reactive power. The power consumed by the load can be understood as the power consumed by the load in the system. Node attribute information can be understood as information related to node attributes such as line length between nodes, conductance, susceptance and voltage phase angle difference between nodes, node voltage, etc. The distribution network attribute information can be understood as attribute information related to the distribution network, such as the total number of nodes contained in the distribution network. The flow constraint sub-condition can be understood as the constraint conditions on the flow distribution on each node and transmission line in the power system to maintain the normal operation and safety of the system.

[0056] Specifically, the processor can determine the injected distributed photovoltaic active output and distributed photovoltaic reactive output, as well as the active power consumed by the load and the reactive power consumed by the load according to the node attribute information of each node in each scenario and the distribution network attribute information of the distribution network. The flow constraint sub-condition is determined by subtracting the distributed photovoltaic active output from the active power consumed by the corresponding load to zero, and subtracting the distributed photovoltaic reactive output from the reactive power consumed by the corresponding load to zero.

[0057] For example, the power flow constraint sub-condition can be determined by the following formula:

[0058]

[0059] Where: They are the distributed photovoltaic active output injected by node i and the active power consumed by the load in the system under scenario s respectively; are the distributed photovoltaic reactive power output of node i and the reactive power consumed by loads in the system under scenario s; G ij , B ij ,δ ij are the conductance, susceptance and voltage phase difference between nodes i and j respectively; N is the total number of system nodes; U i,s is the voltage of node i in scenario s; PV Represents the set of all nodes connected to distributed photovoltaics.

[0060] S203: Determine the indicator constraint sub-conditions according to the relevant indicators of each distribution network and the preset standard information table.

[0061] In this embodiment, the distribution network related indicators can be understood as indicators that ensure the safe and stable operation of the distribution network. The preset standard information table can be understood as an information table containing the maximum constraint standards of each distribution network related indicator. The indicator constraint sub-condition can be understood as a sub-condition formed by constraining the maximum value of the indicator.

[0062] Specifically, the processor may construct the indicator constraint sub-condition corresponding to each distribution network related indicator through the distribution network related indicators and the standard range set in the preset standard information table.

[0063] Exemplarily, the distribution network related indicators may include voltage deviation, voltage fluctuation, branch power flow, distributed photovoltaic output, line thermal capacity and short-circuit capacity, etc. The standard range corresponding to each distribution network related indicator can be determined by searching in the preset standard information table to obtain the indicator constraint sub-condition as shown in the following formula:

[0064] Voltage deviation constraint: The processor can obtain relevant provisions in national standards to constrain voltage deviation. The system node constraints meet the following constraints:

[0065]

[0066] Where: U jmin , U jmax is the minimum and maximum operating voltage values ​​allowed for node j; n Represents a collection of distribution network nodes.

[0067] Voltage fluctuation constraints: The processor can obtain relevant provisions in national standards to constrain voltage fluctuations. The system voltage fluctuation constraints meet the following constraints:

[0068] d k %≤d max %

[0069] Where: d max％ It is the maximum voltage fluctuation value specified by the national standard. k％ It refers to the ratio of the difference between the maximum and minimum effective values ​​of voltage in the power system to the percentage of rated voltage.

[0070] Branch flow constraints: With the integration of distributed photovoltaics into the power distribution network, the distribution network changes from a simple radial network to a multi-terminal active network. The magnitude and direction of the current in each branch of the distribution network will change, which will cause changes in branch capacity. The system branch flow constraint processor can be defined as follows:

[0071]

[0072] Where: S l,s is the apparent power passing through branch l in scenario s; S lmaxis the transmission capacity limit of branch l; Φ b Represents the set of branches l.

[0073] Distributed photovoltaic output constraints: Due to the limitations of their own equipment conditions, the maximum output power of distributed photovoltaic power sources is often limited. The output processor of the distributed photovoltaic system can determine the constraints as follows:

[0074]

[0075] Where: P PVimax It is the upper limit of the active output of distributed photovoltaic power source i.

[0076] Line thermal capacity constraints: When the total capacity of photovoltaic access gradually increases, the distributed photovoltaic output will be higher than the power load, causing the line flow distribution to face the risk of overload, thus affecting the thermal stability limit of the line. The specific constraints for the branch current thermal effect processor can be determined as follows:

[0077]

[0078] Where: I l,s represents the current value of branch l in scenario s; I lmax is the maximum current value that branch l can transmit.

[0079] Short-circuit capacity constraint: The maximum short-circuit current of the distribution network directly affects the short-circuit capacity. The processor can directly express the short-circuit constraint using the maximum short-circuit current and adopt the relevant provisions in the national standard. The specific constraints are as follows:

[0080] I SC ≤I SCmax

[0081] Where: I SC is the system short-circuit current; I SCmax It is the maximum short-circuit current value specified by the national standard.

[0082] S204. Taking each index constraint sub-condition and the flow constraint sub-condition as constraint conditions in the absorption capacity optimization model.

[0083] Specifically, the processor may use the index constraint sub-conditions and the flow constraint sub-conditions together as constraint conditions in the absorption capacity optimization model.

[0084] Exemplarily, the optimization model of photovoltaic absorption capacity of distribution network is established as follows:

[0085]

[0086] S205 . Based on a preset power restriction probability, convert the power flow constraint sub-condition into a probability constraint sub-condition.

[0087] In this embodiment, the preset power-cut probability can be understood as the probability of a predetermined power-cut event occurring. The probability constraint sub-condition can be understood as a sub-condition constrained by the power-cut probability.

[0088] Specifically, the processor may adopt a chance constraint to limit the probability of the system power flow variable exceeding the safety limit by presetting the power curtailment probability.

[0089] Exemplarily, the probability constraint condition can be expressed by the following equation:

[0090]

[0091] Where δ is the preset power curtailment probability. In the constraint equation of probabilistic power flow, Pr is usually used to express chance constraints, that is, the probability of an uncertain event meeting specific conditions should be higher than a preset threshold value (confidence level). (In the probabilistic power flow analysis of power systems, Pr generally represents the probability of an event occurring. Specifically, it can refer to the probability that system power flow variables (such as node voltage, line power flow, etc.) exceed the predetermined safety limit under certain conditions. This representation method enables power system planning and operation analysis to take into account uncertain factors, such as the random output and load fluctuations of renewable energy, thereby improving the reliability and economy of power grid operation.) The above formula represents that the probability of exceeding the limit of the power flow constraint equation does not exceed a given power curtailment probability [(1-δ)×100]%. Generally speaking, the higher the power curtailment probability, the more distributed power sources can be connected to the power grid, but at the same time it will also increase the risk of distributed power curtailment.

[0092] S206. According to the preset scenario indicator variables and preset values, the probability constraint sub-conditions are rewritten into opportunity constraint planning sub-conditions to obtain an intermediate absorption capacity optimization model.

[0093] In this embodiment, the preset scenario indicator variable can be understood as a variable used to indicate whether to eliminate the scenario, for example, 1 for elimination and 0 for retention. The preset value can be understood as a larger number set for using the large number method. The intermediate absorption capacity optimization model can be understood as an optimization model containing the chance-constrained programming sub-condition.

[0094] Specifically, since the obtained probability constraint sub-condition expression involves the probability distribution of photovoltaic output and the non-convexity of the entire model, it is difficult to solve directly. The processor can discretize the probability distribution and introduce an auxiliary binary variable, namely the preset scenario indicator variable, to distinguish whether the opportunity constraint satisfies the discrete scenario. The original continuous probability constraint is converted into a discrete 0-1 integer programming problem by using the large number algorithm (Big-M) through the preset numerical value, and the opportunity constraint planning sub-condition is obtained. Then, the intermediate absorption capacity optimization model is determined through the opportunity constraint planning sub-condition, the indicator constraint sub-condition and the objective function. This conversion greatly simplifies the state space of the problem, making it more suitable for processing by deep learning algorithms.

[0095] For example, the rewritten chance constraint programming sub-condition can be expressed as:

[0096]

[0097] Where: M is the preset value; w s is the preset scene indicator variable, indicating whether to remove the indicator variable of the scene. If w s =1, then the corresponding scenario s does not satisfy the chance constraint, w s = 0 means that the scenario s is included in the model and satisfies the opportunity constraint. The total number of scenarios that satisfy the opportunity constraint is controlled by the above formula, which represents the risk constraint, where π s represents the probability of scenario s occurring. It can be seen that the above two methods for solving chance-constrained programming are equivalent. In particular, when the risk level δ = 0, w s are all 0. At this time, the chance constrained programming model degenerates into a random programming model, that is, all scenarios satisfy the constraints.

[0098] That is, the optimization model of intermediate consumption capacity is:

[0099]

[0100]

[0101] S207. Determine the photovoltaic absorption decision-making intelligent entity based on the intermediate absorption capacity optimization model.

[0102] Specifically, the processor can extract elements for constructing a photovoltaic absorption decision-making intelligent agent and performing a decision-making process from the opportunity constraint planning sub-conditions and the index constraint sub-conditions of the intermediate absorption capacity optimization model through a Markov decision process algorithm.

[0103] Further, based on the above embodiment, the steps of determining the photovoltaic consumption decision-making agent according to the intermediate consumption capacity optimization model can be refined as follows:

[0104] Construct an initial photovoltaic absorption decision-making agent and an initial environment; determine the decision-making process elements used to trigger the initial photovoltaic absorption decision-making agent to make action decisions based on the intermediate absorption capacity optimization model and the initial environment; configure the initial photovoltaic absorption decision-making agent based on the decision-making process elements and the preset decision-making process to obtain the photovoltaic absorption decision-making agent.

[0105] In this embodiment, the initial photovoltaic consumption decision-making agent can be understood as an undefined agent. The initial environment can be understood as an undefined environment, and the feedback is determined by the environment and the corresponding elements. The decision process element can be understood as an influencing element used to indicate the decision process of the initial photovoltaic consumption decision-making agent and the initial environment. The preset decision process can be understood as the process corresponding to the photovoltaic consumption agent and the environment to complete the decision. The Markov decision process uses a state to represent the changes in the environment. The photovoltaic consumption decision-making agent selects an action based on the state value. The environment will feedback the reward value based on the selected action. At the same time, based on the selected action, the environment will also change accordingly, generating a new state value and sending it to the photovoltaic consumption agent.

[0106] Specifically, the processor can first set a neural network or algorithm according to the needs to build an initial photovoltaic consumption decision-making agent and an initial environment. The processor can extract elements that trigger the action of the photovoltaic consumption decision-making agent from the intermediate consumption capacity optimization model, as well as elements that trigger initial environmental changes and feedback, and together constitute the decision-making process elements used to trigger the initial photovoltaic consumption decision-making agent to make action decisions. The processor can configure the decision-making process of the initial photovoltaic consumption decision-making agent according to the decision-making process elements and the preset decision-making process to obtain the photovoltaic consumption decision-making agent.

[0107] Among them, based on the above embodiment, the steps of determining the decision process elements for triggering the initial photovoltaic consumption decision-making agent to make action decisions according to the intermediate consumption capacity optimization model can be further refined as follows:

[0108] Determine the state parameters included in the intermediate absorption capacity optimization model, and construct the state space of the state of the initial environment; determine the variable adjustment actions included in the intermediate absorption capacity optimization model, and construct the action space indicating all actions taken by the initial photovoltaic absorption decision-making agent; use the objective function as the reward function; according to the state vectors and rewards at different times, construct the transition probability function between the new and old states of the initial photovoltaic absorption decision-making agent; use the state space, action space, reward function, state transfer function and preset attenuation function as decision-making process elements that trigger the initial photovoltaic absorption decision-making agent to make action decisions.

[0109] In this embodiment, the state parameters can be understood as parameters that reflect the state of the environment. The state space can be understood as a set of all possible states. The variable adjustment action can be understood as a variable in the intermediate absorption capacity optimization model. The action space can be understood as a set of all possible actions that the photovoltaic absorption decision-making agent can take in a given state. The reward function can be understood as the feedback obtained by the photovoltaic absorption decision-making agent from the environment after taking a certain action, that is, the value of the objective function. The transition probability function can be understood as a function that determines the probability of transition between the old and new states, that is, the probability of transitioning from one state to another after a certain action. The preset attenuation function can be understood as a function used to determine the importance of future rewards relative to immediate rewards.

[0110] Specifically, the processor can determine the state parameters contained in the intermediate absorption capacity optimization model, for example, it can include photovoltaic power generation, load demand, node voltage and line current, and constitute a state space that characterizes the state of the initial environment. By determining the variables contained in the intermediate absorption capacity optimization model, that is, the variable adjustment action, an action space indicating all actions taken by the initial photovoltaic absorption decision-making agent is constructed, such as the action space including the active output adjustment action and reactive power adjustment action of distributed photovoltaics. The processor can use the objective function as a reward function. The processor can construct a transition probability function between the new and old states of the initial photovoltaic absorption decision-making agent based on the state vectors and rewards at different times. By using the state space, action space, reward function, state transfer function and preset attenuation function as decision-making process elements that trigger the initial photovoltaic absorption decision-making agent to make action decisions.

[0111] For example, in order to facilitate understanding of the decision-making process of the photovoltaic consumption decision-making agent, a specific example is used to demonstrate that the decision-making process of the photovoltaic consumption decision-making agent can be determined through the Markov decision process. Figure 3 An example diagram of an intelligent agent decision-making method for constructing a distributed photovoltaic consumption capacity model is provided for the second embodiment of the present invention, such as Figure 3 As shown in the figure, it includes the PV consumption decision-making agent, referred to as the agent, and the environment that interacts with the agent. The Markov decision process consists of five elements M = {S, A, P, R, γ}, where:

[0112] S is the state space, that is, the set of all possible states. In the distributed photovoltaic consumption optimization model, the state space must be able to accurately reflect the state of the environment, including photovoltaic power generation, load demand, node voltage and line current, as shown in the following formula:

[0113]

[0114] In the formula, S t is the state vector at time t, denoted as s; are the distributed photovoltaic active output and reactive power injected into node i at time t respectively; are the active power demand and reactive power demand of the load at node i at time t, respectively; represents the voltage amplitude of node i at time t; represents the current in the branch at time t.

[0115] A is the action space, which is the set of all possible actions that the agent can take in a given state. In reinforcement learning, the action space is generally designed as a variable in the model, including the active output and reactive power of distributed photovoltaics, as shown in the following formula:

[0116]

[0117] In the formula, A t is the action vector at time t, denoted as a; They are the strategic actions of the intelligent agent’s decision on the distributed photovoltaic active output and reactive power adjustment action injected into node i at time t.

[0118] P represents the transition probability function, which refers to the probability of transferring from one state to another state through a certain action, that is, the transition probability between the new and old states of the agent:

[0119]

[0120] In the formula, S t+1 represents the state vector at time t+1, s' represents the next state; R t+1 represents the reward obtained at time t+1, and p represents the probability function.

[0121] R is the reward function, that is, the feedback obtained by the agent from the environment after taking a certain action. In order to achieve the target strategy of maximizing the actual photovoltaic absorption capacity of the system, the objective function is designed as the reward function:

[0122]

[0123] Where r is the reward function at a certain moment, λ is a weight factor that balances power generation and system losses, and w is the penalty term for violating the constraint.

[0124] γ is a decay function that determines the importance of future rewards relative to immediate rewards. The value of γ is between 0 and 1, where 0 means only caring about immediate rewards and 1 means that immediate rewards are equally important as future rewards.

[0125] S208. Based on the photovoltaic consumption decision-making agent, initialize the policy network of the double-delay deep deterministic policy gradient algorithm and the corresponding policy target network.

[0126] In this embodiment, the policy network can be understood as a network for generating actions, and the parameters of the policy network are updated through interaction with the environment. Specifically, the reward signal obtained from the environment is used to guide the optimization of the policy network to maximize long-term returns. The policy target network can be understood as a network used to assist the policy network in collaborative learning. Its network structure is consistent with the policy network, but the parameters are different.

[0127] Specifically, the processor can use the photovoltaic consumption decision-making agent as the policy network of the double-delay deep deterministic strategy and the corresponding policy target network.

[0128] S209: Initialize two preset evaluation networks and two corresponding evaluation target networks.

[0129] In this embodiment, the preset evaluation network can be understood as a network for evaluating the error of the strategy network. The evaluation target network can be understood as a network for evaluating the error of the strategy target network, and its structure is consistent with the preset evaluation network, but the parameters are different.

[0130] Specifically, two preset evaluation networks are usually set in the double-delayed deep deterministic policy gradient algorithm to estimate the state value function, that is, given the state and action, predict the expected value of future rewards. In order to improve the accuracy of the estimation, two independent evaluation networks are used, and the processor can initialize the two preset evaluation networks and the corresponding two evaluation target networks.

[0131] S210. According to the output of the photovoltaic consumption decision-making intelligent agent on the current state variables, sample data is determined and stored in a preset experience pool.

[0132] In this embodiment, the current state variable can be understood as the current state variable that triggers the action change of the photovoltaic consumption decision-making agent. The sample data can be understood as data including the current state variable and the action vector, reward function and next state variable output by the photovoltaic consumption agent. The preset experience pool is used to store the sample data.

[0133] Specifically, the processor can input the current state variables provided by the strategy network into the photovoltaic consumption decision-making agent to obtain the output of the photovoltaic consumption decision-making agent on the current state variables, and package the current state variables, action vectors, reward functions and next state variables as sample data and store them in a preset experience pool.

[0134] S211, according to the preset experience pool, each preset evaluation network, each evaluation target network and the strategy target network, the weight parameters of the strategy network are updated to obtain an updated strategy network.

[0135] In this embodiment, the weight parameter can be understood as a parameter used to improve the performance of the policy network.

[0136] Specifically, the processor can input sample data in the preset experience pool into the policy network to obtain the current state and reward, and use two evaluation networks to estimate the current state and reward respectively to obtain two estimated values; input multiple sample data into the policy target network to obtain the future state and reward, and estimate the future state and reward through two evaluation target networks to obtain two target estimated values, and calculate two errors through the two estimated values ​​and the two target estimated values, and when the error does not reach the set standard, update the weight parameters of the two evaluation networks and the policy network to obtain the updated policy network. Then, the weight parameters of the evaluation target network are soft-updated through the weight parameters of the evaluation network, and the weight parameters of the policy target network are soft-updated through the weight parameters of the policy network.

[0137] S212: Determine whether the training end condition is met. If so, determine the final distributed photovoltaic absorption capacity model based on the updated strategy network.

[0138] In this embodiment, the training end condition can be understood as a condition for determining whether a final model can be output, for example, it can be a set maximum number of training iterations or a standard error value.

[0139] Specifically, the processor can determine whether the error meets the requirement by comparing the determined error value with the set standard error value, or the processor can accumulate the number of training iterations until the preset number of training iterations is reached and the training end condition is met. The processor can arrange the updated strategy network in the intermediate absorption capacity optimization model for solution to obtain the final distributed photovoltaic absorption capacity model.

[0140] S213: If not, obtain the next state variable and return to the sample data determination step.

[0141] In this embodiment, the next state variable can be understood as the state variable corresponding to the next training, which can be obtained through the interaction of the environment with the photovoltaic consumption decision-making agent.

[0142] Specifically, if the training end condition is not met, the processor continues to obtain the next state variable, returns to the step of determining the sample data and storing it in a preset experience pool based on the output of the next state variable by the photovoltaic consumption decision-making agent.

[0143] Further, based on the above embodiment, the weight parameters of the policy target network can be soft updated according to the preset experience pool, each preset evaluation network, each evaluation target network and the policy network to obtain the updated policy target network step as follows:

[0144] Extract target sample data from the preset experience pool, determine the current information to be judged in the target sample data through the policy network, the current information to be judged includes the current state vector and the current action vector, determine the future information to be judged through the policy target network, the future information to be judged includes the future state vector and the future action vector; determine the estimated value of each preset judgment network according to the current information to be judged; determine the target estimated value according to the two judgment target networks and the future information to be judged; determine the loss function of each preset judgment network according to the target estimated value and each estimated value; perform the network weights of the policy network according to each loss function and the weighted gradient algorithm to obtain an updated policy target network.

[0145] In this embodiment, the target sample data can be understood as the sample data to be judged at present. The current information to be judged can be understood as the information used for the judgment network to make an estimate, which may include the current state vector and the current action vector. The future state information to be judged can be understood as the information used for the judgment target network to make an estimate, which may include the future state vector and the future action vector. The estimated value can be understood as the estimated value of the state-action value function calculated by each preset judgment network, which is used to evaluate the expected return of taking an action under a specific state. The target estimated value can be understood as the target estimated value formed by the average value of the estimated value of the state-action value function calculated by the two judgment target networks. The loss function can be understood as a function for calculating the error of the estimated value. The weighted gradient algorithm can be understood as an optimization method for updating the parameters by calculating the gradient of the loss function to the model parameters and combining the weight information.

[0146] Specifically, the processor can extract target sample data from the preset experience pool, determine the current information to be judged in the target sample data through the policy network, and the current information to be judged includes the current state vector and the current action vector. The processor can extract multiple target sample data from the preset experience pool, and determine the future information to be judged through the policy target network, and the future information to be judged includes the future state vector and the future action vector. The processor can input the current information to be judged into two preset judgment networks respectively to obtain the estimated value of each preset judgment network. The processor can input the future information to be judged into two judgment target networks respectively, determine the estimated value of each judgment target network and calculate the average value to obtain the target estimated value. The processor can determine the loss function of each preset judgment network based on the target estimated value, each estimated value and the number of samples. The processor can perform weighted gradient algorithm on the network weights of the policy network according to each loss function to obtain an updated policy target network.

[0147] For example, in order to facilitate understanding of the architecture of the double-delayed deep deterministic policy gradient algorithm for distributed photovoltaic consumption optimization, a specific example is used to demonstrate: Figure 4An example diagram of an algorithm architecture of a method for constructing a distributed photovoltaic consumption capacity model is provided for the second embodiment of the present invention, such as Figure 4 As shown in the figure, it includes a policy network and a corresponding policy target network, two evaluation networks and corresponding evaluation target networks. The specific steps are as follows:

[0148] a. The processor can initialize a policy network and two evaluation networks, and their corresponding target networks;

[0149] b. The processor can use the interaction between the PV consumption decision-making agent and the environment as distributed PV consumption, take actions based on the current state variables through the strategy network, and observe the new state and reward. At the same time, the obtained sample data {s, a, r, s'} is stored in the experience pool;

[0150] c. Randomly extract N groups of sample data from the experience pool and use two judgment networks to estimate the value Q respectively:

[0151]

[0152] In the formula, E is the expected value; R r+k+1 is the reward obtained at time t+k+1; T is the number of time steps;

[0153] e. Use the average of the Q values ​​calculated by the two target evaluation networks to form the target estimate Q, which not only suppresses the problem of overestimation of the network, but also does not make it too conservative:

[0154]

[0155] In the formula, is the i-th judgment target network for the next state s' and action The valuation of ε is random noise;

[0156] f. Use the calculated target Q value Q target Calculate the error and update the judgment network. The loss function is as follows:

[0157]

[0158] Where N is the number of samples;

[0159] g. Use the updated critic network to update the weights of the policy network with weighted delay to maximize Q(s,a);

[0160] h. Use the soft update strategy to regularly update the weights of the target network. The formula for the soft update strategy is:

[0161] θ target =τθ+(1-τ)θ target

[0162] Where τ is a small positive number, which is used to control the speed of updating the main network parameters (main network includes policy network, evaluation network 1 and evaluation network 2) to the target network parameters (target network includes policy target network, evaluation target network 1 and evaluation target network 2); θ is the main network weight parameter; θ target is the target network weight parameter;

[0163] Repeat steps b to g until the agent’s performance no longer improves by L or the preset number of training iterations is reached.

[0164] The technical solution of the embodiment of the present invention proposes a series of index constraints such as voltage quality, short-circuit capacity, and line flow constraints in random scenarios, and comprehensively considers the balance between the maximization of photovoltaic grid-connected capacity and the active network loss of the system. By defining the difference between the photovoltaic absorption capacity and the active loss of the entire system as the actual photovoltaic absorption capacity, the maximum actual photovoltaic absorption capacity is used as the objective function, and a multi-constraint distributed photovoltaic absorption capacity optimization model based on scenario random planning is established, which accurately reflects the impact of photovoltaic grid connection on the actual operation effect of the system, and can quickly determine the maximum boundary capacity value of multiple photovoltaic grid connections at the same time, providing strong support for the planning and operation of the power system. The opportunity constraint planning method is adopted to transform the traditional deterministic power balance constraint into a probabilistic constraint, and an intermediate absorption capacity optimization model based on opportunity constraint planning is constructed, which not only improves the accuracy of absorption capacity evaluation, but also further optimizes the economy of distributed photovoltaic absorption, laying a solid foundation for the intelligent and green development of distribution networks. By transforming the intermediate absorption capacity optimization model into the action decision process of the photovoltaic absorption decision-making agent, for the continuous space action problem of distributed photovoltaic absorption capacity optimization, a double-delayed deep deterministic policy gradient algorithm based on the policy-judgment framework is used. Two sets of judgment networks are used to calculate the average value when estimating the target value, which can suppress the problem of overestimation or over-conservatism of the network, and add noise disturbance to the next state to achieve smooth regularization of the target strategy and obtain more accurate value estimation. At the same time, delayed updates are adopted, and the update frequency of the judgment network is higher than that of the policy network, making the network training more stable. This algorithm provides an efficient, stable and adaptive solution for the distributed photovoltaic absorption optimization model. The accuracy of the final model is improved.

[0165] Embodiment 3

[0166] Figure 5 This is a schematic diagram of a structure of a device for constructing a distributed photovoltaic consumption capacity model provided by the third embodiment of the present invention. Figure 5 As shown, the device includes: a function determination module 51, a condition determination module 52, an agent determination module 53 and a model determination module 54.

[0167] The function determination module 51 is used to determine the objective function of the absorption capacity optimization model according to the maximum absorption capacity of the photovoltaic power generation system of the distribution network and the active power loss of the system;

[0168] A condition determination module 52, used to determine the constraint conditions of the absorption capacity optimization model according to the system power flow distribution related parameters of the photovoltaic power generation system of the distribution network;

[0169] An intelligent agent determination module 53 is used to determine a photovoltaic consumption decision-making intelligent agent according to the consumption capacity optimization model;

[0170] The model determination module 54 is used to determine the final distributed photovoltaic absorption capacity model according to the photovoltaic absorption decision-making agent and the double-delay deep deterministic policy gradient algorithm.

[0171] The technical solution of the embodiment of the present invention determines the objective function of the absorption capacity optimization model according to the maximum absorption capacity of photovoltaic power generation system of distribution network and the system active power loss; determines the constraint conditions of the absorption capacity optimization model according to the system flow distribution related parameters of photovoltaic power generation system of distribution network; determines the photovoltaic absorption decision-making agent according to the absorption capacity optimization model; determines the final distributed photovoltaic absorption capacity model according to the photovoltaic absorption decision-making agent and the double-delay deep deterministic policy gradient algorithm. By setting multiple constraints, the maximum boundary capacity value of multiple photovoltaics connected to the grid at the same time can be quickly determined, reducing the calculation time of the model. Through the double-delay deep deterministic policy gradient algorithm, the processing of continuous action space problems and fine-tuning actions are realized, the algorithm training stability is improved, the problems of over-estimation and over-conservatism are reduced, and it can adapt to dynamic environments, thereby improving the accuracy of the final model and providing strong support for the planning and operation of the power system.

[0172] Furthermore, the condition determination module 52 is specifically used for:

[0173] Determine the power flow constraint sub-condition according to the distributed photovoltaic output injected into each node in each scenario, the power consumed by the load, the node attribute information of each node and the distribution network attribute information of the distribution network;

[0174] Determine the indicator constraint sub-conditions according to the relevant indicators of each distribution network and the preset standard information table;

[0175] Each of the index constraint sub-conditions and the flow constraint sub-conditions is used as a constraint condition in the absorption capacity optimization model.

[0176] Furthermore, the agent determination module 53 includes:

[0177] a sub-condition determination unit, configured to convert the power flow constraint sub-condition into a probability constraint sub-condition based on a preset power restriction probability;

[0178] A model determination unit, used to rewrite the probability constraint sub-condition into the opportunity constraint planning sub-condition according to the preset scenario indicator variable and the preset value, so as to obtain the intermediate consumption capacity optimization model;

[0179] The intelligent agent determination unit is used to determine the photovoltaic absorption decision-making intelligent agent according to the intermediate absorption capacity optimization model.

[0180] Wherein, the agent determination unit includes:

[0181] Construct subunits to construct the initial photovoltaic consumption decision-making intelligent agent and the initial environment;

[0182] An element determination subunit, used to determine a decision process element for triggering the initial photovoltaic absorption decision-making agent to make an action decision according to the intermediate absorption capacity optimization model and the initial environment;

[0183] The intelligent agent determination subunit is used to configure the initial photovoltaic consumption decision-making intelligent agent according to the decision-making process elements and the preset decision-making process to obtain the photovoltaic consumption decision-making intelligent agent.

[0184] Among them, the element determination subunit is specifically used for:

[0185] Determine the state parameters included in the intermediate absorption capacity optimization model, and construct the state space of the state of the initial environment;

[0186] Determining the variable adjustment actions included in the intermediate absorption capacity optimization model, and constructing an action space indicating all actions taken by the initial photovoltaic absorption decision-making agent;

[0187] Using the objective function as a reward function;

[0188] According to the state vectors and rewards at different times, a transition probability function between the new and old states of the initial photovoltaic consumption decision-making agent is constructed;

[0189] The state space, action space, reward function, state transfer function and preset attenuation function are used as decision-making process elements to trigger the initial photovoltaic consumption decision-making agent to make action decisions.

[0190] Furthermore, the model determination module 54 includes:

[0191] A first determination unit is used to initialize a policy network and a corresponding policy target network of a double-delayed deep deterministic policy gradient algorithm based on the photovoltaic consumption decision-making agent;

[0192] A second determination unit, used to initialize two preset evaluation networks and two corresponding evaluation target networks;

[0193] A third determination unit is used to determine sample data and store it in a preset experience pool according to the output of the photovoltaic consumption decision-making intelligent agent on the current state variable;

[0194] A fourth determining unit, configured to update the weight parameters of the policy network according to the preset experience pool, each of the preset evaluation networks, each of the evaluation target networks and the policy target network to obtain an updated policy network;

[0195] A fifth determination unit, configured to determine whether a training end condition is met, and if so, determine a final distributed photovoltaic consumption capacity model based on the updated strategy network;

[0196] The sixth determination unit is used to obtain the next state variable if no, and return to the determination step of the sample data.

[0197] The fourth determining unit is specifically used for:

[0198] Extracting target sample data from the preset experience pool, determining current information to be judged in the target sample data through the strategy network, wherein the current information to be judged includes a current state vector and a current action vector, and determining future information to be judged through the strategy target network, wherein the future information to be judged includes a future state vector and a future action vector;

[0199] Determining the estimated value of each of the preset evaluation networks according to the current information to be evaluated;

[0200] Determining a target estimation value according to the two evaluation target networks and the future information to be evaluated;

[0201] Determining the loss function of each of the preset evaluation networks according to the target estimated value and each of the estimated values;

[0202] According to each of the loss functions and the weighted gradient algorithm, the network weights of the policy network are adjusted to obtain an updated policy target network.

[0203] The apparatus for constructing a distributed photovoltaic accommodation capacity model provided in an embodiment of the present invention can execute the method for constructing a distributed photovoltaic accommodation capacity model provided in any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

[0204] Embodiment 4

[0205] Figure 6A schematic diagram of the structure of an electronic device 60 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0206] like Figure 6 As shown, the electronic device 60 includes at least one processor 61, and a memory connected to the at least one processor 61, such as a read-only memory (ROM) 62, a random access memory (RAM) 63, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 61 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 62 or the computer program loaded from the storage unit 68 to the random access memory (RAM) 63. In the RAM 63, various programs and data required for the operation of the electronic device 60 can also be stored. The processor 61, the ROM 62, and the RAM 63 are connected to each other via a bus 64. An input / output (I / O) interface 65 is also connected to the bus 64.

[0207] A number of components in the electronic device 60 are connected to the I / O interface 65, including: an input unit 66, such as a keyboard, a mouse, etc.; an output unit 67, such as various types of displays, speakers, etc.; a storage unit 68, such as a disk, an optical disk, etc.; and a communication unit 69, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 69 allows the electronic device 60 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0208] The processor 61 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 61 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The processor 61 executes the various methods and processes described above, such as a method for constructing a distributed photovoltaic consumption capacity model.

[0209] In some embodiments, the method for constructing a distributed photovoltaic accommodation capacity model may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 68. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 60 via the ROM 62 and / or the communication unit 69. When the computer program is loaded into the RAM 63 and executed by the processor 61, one or more steps of the method for constructing a distributed photovoltaic accommodation capacity model described above may be performed. Alternatively, in other embodiments, the processor 61 may be configured to execute the method for constructing a distributed photovoltaic accommodation capacity model in any other appropriate manner (e.g., by means of firmware).

[0210] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0211] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0212] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0213] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0214] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0215] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0216] In one embodiment, the embodiment of the present invention also includes a computer program product, the computer program product includes a computer program, and when the computer program is executed by a processor, it implements the method for constructing a distributed photovoltaic absorption capacity model of any embodiment of the present invention.

[0217] In the process of implementation, the computer program product can be written in one or more programming languages ​​or a combination thereof to perform the computer program code of the present invention, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet).

[0218] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0219] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for constructing a distributed photovoltaic consumption capacity model, characterized in that: include: According to the maximum photovoltaic absorption capacity of the photovoltaic power generation system of the distribution network and the active power loss of the system, the objective function of the absorption capacity optimization model is determined; Determining the constraint conditions of the absorption capacity optimization model according to the system power flow distribution related parameters of the photovoltaic power generation system of the distribution network; Determine a photovoltaic consumption decision-making intelligent agent according to the consumption capacity optimization model; According to the photovoltaic absorption decision-making intelligent agent and the double-delay deep deterministic policy gradient algorithm, the final distributed photovoltaic absorption capacity model is determined.

2. The method according to claim 1, characterized in that: Determining the constraint conditions of the absorption capacity optimization model according to the system power flow distribution related parameters of the photovoltaic power generation system of the distribution network includes: Determine the power flow constraint sub-condition according to the distributed photovoltaic output injected into each node in each scenario, the power consumed by the load, the node attribute information of each node and the distribution network attribute information of the distribution network; Determine the indicator constraint sub-conditions according to the relevant indicators of each distribution network and the preset standard information table; Each of the index constraint sub-conditions and the flow constraint sub-conditions is used as a constraint condition in the absorption capacity optimization model.

3. The method according to claim 2, characterized in that Determining the photovoltaic consumption decision-making intelligent agent according to the consumption capacity optimization model includes: Based on a preset power curtailment probability, converting the power flow constraint sub-condition into a probability constraint sub-condition; According to the preset scenario indicator variables and preset values, the probability constraint sub-conditions are rewritten into opportunity constraint planning sub-conditions to obtain an intermediate consumption capacity optimization model; According to the intermediate absorption capacity optimization model, the photovoltaic absorption decision-making intelligent agent is determined.

4. The method according to claim 3, characterized in that Determining the photovoltaic consumption decision-making intelligent agent according to the intermediate consumption capacity optimization model includes: Construct the initial photovoltaic consumption decision-making intelligent agent and initial environment; Determining, according to the intermediate absorption capacity optimization model and the initial environment, a decision process element for triggering the initial photovoltaic absorption decision-making agent to make an action decision; According to the decision-making process elements and the preset decision-making process, the initial photovoltaic consumption decision-making intelligent agent is configured to obtain the photovoltaic consumption decision-making intelligent agent.

5. The method according to claim 4, characterized in that The step of determining, according to the intermediate absorption capacity optimization model, a decision process element for triggering the initial photovoltaic absorption decision-making agent to make an action decision comprises: Determine the state parameters included in the intermediate absorption capacity optimization model, and construct the state space of the state of the initial environment; Determining the variable adjustment actions included in the intermediate absorption capacity optimization model, and constructing an action space indicating all actions taken by the initial photovoltaic absorption decision-making agent; Using the objective function as a reward function; According to the state vectors and rewards at different times, a transition probability function between the new and old states of the initial photovoltaic consumption decision-making agent is constructed; The state space, action space, reward function, state transfer function and preset attenuation function are used as decision-making process elements to trigger the initial photovoltaic consumption decision-making agent to make action decisions.

6. The method according to claim 1, characterized in that The method of determining the final distributed photovoltaic consumption capacity model according to the photovoltaic consumption decision-making agent and the double-delayed deep deterministic policy gradient algorithm includes: Based on the photovoltaic consumption decision-making agent, the policy network and the corresponding policy target network of the double-delay deep deterministic policy gradient algorithm are initialized; Initialize two preset evaluation networks and two corresponding evaluation target networks; According to the output of the photovoltaic consumption decision-making intelligent agent on the current state variable, sample data is determined and stored in a preset experience pool; According to the preset experience pool, each of the preset evaluation networks, each of the evaluation target networks and the strategy target network, the weight parameters of the strategy network are updated to obtain an updated strategy network; Determine whether the training end condition is met, and if so, determine the final distributed photovoltaic consumption capacity model based on the updated strategy network; If not, the next state variable is obtained and the process returns to the step of determining the sample data.

7. The method according to claim 6, characterized in that The step of soft-updating the weight parameters of the policy target network according to the preset experience pool, each of the preset evaluation networks, each of the evaluation target networks and the policy network to obtain an updated policy target network includes: Extracting target sample data from the preset experience pool, determining current information to be judged in the target sample data through the strategy network, wherein the current information to be judged includes a current state vector and a current action vector, and determining future information to be judged through the strategy target network, wherein the future information to be judged includes a future state vector and a future action vector; Determining the estimated value of each of the preset evaluation networks according to the current information to be evaluated; Determining a target estimation value according to the two evaluation target networks and the future information to be evaluated; Determining the loss function of each of the preset evaluation networks according to the target estimated value and each of the estimated values; According to each of the loss functions and the weighted gradient algorithm, the network weights of the policy network are adjusted to obtain an updated policy target network.

8. A device for constructing a distributed photovoltaic consumption capacity model, characterized in that: include: A function determination module is used to determine the objective function of the absorption capacity optimization model according to the maximum absorption capacity of the photovoltaic power generation system of the distribution network and the active power loss of the system; A condition determination module, used to determine the constraint conditions of the absorption capacity optimization model according to the system power flow distribution related parameters of the photovoltaic power generation system of the distribution network; An intelligent agent determination module is used to determine a photovoltaic consumption decision-making intelligent agent according to the consumption capacity optimization model; The model determination module is used to determine the final distributed photovoltaic absorption capacity model according to the photovoltaic absorption decision-making agent and the double-delay deep deterministic policy gradient algorithm.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the method for constructing a distributed photovoltaic absorption capacity model according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the method for constructing a distributed photovoltaic absorption capacity model according to any one of claims 1 to 7 when executed.

Images