Distributed photovoltaic cluster coordination control method, device, equipment and storage medium

Abstract

The invention belongs to the field of power distribution networks, and particularly relates to a distributed photovoltaic cluster coordination control method, device and equipment and a storage medium, and the method comprises the steps: obtaining the photovoltaic state data of each node in a photovoltaic cluster, and inputting the photovoltaic state data into a pre-established enhanced containment control model; a pinning node is selected from all the nodes, and an optimal action value is selected for realizing pinning control for a pinning consistency algorithm through a greedy control strategy of a reinforcement learning algorithm adopted by the reinforcement pinning control model; and executing greedy action on the contained nodes based on a contained consistency algorithm, and taking the optimal action value as a control instruction to coordinate each node of the photovoltaic cluster. Through the reinforcement learning and containment control method, the power correction can be quickly and accurately generated, and the inter-group interaction power is optimized, so that the system frequency and the bus voltage are stabilized near rated values, and the reactive elastic support capability of the distributed photovoltaic cluster in the peer-to-peer control mode is improved.

Description

Technical Field

[0001] The present invention belongs to the field of power distribution networks, relates to a voltage prediction method, and in particular to a distributed photovoltaic cluster coordinated control method, device, equipment and storage medium. Background Art

[0002] Distributed photovoltaics have the technical advantages of low pollution, outstanding environmental benefits and the ability to alleviate local electricity shortages. However, due to the intermittent and uncertain output of distributed photovoltaics, distribution networks are prone to power flow reverse transmission and voltage over-limit problems, which in turn increases the difficulty of grid regulation.

[0003] In the related art, the voltage control method of the distribution network containing distributed photovoltaics using the traditional droop control method can keep the voltage fluctuating within the specified range, but the specified fluctuation range is large and easily affects the safe operation of the power grid. In addition, the traditional droop control method requires the incorporation of loads, which will cause the voltage fluctuation to exceed the specified range, thereby causing voltage instability. Summary of the invention

[0004] In view of this, the present invention discloses a distributed photovoltaic cluster coordinated control method, device, equipment and storage medium, which can solve the deficiencies existing in the related art.

[0005] To achieve the above purpose, the present invention discloses the following technical solutions:

[0006] According to a first aspect of the present invention, a distributed photovoltaic cluster coordinated control method is proposed, the method comprising:

[0007] Acquire photovoltaic status data of each node in the photovoltaic cluster, and input the photovoltaic status data into a pre-established enhanced control model; wherein the photovoltaic status data includes voltage data and frequency data of each node, and the enhanced control model is established based on a reinforcement learning algorithm and a control consistency algorithm;

[0008] Selecting a pinning node from each node, and selecting an optimal action value for the pinning consistency algorithm to implement pinning control through a greedy control strategy of a reinforcement learning algorithm adopted by the reinforced pinning control model;

[0009] A greedy action is performed on the pinned node based on a pinning consistency algorithm, and the optimal action value is used as a control instruction to coordinate various nodes of the photovoltaic cluster.

[0010] According to a second aspect of the present invention, a distributed photovoltaic cluster coordination control device is proposed, the device comprising:

[0011] An acquisition unit: acquires photovoltaic status data of each node in the photovoltaic cluster, and inputs the photovoltaic status data into a pre-established enhanced control model; wherein the photovoltaic status data includes voltage data and frequency data of each node, and the enhanced control model is established based on a reinforcement learning algorithm and a control consistency algorithm;

[0012] A selection unit: selecting a pinning node from each node, and selecting an optimal action value for the pinning consistency algorithm to implement pinning control through a greedy control strategy of a reinforcement learning algorithm adopted by the reinforcement pinning control model;

[0013] The computing unit executes a greedy action on the pinned node based on a pinning consistency algorithm, and uses the optimal action value as a control instruction to coordinate the various nodes of the photovoltaic cluster.

[0014] According to a third aspect of the present invention, an electronic device is provided, comprising:

[0015] processor;

[0016] a memory for storing processor-executable instructions;

[0017] The processor implements the steps of the method described in the first aspect by running the executable instructions.

[0018] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, on which computer instructions are stored, and when the instructions are executed by a processor, the steps of the method described in the first aspect are implemented.

[0019] It can be seen from the above technical solutions that the distributed photovoltaic cluster coordination control method disclosed in the present invention establishes a reinforcement control model based on the reinforcement learning algorithm and the control consistency algorithm, and coordinates the photovoltaic cluster through the model. On the one hand, since the reinforcement control model has the advantages of both the reinforcement learning algorithm and the control consistency algorithm, it can not only realize autonomous learning under the conditions of weak communication and network structure changes, but also reduce the workload of offline training; on the other hand, when the power grid is disturbed by external factors, the reinforcement learning and control methods can quickly and accurately generate power corrections, optimize the interaction power between clusters, so that the system frequency and bus voltage are stabilized near the rated value, thereby improving the reactive elastic support capacity of the distributed photovoltaic cluster under the peer control mode. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a flow chart of a distributed photovoltaic cluster coordinated control method provided by an exemplary embodiment.

[0021] Figure 2 It is a schematic diagram of a fusion mechanism of containment and reinforcement learning provided by an exemplary embodiment.

[0022] Figure 3 It is a schematic diagram of a collaborative control based on enhanced containment provided by an exemplary embodiment.

[0023] Figure 4 It is a schematic diagram of a load-side resource regulation potential assessment provided by an exemplary embodiment.

[0024] Figure 5 It is a schematic diagram of coordinated control among distributed photovoltaic clusters provided by an exemplary embodiment.

[0025] Figure 6 It is a schematic structural diagram of a device provided by an exemplary embodiment.

[0026] Figure 7 It is a block diagram of a distributed photovoltaic cluster coordination control device provided by an exemplary embodiment. DETAILED DESCRIPTION

[0027] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with one or more embodiments of the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of one or more embodiments of the present invention as detailed in the appended claims.

[0028] It should be noted that: in other embodiments, the steps of the corresponding method are not necessarily performed in the order shown and described in the present invention. In some other embodiments, the steps included in the method may be more or less than those described in the present invention. In addition, a single step described in the present invention may be decomposed into multiple steps for description in other embodiments; and multiple steps described in the present invention may be combined into a single step for description in other embodiments.

[0029] To further illustrate the present invention, the following examples are provided:

[0030] Distributed photovoltaics have the technical advantages of low pollution, outstanding environmental benefits and the ability to alleviate local electricity shortages. However, due to the intermittent and uncertain output of distributed photovoltaics, distribution networks are prone to power flow reverse transmission and voltage over-limit problems, which in turn increases the difficulty of grid regulation.

[0031] In the related art, the voltage control method of the distribution network containing distributed photovoltaics using the traditional droop control method can keep the voltage fluctuating within the specified range, but the specified fluctuation range is large and easily affects the safe operation of the power grid. In addition, the traditional droop control method requires the incorporation of loads, which will cause the voltage fluctuation to exceed the specified range, thereby causing voltage instability.

[0032] In order to solve the deficiencies existing in the related art, the present invention proposes a coordinated control method for a distributed photovoltaic cluster.

[0033] Figure 1 FIG. 1 is a flow chart of a distributed photovoltaic cluster coordinated control method provided by an exemplary embodiment, wherein the distribution network includes photovoltaic nodes and non-photovoltaic nodes. Figure 1 As shown, the method may include the following steps:

[0034] Step 102, obtaining photovoltaic status data of each node in the photovoltaic cluster, and inputting the photovoltaic status data into a pre-established enhanced control model; wherein the photovoltaic status data includes voltage data and frequency data of each node, and the enhanced control model is established based on a reinforcement learning algorithm and a control consistency algorithm.

[0035] Reinforcement learning algorithm, including action set A, system state set S, and immediate reward r, uses trial and error to find an action sequence that produces the maximum cumulative reward. In general, the expected total reward Q is obtained by accumulating the rewards obtained by the corresponding actions in state s:

[0036]

[0037] Among them, r(s t ,a t ) is the agent in state s t Take action when t The discount parameter β is used to exponentially reduce the weight of the reward obtained in the next search. β∈[0,1]. When β=0, the system only considers the immediate benefit without considering the long-term benefit. When β=1, the system considers the long-term benefit to be equally important as the immediate benefit.

[0038] Reinforcement learning usually uses a greedy algorithm. In practice, the true optimal value of the total reward Q* is usually unknown, but it can be iteratively approximated by the reinforcement learning algorithm.

[0039] The pinning consensus algorithm is a distributed agent interaction rule. Controllable target values ​​are preset for some agents, and adjacent agents communicate with each other, so as to control the entire network by pinning down this small number of agents. The main goal of consistency control is to reach a consensus agreement among all agents in the network, which only requires interaction between neighbors. This method is to establish information exchange rules between a unit and all its neighbors, and realize the transmission of global information through local information.

[0040] The linear expression of the pinning consensus algorithm is:

[0041]

[0042] Where Δx i (t) is the deviation (adjustment amount) of node i at time t; N i is the set of nodes close to node i; a ij The information exchange coefficient between nodes i and j is x * is the containment value; i is the control gain of node i, δ i ∈[0,1], when node i is a pinning point, take δ i =1, otherwise δ i =0.

[0043] The reinforcement control model is based on the reinforcement learning algorithm and the control consistency algorithm, such as Figure 2 As shown in the figure, the action of the reinforcement learning method is realized through the control of the pinning. The core of this method is to reduce the number of controllers by adopting the pinning control strategy, while adapting to the weak communication conditions and the changes in the network structure. The implementation of reinforcement control learning only needs to be updated regularly at specific locations, thereby reducing the burden of offline computing training and the frequency of use of the reinforcement learning controller. At the same time, by adjusting the pinning matrix, the reinforcement control learning constructed by the distributed control structure can be adapted to the power plug-and-play operation.

[0044] The following are the steps for the fusion mechanism of pinning and reinforcement learning:

[0045] Step 1021, obtain the current state S t , initialize all Q(s t ,a t ); Step 1022, adopting the greedy control strategy, by selecting the optimal action value To achieve the maximum behavior value function Q(s t ,a t ); Step 1023, using the containment control method to implement the greedy action Step 1024, after the control is completed, the agent enters s t+1 state, calculate the reward function r(s t ,a t ); Update the behavior value function Q(s t ,a t ); Step 1025, update the state and return to step 1022, until the updated s is the terminal state, and the loop ends.

[0046] Step 104 , selecting a pinning node from each node, and selecting an optimal action value for the pinning consistency algorithm to implement pinning control through the greedy control strategy of the reinforcement learning algorithm adopted by the reinforced pinning control model.

[0047] Step 106 : executing a greedy action on the pinned node based on a pinning consistency algorithm, and using the optimal action value as a control instruction to coordinate various nodes of the photovoltaic cluster.

[0048] The inverter control mode of photovoltaic power generation often adopts a double closed-loop control strategy. In the present invention, the outer loop control adopts voltage loop control to achieve the voltage stabilization function; the inner loop control is current loop control to achieve the ability to improve the dynamic response of the system.

[0049] Generally, the main leading node in the cluster is selected as the containment point, and the corresponding containment reference value is related to the actual problem. For example, in a photovoltaic cluster, a spectrum node is usually selected as the containment point.

[0050] The primary control process of distributed photovoltaic power sources in the peer control mode is mainly the coordinated increase and decrease of output by distributed power sources. However, since droop control is a differential regulation, a corresponding method is required to achieve secondary control. The present invention adopts a voltage / frequency reference value adjustment method. According to the principle of droop control, the droop control formula taking into account the secondary control is:

[0051]

[0052] Among them, Δf, ΔU i They are the secondary control frequency and voltage reference control instructions respectively; f i,ref , U i,ref are the frequency and voltage reference values ​​of primary control respectively; m i 、n i The present invention collects the actual voltage frequency according to the main control agent, performs difference calculation with the rated value, and generates a secondary control value.

[0053] In one embodiment, using the optimal action value as a control instruction to coordinate the nodes of the photovoltaic cluster includes: forming an error matrix according to a predefined state deviation, and iteratively calculating an action value for the next control according to the optimal action value.

[0054] According to the optimal action selected by reinforcement control learning, all agents will achieve the optimal frequency and voltage control of the system. It is only necessary to design a reinforcement control learning controller in one agent, and other agent points can be synchronized through communication. When under distributed control, the system states all converge gradually to the corresponding rated value of 0.

[0055] like Figure 3 As shown, the specific control process is:

[0056] First, initialize the Q table and determine the pinning information matrix W p ;

[0057] Next, calculate the reward value r f 、r U , r f is the frequency reward value, r U is the voltage reward value;

[0058] Select greedy action in voltage-frequency control:

[0059]

[0060] Use the containment control method to implement control and define the state deviation And form the error matrix According to the iterative control instruction of the control, the action value of the k+1th control is calculated, that is, the control instruction:

[0061]

[0062] Among them, W p To contain the information matrix.

[0063] According to the control gain δ of the pinning point i , the information exchange coefficient a between nodes i and j ij , determine the update matrix W corresponding to the i-th agent being pinned down p The expression is:

[0064]

[0065] After calculating the action value, the corresponding adjustment value f is updated i, ' ref , U i ' .ref, then calculate the instant reward value of voltage-frequency control, and update the function table according to the selected greedy action.

[0066] In this embodiment, a reinforcement pinning control model is established based on the reinforcement learning algorithm and the pinning consistency algorithm, and the photovoltaic cluster is coordinated through the model. On the one hand, since the reinforcement pinning control model has the advantages of both the reinforcement learning algorithm and the pinning consistency algorithm, it can not only realize autonomous learning under the conditions of weak communication and network structure changes, but also reduce the workload of offline training; on the other hand, when the power grid is disturbed by the outside world, the reinforcement learning and pinning control methods can quickly and accurately generate power corrections, optimize the interaction power between clusters, thereby stabilizing the system frequency and bus voltage near the rated value, and improving the reactive elastic support capacity of the distributed photovoltaic cluster under the peer control mode.

[0067] In one embodiment, the greedy control strategy of the reinforcement learning algorithm adopted by the reinforced containment control model selects the optimal action value for the containment consistency algorithm to implement containment control, including: calculating the reward value of the coordinated photovoltaic cluster; if the reward value does not reach a preset threshold, re-coordinating and controlling the photovoltaic cluster until the reward value reaches the preset threshold.

[0068] In one embodiment, before obtaining photovoltaic status data of each node in the photovoltaic cluster, that is, before coordinating each node in the photovoltaic cluster, the method further includes: evaluating the load-side resource regulation potential of the photovoltaic cluster, and optimizing and regulating the photovoltaic cluster according to the evaluation results.

[0069] Furthermore, the evaluating the load-side resource regulation potential of the photovoltaic cluster includes: analyzing, evaluating, and predicting the photovoltaic cluster in sequence to calculate the load-side resource regulation potential value of the photovoltaic cluster.

[0070] Considering the uncertainty of load-side resource response, the load-side resource regulation potential is evaluated. The load-side resource response uncertainty evaluation method is:

[0071] As demand response continues to deepen, the load side changes from rigid uncontrollable to flexible controllable, the adjustable resources increase, and actively participate in grid regulation to improve the safety and stability of the grid. The load is affected by random factors such as the external environment and economic incentives, and the uncertainty of the response reduces the controllable performance of the grid. Therefore, according to the influencing factors of different types of load-side resource demand response, a load-side resource response uncertainty assessment method based on Markov process is proposed to quantify the uncertainty of controllable load response behavior, which is conducive to improving the system control accuracy.

[0072] Considering that various types of electricity users have different evaluation criteria for electricity comfort and electricity satisfaction, and the electricity consumption status of users in different time periods is affected by a variety of random factors. Therefore, using a suitable user electricity consumption state transfer matrix, it is possible to describe the user's electricity consumption behavior characteristics under different influencing factors, that is, the user's load transfer characteristics. The user demand response uncertainty assessment method based on Markov process is to establish a model based on the different discrete degrees of the user's electricity consumption state space:

[0073] The user's electricity consumption sequence satisfies the Markov property of the Markov chain, and the expression is:

[0074] P{I i (t+1)=x i (t+1)|I i (1) = x i (1),…,I i (t) = x i (t)}

[0075] =P{I i (t+1)=x i (t+1)|I i (t) = x i (t)};

[0076] Where t is the tth time period in the user's electricity consumption sequence; i is the user number; I i (t) is the power consumption state space of user i in time period t; x i (t)∈I i (t) is the electricity consumption of user i at time t.

[0077] Considering that the electricity consumption of users in different time periods has a certain range, It represents the lower limit of the power consumption of user i in time period t, that is, the power of the basic load on the user side, represents the upper limit of the power consumption of user i in time period t, that is, the maximum power consumption of user i in time period t when all appliances on the user side are turned on. The user's electricity usage range Discretized into n kinds of power consumption states, Indicates. represents the probability distribution of n kinds of power consumption states of user i. The power consumption state of user i in time period t is The probability of

[0078] Due to the different regulation requirements and incentives of the power system in different periods of time, the power consumption status of users in different periods of time is also different. Indicates that user i changes from power consumption state in time period t Transfer to the power consumption state of the time period (t+Δt) The state transition probability.

[0079]

[0080] Among them, P i t,t+Δt (λ i (t)) is the influencing factor λ i (t) is the power consumption state transfer matrix of user i, a 11 ,a 12 ,…,a nn ,b 11 ,b 12 ,…,b nn is the weight coefficient of different influencing factors.

[0081] The load-side resource regulation potential is evaluated from three aspects: intelligent analysis, intelligent evaluation and intelligent prediction based on data mining.

[0082] Intelligent analysis is to collect power grid operation data, clean the data, identify outliers based on confidence interval and fill missing value data with KNN algorithm. Finally, the improved K-means algorithm is used to perform preliminary clustering on the preprocessed data.

[0083] like Figure 4 As shown in the figure, the intelligent evaluation model consists of two parts. The first part builds a load regulation potential evaluation index system: environmental characteristic index, load characteristic index and behavioral characteristic index. The second part is based on the improved K-means clustering model. Clustering is used for various types of load data to study customer power consumption patterns and load characteristics. Combined with the regulation potential reported by users, a load regulation potential method is constructed to obtain the regulation potential results of various models.

[0084] Intelligent prediction is based on the load regulation potential obtained from intelligent evaluation, and the load regulation potential needs to be predicted. Considering the data characteristics of the regulation potential itself, the transfer probability of each load state is established, and the VMD variational mode decomposition method is used to perform mode decomposition on the regulation potential data. VMD can obtain the optimal solution in the variational model through repeated iterations. Then the regulation potential is predicted through the time convolution network TCN.

[0085] The upper and lower limits of load regulation potential are calculated through intelligent analysis, intelligent evaluation and intelligent prediction. The calculation formula for the upper and lower limits of load regulation potential is:

[0086]

[0087] Among them, P up and P downare the upper and lower limits of load regulation potential, m u and m d are the index upward and downward adjustment coefficients, respectively, with a value range of (0,1], C is the normalized index value, P max,d and P min,d are the maximum load value and the minimum load value in period d, respectively, ave,d is the average load value within period d.

[0088] Specifically, the photovoltaic cluster is optimized and regulated according to the evaluation results, including: optimizing and regulating the photovoltaic cluster based on a pre-established distributed photovoltaic cluster optimization and regulation model. The distribution network is optimized and regulated by combining the distributed photovoltaic, the active reduction and reactive compensation of the adjustable load and the reactive output power of the static VAR compensator (SVC), with the goal of minimizing network loss, voltage deviation at the restraining point and photovoltaic abandoned light. The objective function of the distributed photovoltaic cluster optimization and regulation model is:

[0089] F=min(λ 1 f 1 +λ 2 f 2 +λ 3 f 3 )

[0090]

[0091] f 3 =sum(P PV -P PV1 );

[0092] Among them, f 1 、f 2 and f 3 are network loss, voltage deviation at the tie point and photovoltaic abandonment, respectively, 1 , 2 and λ 3 is the corresponding weight coefficient, P ij , Q ij and r ij are the active power, reactive power and line impedance between node i and node j respectively, U i is the voltage at node i, U base is the line reference voltage, P PV is the photovoltaic output, P PV1 is the photovoltaic power consumption.

[0093] From the perspective of safe operation of distribution network and distributed photovoltaic cluster, the constraints of the optimization control model between distributed photovoltaic clusters take into account photovoltaic power station (PV) constraints, power flow constraints, node power constraints, system voltage and current constraints, and static VAR compensator (SVC) constraints.

[0094] Among them, the PV constraint can be expressed as:

[0095]

[0096] Among them, P i,PV and P i,PVmax are the maximum active power and active power of PV at node i respectively; Q i,PV is the maximum value of the PV reactive power of node i; tanθ is the current power factor angle.

[0097] The power flow constraint can be expressed as:

[0098]

[0099] Among them, P i and Q i are the total active and reactive power of node i respectively; P j and Q j are the total active and reactive power of node j, U i and U j are the voltages at nodes i and j.

[0100] The node power constraint can be expressed as:

[0101]

[0102] Among them, P i sub and Q i sub are the active and reactive power injected by the i-node substation, P i,up and P i,down are the maximum and minimum load regulation of node i, Q i,SVC,min and Q i,SVC,max are the minimum and maximum values ​​of the SVC power at the i-node respectively.

[0103] The system voltage and current constraints can be expressed as:

[0104]

[0105] Among them, I ij and I ijmax are the current current ij and the maximum current allowed to pass through the branch ij respectively; U imax and U iminThe maximum and minimum voltages that the system is allowed to carry.

[0106] In this embodiment, on the basis of establishing an inter-cluster optimization scheduling model, energy complementarity among photovoltaic clusters within multiple clusters is coordinated by strengthening the containment control, so as to achieve coordinated control among clusters, ensure the normal operation of photovoltaic and load, promote the local consumption of new energy, and support the power supply of the power grid, thereby improving the power supply reliability and economy of the distribution network.

[0107] like Figure 5 As shown in the figure, when the distribution network exceeds the limit, under the optimized dispatching of the power grid, the power is transmitted through the network inside and outside the cluster. The coordinated optimization control between clusters is realized through the optimized dispatching between clusters. The control nodes between clusters exchange agent boundary information, coordinate the imbalance of active and reactive power between clusters in the entire distribution network, and implement frequency and voltage control. When the frequency / voltage of the power grid exceeds the limit, the power output of the nodes in the cluster is adjusted according to the frequency / voltage change through the strengthened control within the cluster, and the control nodes transmit power with the nodes in the cluster and the nodes that exceed the limit. On this basis, through the optimized coordinated regulation between clusters, the power exchange between clusters is realized, the frequency / voltage deviation of the power grid is reduced, and the frequency / voltage of the power grid is ensured to be stable within the safe operating range.

[0108] Figure 6 is a schematic structural diagram of a device provided by an exemplary embodiment. Figure 6 At the hardware level, the device includes a processor 602, an internal bus 604, a network interface 606, a memory 609, and a non-volatile memory 610, and may also include hardware required for other functions. One or more embodiments of the present invention may be implemented based on software, such as the processor 602 reading the corresponding computer program from the non-volatile memory 610 into the memory 608 and then running it. Of course, in addition to the software implementation, one or more embodiments of the present invention do not exclude other implementations, such as logic devices or a combination of software and hardware, etc., that is, the execution subject of the following processing flow is not limited to each logic unit, but may also be hardware or logic devices.

[0109] Please refer to Figure 7 , a distributed photovoltaic cluster coordination control device can be applied to Figure 7 In the device shown, to implement the technical solution of the present invention, the distribution network includes photovoltaic nodes and non-photovoltaic nodes, and the device may include:

[0110] An acquisition unit 702 is used to acquire photovoltaic status data of each node in the photovoltaic cluster, and input the photovoltaic status data into a pre-established enhanced control model; wherein the photovoltaic status data includes voltage data and frequency data of each node, and the enhanced control model is established based on a reinforcement learning algorithm and a control consistency algorithm;

[0111] A selection unit 704 is used to select a pinning node from each node, and select an optimal action value for the pinning consistency algorithm to implement pinning control through a greedy control strategy of the reinforcement learning algorithm adopted by the reinforcement pinning control model;

[0112] The calculation unit 706 is used to perform a greedy action on the pinned node based on a pinning consistency algorithm, and use the optimal action value as a control instruction to coordinate the nodes of the photovoltaic cluster.

[0113] Optionally, the selection unit 704 is specifically configured to:

[0114] Calculating the reward value of the PV cluster after coordination;

[0115] When the reward value does not reach the preset threshold, the photovoltaic cluster is coordinated and controlled again until the reward value reaches the preset threshold.

[0116] Optionally, before obtaining the photovoltaic status data of each node in the photovoltaic cluster, the device further includes:

[0117] The evaluation unit 708 is used to evaluate the load-side resource regulation potential of the photovoltaic cluster and optimize and regulate the photovoltaic cluster according to the evaluation result.

[0118] Optionally, the evaluation unit 708 is specifically configured to:

[0119] The photovoltaic cluster is analyzed, evaluated and predicted in turn to calculate the load-side resource regulation potential value of the photovoltaic cluster.

[0120] Optionally, the evaluation unit 708 is specifically configured to:

[0121] The photovoltaic cluster is optimized and regulated based on a pre-established distributed photovoltaic cluster optimization and regulation model; wherein the objective function of the distributed photovoltaic cluster optimization and regulation model is established with the goal of minimizing network loss, restraint point voltage deviation and photovoltaic abandoned light amount.

[0122] Optionally, the device further comprises:

[0123] The control unit 710 is used to coordinate the nodes of the photovoltaic cluster according to the principle of droop control and the control instruction. The droop control formula is:

[0124]

[0125] Among them, Δf, ΔU i They are the secondary control frequency and voltage reference control instructions respectively; f i,ref , U i,ref are the frequency and voltage reference values ​​of primary control respectively; m i 、n i They are the coefficients of active and reactive regulation of droop control respectively.

[0126] Optionally, the calculation unit 706 is specifically configured to:

[0127] An error matrix is ​​formed according to the predefined state deviation, and the action value of the next control control is iteratively calculated according to the optimal action value.

[0128] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer, which may be in the form of a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email transceiver, a game console, a tablet computer, a wearable device or a combination of any of these devices.

[0129] In a typical configuration, a computer includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0130] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0131] Computer-readable media include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, disk storage, quantum memory, graphene-based storage media or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined in the present invention, computer-readable media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0132] With respect to the computer-readable medium (or computer-readable storage medium) as described above or in any other form, computer instructions may be stored thereon, and when the instructions are executed by a processor, one or more of the above-mentioned embodiments are implemented, thereby realizing the technical solution of the present invention.

[0133] The present invention also proposes a computer program, which, when executed by a processor, implements one or more of the above embodiments, thereby realizing the technical solution of the present invention. The computer program may be specifically recorded in the above or any other form of computer-readable medium, and the present invention is not limited thereto.

[0134] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0135] The above describes a specific embodiment of the present invention. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0136] The terms used in one or more embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of the present invention. The singular forms of "a", "said" and "the" used in one or more embodiments of the present invention and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in the present invention refers to and includes any or all possible combinations of one or more associated listed items.

[0137] It should be understood that although the terms first, second, third, etc. may be used to describe various information in one or more embodiments of the present invention, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of the present invention, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0138] The above description is merely a preferred embodiment of one or more embodiments of the present invention and is not intended to limit one or more embodiments of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of the present invention shall be included in the scope of protection of one or more embodiments of the present invention.

Claims

1. A distributed photovoltaic cluster coordinated control method, characterized in that: The method comprises: Acquire photovoltaic status data of each node in the photovoltaic cluster, and input the photovoltaic status data into a pre-established enhanced control model; wherein the photovoltaic status data includes voltage data and frequency data of each node, and the enhanced control model is established based on a reinforcement learning algorithm and a control consistency algorithm; Selecting a pinning node from each node, and selecting an optimal action value for the pinning consistency algorithm to implement pinning control through a greedy control strategy of a reinforcement learning algorithm adopted by the reinforced pinning control model; A greedy action is performed on the pinned node based on a pinning consistency algorithm, and the optimal action value is used as a control instruction to coordinate various nodes of the photovoltaic cluster.

2. The method according to claim 1, characterized in that The greedy control strategy of the reinforcement learning algorithm adopted by the reinforced containment control model selects the optimal action value for the containment consistency algorithm to implement containment control, including: Calculating the reward value of the PV cluster after coordination; When the reward value does not reach the preset threshold, the photovoltaic cluster is coordinated and controlled again until the reward value reaches the preset threshold.

3. The method according to claim 1, characterized in that Before obtaining the photovoltaic status data of each node in the photovoltaic cluster, the method further includes: The load-side resource regulation potential of the photovoltaic cluster is evaluated, and the photovoltaic cluster is optimized and regulated based on the evaluation results.

4. The method according to claim 3, characterized in that The evaluation of the load-side resource regulation potential of the photovoltaic cluster includes: The photovoltaic cluster is analyzed, evaluated and predicted in turn to calculate the load-side resource regulation potential value of the photovoltaic cluster.

5. The method according to claim 3, characterized in that: The optimizing and regulating the photovoltaic cluster according to the evaluation result includes: The photovoltaic cluster is optimized and regulated based on a pre-established distributed photovoltaic cluster optimization and regulation model; wherein the objective function of the distributed photovoltaic cluster optimization and regulation model is established with the goal of minimizing network loss, restraint point voltage deviation and photovoltaic abandoned light amount.

6. The method according to claim 1, characterized in that The method further comprises: According to the principle of droop control and the control instruction, the nodes of the photovoltaic cluster are coordinated. The droop control formula is: Among them, Δf, ΔU i They are the secondary control frequency and voltage reference control instructions respectively; f i,ref , U i,ref are the frequency and voltage reference values ​​of primary control respectively; m i 、n i They are the coefficients of active and reactive regulation of droop control respectively.

7. The method according to claim 1, characterized in that The method of using the optimal action value as a control instruction to coordinate the nodes of the photovoltaic cluster includes: An error matrix is ​​formed according to the predefined state deviation, and the action value of the next control control is iteratively calculated according to the optimal action value.

8. A distributed photovoltaic cluster coordination control device, characterized in that: The device comprises: An acquisition unit: acquires photovoltaic status data of each node in the photovoltaic cluster, and inputs the photovoltaic status data into a pre-established enhanced control model; wherein the photovoltaic status data includes voltage data and frequency data of each node, and the enhanced control model is established based on a reinforcement learning algorithm and a control consistency algorithm; A selection unit: selecting a pinning node from each node, and selecting an optimal action value for the pinning consistency algorithm to implement pinning control through a greedy control strategy of a reinforcement learning algorithm adopted by the reinforcement pinning control model; The computing unit executes a greedy action on the pinned node based on a pinning consistency algorithm, and uses the optimal action value as a control instruction to coordinate the various nodes of the photovoltaic cluster.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor implements the steps of the method according to any one of claims 1 to 7 by running the executable instructions.

10. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the instruction is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images