Distributed photovoltaic hierarchical cluster frequency-voltage cooperative support control method

Abstract

The invention discloses a distributed photovoltaic hierarchical cluster frequency-voltage cooperative support control method. According to the method, the influence of distributed photovoltaic participation in system frequency modulation on distribution network voltage is considered, and distributed photovoltaic cluster division is realized based on a non-negative matrix factorization technology; in a global optimization layer, establishing a distributed photovoltaic active / frequency and reactive / voltage local droop control parameter optimization model, decomposing the distributed photovoltaic active / frequency and reactive / voltage local droop control parameter optimization model into a plurality of sub-models, and performing coordinated optimization based on an alternating direction multiplier method; in the local coordination layer, local voltage out-of-limit of the distribution network is relieved through voltage self-correction control in a cluster; and in the local control layer, distributed photovoltaic droop control parameters are configured based on a calculation result of the global optimization layer, and then the distributed photovoltaic is combined with a droop control strategy to provide rapid frequency-voltage cooperative support. According to the method, large-scale distributed photovoltaic can be efficiently coordinated, frequency support is provided for a power transmission network, and the voltage out-of-limit problem of the distribution network is relieved.

Description

Technical Field

[0001] The present invention relates to a distributed photovoltaic hierarchical cluster frequency-voltage coordinated support control method, belonging to the field of power system operation control. Background Art

[0002] The massive access of distributed photovoltaics seriously threatens the safe operation of the system. Distributed photovoltaics have flexible power regulation capabilities and can participate in both system frequency regulation and local voltage control simultaneously. Efficient coordinated control of large-scale distributed photovoltaics to fully exploit and utilize their power grid support potential is of great significance for improving the global frequency stability of the system and ensuring the local voltage safety of the distribution network.

[0003] In existing technical solutions, generally, the active and reactive power outputs of distributed photovoltaics are changed to participate in system frequency control and voltage regulation. However, the existing control technologies or solutions have the following problems: Distributed photovoltaics have small single-unit capacities, large numbers, and wide distributions. The coordinated control problem of distributed photovoltaics has characteristics such as high dimensionality and strong nonlinearity, and it is difficult for traditional centralized control schemes to achieve efficient coordinated control; Existing zoning / hierarchical control schemes do not consider the coupling problems brought about by the simultaneous participation of distributed photovoltaics in frequency and voltage control, and cannot guarantee the power grid support effect of distributed photovoltaics; The output power and load demand of distributed photovoltaics fluctuate randomly, resulting in rapid changes in the system operation state. Traditional optimal control schemes have good effects but slow response speeds, while pure local control has a fast response speed but poor adaptability, and it is impossible to achieve the multi-time-scale optimal coordinated control of distributed photovoltaics. Summary of the Invention

[0004] In view of the limitations of the related background art, the present invention provides a distributed photovoltaic hierarchical cluster frequency-voltage coordinated support control method. This method proposes a "global optimization - local coordination - local control" distributed photovoltaic hierarchical cluster coordinated control scheme, and realizes the coordinated support for system frequency stability and distribution network voltage security through the coordinated control of large-scale distributed photovoltaics. The method of the present invention considers the influence of distributed photovoltaics participating in system frequency regulation on the distribution network voltage, improves the electrical distance index, and realizes cluster division based on non-negative matrix factorization to enhance the rationality of distributed photovoltaic cluster division. To improve the adaptability of droop control, an optimization model for distributed photovoltaic active / frequency and reactive / voltage droop control parameters is constructed in the global optimization layer, and inter-group coordinated optimization is carried out based on the alternating direction multiplier method to optimize and adjust the droop control parameters on the hourly time scale. Further, in view of the problem of local voltage over-limit in the distribution network caused by the rapid change of distributed photovoltaic power generation or load demand, a voltage self-correction control strategy within the cluster is applied in the local coordination layer to quickly adjust the local voltage through a small amount of calculation. Finally, in the local control layer, combined with the calculation results of the global optimization layer, the distributed photovoltaic active / frequency and reactive / voltage droop control parameters are configured to provide real-time active frequency-voltage support based on droop control. Through hierarchical cluster coordinated control, the multi-time scale efficient coordinated control of large-scale distributed photovoltaics is realized, providing aggregated frequency support for the transmission network (that is, by adjusting the output power of multiple distributed photovoltaics, the power at the outlet of the distribution network substation is adjusted, thereby providing frequency regulation power support for the transmission network), and at the same time alleviating the problem of distribution network voltage over-limit under complex operating conditions.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A distributed photovoltaic hierarchical cluster frequency-voltage coordinated support control method, which is a large-scale distributed photovoltaic voltage-frequency hierarchical coordinated support control method including a "global optimization layer - local coordination layer - local control layer", and specifically includes the following steps:

[0007] 1) Considering the influence of distributed photovoltaics participating in system frequency regulation on the distribution network voltage, cluster division of distributed photovoltaics is carried out based on non-negative matrix factorization technology to simultaneously integrate network topology and node attribute information, thereby enhancing the rationality of the distributed photovoltaic cluster division result;

[0008] 2) Construct a hierarchical cluster coordinated control scheme including a global optimization layer, a local coordination layer and a local control layer to realize the multi-time scale coordinated control of distributed photovoltaics in hours - minutes - real time, so as to provide aggregated frequency support and alleviate the problem of distribution network voltage over-limit;

[0009] 3) At the global optimization layer, aiming to minimize the loss of distributed photovoltaic power generation benefits, the use of reactive power, and network losses, and with the power flow equation of the distribution network, the operating characteristics of distributed photovoltaics, and the local droop control function as constraints, an optimization model for active / frequency and reactive / voltage local droop control parameters is constructed;

[0010] 4) Based on the cluster division results, reconstruct the constraints related to frequency regulation in the optimization model of active / frequency and reactive / voltage local droop control parameters, and add consistency constraints, thereby decomposing the optimization model of active / frequency and reactive / voltage local droop control parameters into multiple sub-models, and coordinating and optimizing all sub-models based on the alternating direction multiplier method to reduce the computational burden, so as to finally achieve the fair distribution of frequency regulation power and the optimal design of local reactive / voltage droop control parameters, and obtain the optimized active / frequency and reactive / voltage droop control parameters;

[0011] 5) At the local coordination layer, for distributed photovoltaic clusters with voltage violations, by modifying the reactive power reference value of the reactive / voltage droop control of distributed photovoltaics within the cluster, voltage self-correction control within the distributed photovoltaic clusters with voltage violations is realized;

[0012] 6) At the local control layer, based on the parameters calculated by the global optimization layer (active / frequency droop control gain, local reactive / voltage control gain, voltage dead zone, and saturation parameters), configure the active / frequency and reactive / voltage droop control parameters of each distributed photovoltaic, and then each distributed photovoltaic combines local frequency and voltage measurement values to quickly respond to frequency and voltage deviations, realizing immediate frequency and voltage coordinated support.

[0013] In the above technical solution, further, in step 1), considering the influence of distributed photovoltaics participating in system frequency regulation on the distribution network voltage, the electrical distance index is improved, and based on the non-negative matrix factorization technology, the distributed photovoltaics are clustered. The specific implementation process is as follows:

[0014] Assume the number of network nodes is , and the electrical distance index between nodes can be defined based on the reactive power-voltage sensitivity of the distribution network:

[0015]

[0016] In the formula: represents the electrical distance index between node and ; and respectively represent the th and th diagonal elements of the reactive power-voltage sensitivity matrix of the distribution network; represents The rd row and th column element. Since the active power adjustment of distributed PV has a significant impact on the node voltage, this impact must be considered during cluster division. Assume that the aggregated supporting power of the distribution network (i.e., the power adjustment amount at the outlet of the distribution network substation) is , and each distributed PV fairly undertakes the frequency regulation task according to its capacity. Then, the impact of the distributed PV located at node

[0017]

[0018] on the voltage of distribution network node can be expressed as: In the formula: represents the voltage change at node ; is a constant used to quantify the impact of the injection power adjustment at node on the power at the substation outlet; represents the impact of the injection of unit power by the distributed PV located at node on the voltage of node ; is the th row and

[0019] th

[0020] column element of the active-voltage sensitivity matrix ; represents the installed capacity of the distributed PV at node . Based on this, the electrical distance index can be modified as: In the formula: represents the improved electrical distance index; represents the impact of the injection of unit power by the distributed PV located at node

[0021] on the voltage of node ; and

[0022] respectively represent the impacts of the injection of unit power by the distributed PVs located at node on the voltages of nodes and

[0023] ; Denote the set consisting of all nodes in the distribution network except the reference node (i.e., the node at the substation outlet).

[0024] The diagonal elements of the voltage sensitivity matrix reflect the impact of the adjustment of each distributed PV power on the local voltage, while the linearization coefficient approximately characterizes the contribution of the active power regulation of distributed PV to the aggregated support power of the distribution network. Without loss of generality, here the distribution network node attribute matrix is defined as:

[0025]

[0026] In the formula: is a constant coefficient vector; denotes taking the diagonal elements of the matrix and forming a column vector.

[0027] Node similarity matrix can be constructed in combination with cosine similarity, and its elements are , representing the similarity between node and node , and can be specifically expressed as:

[0028]

[0029] In the formula: and are the th nd and th column vectors of the node attribute matrix

[0030] In order to find a set of nodes within the clusters that are closely connected and have homogeneous attributes, the cluster partitioning task is constructed as a non-negative matrix factorization problem. By decomposing the node-edge adjacency matrix of the network and the node similarity matrix , the node-cluster membership matrix is obtained. The element in the th row and th column of the node-cluster membership matrix represents the membership degree of the distribution network node to the cluster . The row number corresponding to the largest element in each column of the node-cluster membership matrix is the number of the cluster to which the corresponding node belongs. The non-negative matrix factorization problem for distributed PV cluster partitioning is expressed as:

[0031]

[0032] In the formula: and are two non - negative basic matrices; represents the Frobenius norm; is a positive real parameter used to balance the contributions of topological and attribute information; the third term of the objective function is a regularization term used to control the number of nodes within each cluster. is a regularization constant; represents the matrix the -th element of represents the total membership degree of each cluster; is a constant vector with all elements equal to 1.

[0033] Furthermore, in step 2), based on the hierarchical cluster coordinated control scheme, the specific method for implementing distributed photovoltaic hour - minute - real - time multi - time - scale coordinated control is as follows:

[0034] The hierarchical cluster coordinated control scheme includes three control levels: the global optimization layer, the local coordination layer, and the local control layer. The global optimization layer performs distributed optimization once per hour to adjust the active / frequency and reactive / voltage droop control parameters of distributed photovoltaics. The local coordination layer, on the minute - level time scale, performs in - cluster voltage self - correction control by modifying the reactive power reference value of the reactive / voltage droop control of distributed photovoltaics within the cluster, thereby controlling the reactive power output of distributed photovoltaics within the cluster to alleviate the over - limit of the distribution network voltage. In the local control layer, combined with the optimization design results of the global optimization layer, the active / frequency and reactive / voltage droop control parameters of each distributed photovoltaic are configured. The distributed photovoltaic combines local frequency and voltage measurement values and responds to the system frequency and voltage deviations in real time. Through the mutual cooperation of the three control levels, the coordinated control of distributed photovoltaics at multiple time scales is achieved, thereby providing aggregated frequency support for the system and effectively alleviating the problem of over - limit of the distribution network voltage caused by the fluctuations of distributed photovoltaic power generation and load demand.

[0035] Furthermore, in step 3), the optimization model of the local active / frequency and reactive / voltage droop control parameters can be specifically expressed as:

[0036]

[0037]

[0038]

[0039]

[0040]

[0041]

[0042]

[0043]

[0044]

[0045]

[0046]

[0047]

[0048]

[0049]

[0050]

[0051]

[0052]

[0053]

[0054]

[0055]

[0056]

[0057]

[0058]

[0059] Wherein: is the objective function of the optimization model for the active / frequency and reactive / voltage local droop control parameters; , and are non - negative weight factors for distributed photovoltaic active reserve, reactive power, and network loss respectively, and satisfy ; The subscript represents the time; and represent respectively the upward and downward reserve powers of node at time is the inverter reactive power adjustment amount at time represents the square of the current amplitude on line ; represents the resistance of line ; Represents the set of distribution network lines; Represents the set of nodes installed with distributed photovoltaic; Is the optimization time domain; And Respectively represent the active and reactive power flows On each line Represents the downstream node set That has a line connection with node And Respectively represent the active and reactive power injected On node Represents the square of the voltage amplitude of node , Is the voltage amplitude of node ; Represents the reactance of line ; And Respectively represent the upper and lower limits of voltage; And Respectively are The maximum available and output active power of the distributed photovoltaic at time Of node And Respectively are the maximum allowable upward and downward active reserve rates; And Respectively represent The inverter output reactive power and reactive power reference value at time Of node Represents the distributed photovoltaic reactive power adjustment amount; Represents The maximum available reactive power of the inverter at time Of node And Respectively represent The upward and downward frequency modulation participation degrees of the distributed photovoltaic at time And Respectively represent The upward and downward frequency modulation gains at time Of node And Respectively are The column vectors composed of the upper / lower frequency modulation gains of each distributed photovoltaic at time And Respectively are The aggregated active / frequency upward and downward frequency modulation gains expected by the transmission network operator at time And Respectively are The distributed photovoltaic at time The upper and lower limits of the corresponding frequency dead zone; and are respectively the upper and lower limits of the frequency saturation threshold corresponding to the distributed photovoltaic at time ; and respectively represent the upward and downward voltage droop control gains of the node at time ; and are respectively the upper and lower saturation thresholds of the reactive power / voltage droop control of the node at time ; and are respectively the upper and lower limits of the dead zone of the reactive power / voltage droop control of the node at time ; is the frequency measurement value; are respectively the 0-1 indicator variables related to the 5 linear segments of the reactive power / voltage droop control function of the node at time ; is a relatively large constant; represents the reactive power reference command that may be output by the distributed photovoltaic reactive power / voltage droop control of the node at time ; represents the reactive power reference command actually output by the voltage droop control of the node at time ; represents the Hadamard product; is the th element of; and are respectively the upper and lower limits of the voltage dead zone and saturation parameters; is the scaling factor.

[0060] Furthermore, in step 4), based on the cluster division result, the constraints related to frequency regulation in the optimization model of the active power / frequency and reactive power / voltage local droop control parameters are reconstructed, and the consistency constraint is added. The specific process is as follows:

[0061] Assume that the cluster division method in step 1) divides the distributed photovoltaic into clusters, then the distribution network can also be correspondingly divided into sub-networks. Here, the tie-line splitting method is used to divide the distribution network. If there is a common line between two clusters m and n, then at A virtual node is introduced in the middle and the line is disconnected at this virtual node. The entire network is divided into subnets m and n, and each subnet has a connection line and a boundary node. In distributed optimization, the subnets and each have an optimization model, called a sub-model. The impedance value of each connection line is half of that of the common line , so the consistency constraint between subnets m and n is:

[0062]

[0063] In the formula: and respectively represent the active power and reactive power flowing through the connection line between subnets and and calculated by subnet and respectively represent the active power and reactive power flowing through the connection line between subnets and and calculated by subnet and are respectively the voltages at the boundary nodes calculated by subnet and subnet .

[0064] Therefore, the constraints related to frequency regulation in the optimization model of the active / frequency and reactive / voltage local droop control parameters are reconstructed (i.e., the equality constraint corresponding to the active / frequency droop control gain: ), which is specifically expressed as:

[0065]

[0066]

[0067] In the formula: and respectively represent the downward and upward frequency regulation reference commands issued by the transmission network operator at time and respectively represent the downward and upward frequency regulation reference commands sent from subnet to subnet at time and respectively represent the downward and upward frequency regulation reference commands sent from subnet to subnet Downward and upward frequency modulation reference commands; Represents the linearization constant coefficient corresponding to subnet m, which can be calculated separately for each subnet; and Respectively represent the Upstream and downstream subnet sets of subnet; Is a 0-1 indicator variable. 1 indicates that subnet m is connected to the substation, and this subnet receives the frequency modulation command issued by the transmission network operator. Otherwise, subnet Only receives the frequency modulation commands sent by adjacent subnets; Represents subnet Set of nodes installed with distributed photovoltaics in;

[0068] At the same time, additional consistency constraints are added. The purpose of adding consistency constraints is to make the tie-line power, boundary node voltage, frequency modulation reference command, and distributed photovoltaic frequency modulation participation degree of each subnet the same to ensure the physical feasibility of the calculation results. Specifically, it is expressed as:

[0069]

[0070]

[0071] In the formula: and Respectively represent At time, the Calculated by subnet Downward and upward frequency modulation reference commands sent to subnet; and Respectively represent At time, the Calculated by subnet Downward and upward frequency modulation reference commands sent to subnet; and Respectively represent the upward and downward frequency modulation participation degrees of distributed photovoltaics within subnet m; and Respectively represent the upward and downward frequency modulation participation degrees of distributed photovoltaics calculated by subnet ;

[0072] Define the vector composed of tie-line power, boundary node voltage, frequency modulation reference command, and distributed photovoltaic frequency modulation participation degree as the coupling variable. The above consistency constraints can be rewritten in vector form, specifically expressed as:

[0073]

[0074] In the formula: Represents the The vector composed of variables such as the relevant tie-line power, boundary node voltage, frequency regulation reference instruction, and distributed PV frequency regulation participation is called the coupling variable; similarly, represents the tie-line power, boundary node voltage, frequency regulation reference instruction, and the vector composed of variables such as distributed PV frequency regulation participation related to subnet ; and respectively represent the values of the coupling variables and calculated from subnet and .

[0075] Furthermore, in step 4), the objective function of each sub-model is:

[0076]

[0077] In the formula: represents the objective function of the sub-model corresponding to subnet m; , and are the non-negative weight factors for distributed PV active reserve, reactive power, and network loss of subnet , respectively, and satisfy ; represents the set of lines within subnet m; Specifically, it is obtained by solving the sub-model corresponding to subnet ; is the Lagrange multiplier vector; represents the penalty coefficient. Except for the equality constraints corresponding to the active / frequency droop control gain, the constraints in the active / frequency and reactive / voltage local droop control parameter optimization model are all included in the constraints of each sub-model. In addition, the constraints of each sub-model also include the above consistency constraints.

[0078] Furthermore, in step 4), the coordinated optimization of all sub-models based on the alternating direction multiplier method is as follows. The specific iteration process is:

[0079]

[0080] In the formula: and respectively represent the feasible decision spaces of sub-model m and sub-model , which are composed of the constraints of each sub-model; and represent the optimization variables of sub-model and sub-model ; the subscript represents the current time; and respectively represent the sub - model of the th iteration and the coupled variable values calculated by the sub - model ; the superscript represents the number of iterations; and respectively represent the optimization results obtained by the sub - model at the th iteration; and respectively represent the Lagrange multiplier vectors of the sub - model and the sub - model during the th iteration calculation process; represents the penalty coefficient of the sub - model during the

[0081] Define the th iteration calculation of the original residual and the dual residual of the sub - model

[0082]

[0083] where: is the coupled variable related to the sub - network ; represents the value of the coupled variable calculated by the sub - network and sent to the sub - network ; and are the Lagrange multipliers calculated at the current adjacent iteration step and the previous iteration step respectively.

[0084] Then the judgment condition for whether the algorithm converges after the th iteration is expressed as:

[0085] ,,

[0086] ,,

[0087] ,,

[0088] where: and respectively represent the residual vectors composed of the original residuals and dual residuals of all sub - models after the th iteration; is the maximum allowable residual of the alternating direction multiplier method and can be set to 1×10 -3 ; represents the infinity norm; represents the number of sub - models, that is, the number of distributed photovoltaic clusters.

[0089] To accelerate iterative convergence and reduce the number of iterations, a dynamic update strategy for the penalty coefficient is introduced. Taking the cluster as an example, its corresponding dynamic update strategy for the penalty coefficient is specifically expressed as:

[0090]

[0091] In the formula: is the adjustment factor; is the balance factor; represents the two - norm.

[0092] Furthermore, in step 5), by modifying the reactive power reference value of the internal distributed photovoltaic reactive power / voltage droop control, the reactive power output of the distributed photovoltaic is further adjusted to achieve the voltage self - correction control within the distributed photovoltaic cluster with voltage violation. The derivation process of the correction amount of the reactive power reference value of the distributed photovoltaic reactive power / voltage droop control is as follows:

[0093] Assume that due to the increase in the distributed photovoltaic output power, the voltage of node in the cluster rises from to , and each reactive power / voltage droop controller operates in the linear region. Then the distributed photovoltaic will change its reactive power output, and the steady - state voltage is expressed as:

[0094]

[0095] In the formula: represents the reactive power / voltage droop control gain of the distributed photovoltaic ; and respectively represent the initial voltage of node and the voltage without control after perturbation; then represents the steady - state voltage on node after reactive power / voltage droop control.

[0096] If the steady - state voltage exceeds the upper safety limit or the lower limit of the distribution network voltage, such as 1.05 p.u. or 0.95 p.u., then it is necessary to adjust the distributed photovoltaic reactive power / voltage droop control to compensate for reactive power and adjust the voltage within the safe range. When the voltage violates the limit, the reactive power reference value of the reactive power / voltage droop control can be modified, and the droop control curve is translated downward along the Q - axis to make the distributed photovoltaic inverter absorb more reactive power. The actual reactive power compensation amount required to maintain voltage safety is:

[0097]

[0098]

[0099] In the formula: represents the voltage correction amount; represents the actual reactive power compensation amount. If the voltage drops , then the reduction amount of the reactive power output of the distributed PV can be calculated according to the following formula:

[0100]

[0101] Therefore, the correction amount of the reactive power reference value for the reactive power / voltage droop control of the distributed PV should satisfy:

[0102]

[0103] In the formula: represents the correction amount of the reactive power reference value for the reactive power / voltage droop control of the distributed PV.

[0104] The above analysis is only for the case where a single distributed PV participates in the distribution network voltage regulation. When multiple distributed PVs participate in voltage regulation simultaneously, considering the interaction effects among multiple distributed PVs, the correction amount of the reactive power reference value for the reactive power / voltage droop control of each distributed PV can be calculated by the following formula:

[0105]

[0106] In the formula: represents the vector composed of the droop control gains of each distributed PV in the cluster; represents the diagonal matrix with the diagonal elements being ; ; represents the distributed PV cluster the number of nodes within.

[0107] By calculating the correction amount of the reactive power reference value for the reactive power / voltage droop control of the distributed PV and sending it to each distributed PV in the cluster to modify the reactive power reference value of the reactive power / voltage droop control of each distributed PV, the reactive power output of the distributed PV can be adjusted to solve the problem of voltage over-limit within the cluster.

[0108] Finally, in step 5), in the local control layer, based on the parameters calculated by the global optimization layer, the active / frequency and reactive / voltage droop control parameters of each distributed photovoltaic are configured; the local frequency and voltage measurement values are measured in real time. Once the frequency or voltage measurement value exceeds the frequency or voltage dead zone, the distributed photovoltaic will quickly adjust its active or reactive output according to the active / frequency or reactive / voltage droop control, so as to participate in the system frequency modulation and distribution network voltage control in real time. Among them, the active / frequency and reactive / voltage droop control of the local control layer are specifically expressed as:

[0109]

[0110]

[0111] In the formula: and respectively represent the real-time active and reactive adjustment amounts corresponding to the frequency and voltage deviations of the distributed photovoltaic at moment; represents the system frequency measurement value at moment; represents the voltage measurement value of node

[0112] The beneficial effects of the present invention are as follows:

[0113] Facing the frequency-voltage coordinated support control requirements of large-scale distributed photovoltaics, the present invention proposes a method for dividing and hierarchically coordinating the control of distributed photovoltaic clusters. Considering the influence of distributed photovoltaics participating in system frequency modulation on the distribution network voltage, and integrating network topology and node attribute information, a reasonable division of distributed photovoltaic clusters is realized based on the non-negative matrix factorization technology. On this basis, through the hierarchical cluster coordination control of "global optimization - local coordination - local control", the multi-time scale efficient coordination control of large-scale distributed photovoltaics and the frequency-voltage coordinated support are realized. Specifically, in the global optimization layer, an optimization model for reactive / voltage and active / frequency droop control parameters is constructed, and multiple clusters are coordinated based on the alternating direction multiplier method to balance the control speed and optimality. In the local coordination layer, a voltage self-correction control strategy within the cluster is proposed. By correcting the reactive reference value of the reactive / voltage droop control of the distributed photovoltaic, the active and reactive outputs of the distributed photovoltaic are changed, so as to cope with the local voltage over-limit problem caused by the real-time fluctuations of the source load. In the local control layer, the distributed photovoltaic quickly responds to the system frequency and voltage deviations based on the optimized and adjusted active / frequency and reactive / voltage droop control. Description of the Drawings

[0114] Figure 1 is the implementation process of the hierarchical cluster frequency-voltage coordinated support control method for distributed photovoltaics according to an embodiment of the present invention;

[0115] Figure 2 This is a comparison of the control effects of the method of the present invention and the existing control scheme on the improved IEEE 33-node test system, where (a) is the comparison of the frequency control effect and (b) is the comparison of the voltage control effect. Detailed implementation manners

[0116] The following further describes the present invention in conjunction with the drawings and embodiments. The flow of a distributed photovoltaic hierarchical cluster frequency-voltage coordinated support control method provided by the embodiments of the present invention is as Figure 1 shown.

[0117] A distributed photovoltaic hierarchical cluster frequency-voltage coordinated support control method specifically includes the following steps:

[0118] (1) Considering the influence of distributed photovoltaics participating in system frequency regulation on the distribution network voltage, improve the electrical distance index. The specific method is to initialize basic data such as the distribution network topology, line parameters, and source and load data, obtain the distribution network voltage sensitivity information, and calculate the improved electrical distance index, so as to form an adjacency matrix and a node attribute matrix; and use the non-negative matrix factorization technology to cluster the distributed photovoltaics, specifically as follows:

[0119] Assume that the number of distribution network nodes is , and the non-negative matrix factorization problem for realizing the clustering of distributed photovoltaics is expressed as:

[0120] ,

[0121] ,

[0122] ,

[0123] ,

[0124] ,

[0125] ,

[0126] ,

[0127] In the formula: represents the number of distributed photovoltaic clusters; represents the number of network nodes; represents the set composed of the remaining nodes in the distribution network except the reference node (i.e., the node at the substation outlet); and are two non-negative basic matrices; represents the Frobenius norm; is a positive real parameter used to balance the contributions of topology and attribute information; the third term of the objective function is a regularization term used to control the number of nodes within each cluster. is a regularization constant; represents the -th element of the matrix ; represents the total membership degree of each cluster; is a constant vector with all elements equal to 1; represents the node-cluster membership matrix; represents the node similarity matrix; represents the distribution network node attribute matrix; and are the -th and -th column vectors of the distribution network node attribute matrix respectively; is a constant coefficient vector, and its element is used to quantify the impact of the injection power adjustment of node on the substation outlet power; represents taking the diagonal elements of the matrix and forming a column vector; represents the node-edge adjacency matrix of the network, and its element is , indicating the electrical connection between node and node ; represents the improved electrical distance index; is a constant used to quantify the impact of the injection power adjustment of node on the substation outlet power; and represent the impacts of injecting a unit power of distributed PV located at node on the voltages of node and node respectively, and represent the impacts of injecting a unit power of distributed PV located at node on the voltages of node and node respectively; represents the active-voltage sensitivity matrix, is its element at the -th row and -th column; represents the installed capacity of distributed PV at node ; represents the reactive-voltage sensitivity; represents the electrical distance index between node and ; and respectively represent the th and th diagonal elements; represents the th row and th column element.

[0128] (2)Construct a hierarchical cluster coordinated control scheme including a global optimization layer, a local coordination layer, and a local control layer to achieve distributed PV multi-time scale coordinated control for providing fast frequency support and alleviating the problem of distribution network voltage over-limit.

[0129] The hierarchical cluster coordinated control scheme includes three control levels: the global optimization layer, the local coordination layer, and the local control layer; the global optimization layer performs distributed optimization once per hour to optimize the active / frequency and reactive / voltage droop control parameters of distributed PV; the local coordination layer performs intra-cluster voltage self-correction control on a minute-level time scale by modifying the reactive reference value of the reactive / voltage droop control of distributed PV within the cluster to adjust the reactive power output of distributed PV within the cluster, thereby alleviating the distribution network voltage over-limit; in the local control layer, combined with the optimization design results of the global optimization layer, the active / frequency and reactive / voltage droop control parameters of each distributed PV are configured, and the distributed PV combines local frequency and voltage measurement values to respond to system frequency and voltage deviations in real time; through the mutual cooperation of the three control levels, the coordinated control of distributed PV at multiple time scales is achieved.

[0130] (3)In the global optimization layer, aiming to minimize the loss of distributed PV power generation benefits, the use of reactive power, and network losses, and with the distribution network power flow equation, the operating characteristics of distributed PV, and the local droop control function as constraints, an optimization model for active / frequency and reactive / voltage local droop control parameters is constructed.

[0131] The optimization model for active / frequency and reactive / voltage local droop control parameters is specifically as follows:

[0132]

[0133]

[0134]

[0135]

[0136]

[0137]

[0138]

[0139]

[0140]

[0141]

[0142]

[0143]

[0144]

[0145]

[0146]

[0147]

[0148]

[0149]

[0150]

[0151]

[0152]

[0153]

[0154]

[0155] Wherein: is the objective function of the optimization model for the active / frequency and reactive / voltage local droop control parameters; , and are both non - negative weight factors regarding the active reserve, reactive power and network loss of distributed photovoltaic, and satisfy ; the subscript represents the time; and represent the upward and downward reserve powers respectively; is the reactive power adjustment amount of the inverter; represents the square of the current amplitude on line ; represents the resistance of line ; represents the set of distribution network lines; Denote the set of nodes installed with distributed PVs; For optimizing the time domain; and respectively represent the active and reactive power flows on each line; Denote the set of downstream nodes that are connected to node by lines; and respectively represent the active and reactive powers injected into node ; Denote the square of the voltage magnitude of node , where is the voltage magnitude of node ; Denote the reactance of line ; Denote the upper and lower limits of voltage and respectively; and respectively represent the maximum available and output active powers of distributed PVs; and respectively represent the maximum allowable upward and downward active reserve ratios; and respectively represent the inverter output reactive power and reactive power reference value; Denote the reactive power adjustment amount of distributed PVs; Denote the maximum available reactive power of the inverter; and respectively represent the upward and downward frequency regulation participation degrees of distributed PVs; and respectively represent the upward and downward frequency regulation gains; and respectively represent the column vectors composed of the upward / downward frequency regulation gains of each distributed PV; and respectively represent the aggregated active / frequency upward and downward frequency regulation gains expected by the transmission network operator; and respectively represent the upper and lower limits of the frequency dead zone corresponding to distributed PV at time ; and respectively represent the upper and lower limits of the frequency saturation threshold corresponding to distributed PV at time ; and respectively represent the upward and downward voltage droop control gains; and respectively represent the upper and lower saturation thresholds of the reactive power / voltage droop control; and are the upper and lower limits of the dead zone of the reactive power / voltage droop control respectively; is the measured frequency value; are the 0-1 indicator variables related to the 5 linear segments of the reactive power / voltage droop control function respectively; is a relatively large constant; represents the reactive power reference command that may be output by the distributed PV reactive power / voltage droop control; represents the reactive power reference command actually output by the voltage droop control; represents the Hadamard product; is the th element of; and are the upper and lower limits of the voltage dead zone and the saturation parameter respectively; is the scaling factor.

[0156] (4) Based on the cluster division results, reconstruct the constraints related to frequency regulation in the active power / frequency and reactive power / voltage local droop control parameter optimization model, and add consistency constraints, so as to decompose the active power / frequency and reactive power / voltage local droop control parameter optimization model into multiple sub-models, and based on the alternating direction multiplier method, conduct coordinated optimization of multiple sub-models, so as to realize the optimal design of the active power / frequency and reactive power / voltage droop control parameters of each distributed PV.

[0157] Reconstruct the constraints related to frequency regulation in the active power / frequency and reactive power / voltage local droop control parameter optimization model (i.e., the equality constraint corresponding to the active power / frequency droop control gain: ), which is specifically expressed as:

[0158] ,

[0159] ,

[0160] In the formula: the subscript represents the time, and respectively represent the downward and upward frequency regulation reference commands sent from subnet to subnet ; and respectively represent the downward and upward frequency regulation reference commands sent from subnet to subnet ; represents the linearized constant coefficient corresponding to subnet m, which can be calculated separately by each subnet; and respectively represent the downward and upward active power / frequency droop control gains of the distributed PV in the cluster; and respectively represent the upstream and downstream subnet sets of the subnetwork ; is a 0-1 indicator variable. 1 indicates that subnetwork m is connected to the substation, and this subnetwork receives the frequency regulation instructions issued by the transmission grid operator. Otherwise, the subnetwork only receives the frequency regulation instructions sent by adjacent subnets; represents the set of nodes with distributed photovoltaics installed in the subnetwork .

[0161] The consistency constraint is expressed as:

[0162]

[0163] ,

[0164] ,

[0165] In the formula: and respectively represent the active power and reactive power flowing through the tie line between the subnets calculated by the subnetwork and ; and respectively represent the active power and reactive power flowing through the tie line between the subnets calculated by the subnetwork and ; and respectively represent the voltages at the boundary nodes calculated by the subnetwork and the subnetwork ; and respectively represent the downward and upward frequency regulation reference instructions calculated by the subnetwork and sent to the subnetwork ; and respectively represent the downward and upward frequency regulation reference instructions calculated by the subnetwork and sent to the subnetwork ; and respectively represent the upward and downward frequency regulation participation degrees of distributed photovoltaics in subnetwork m; and respectively represent the upward and downward frequency regulation participation degrees of distributed photovoltaics calculated by the subnetwork .

[0166] Define the vector composed of tie-line power, boundary node voltage, frequency regulation reference command, and distributed PV frequency regulation participation as the coupling variable. The above-mentioned consistency constraints can be rewritten in vector form as follows:

[0167]

[0168] In the formula: represents the vector composed of variables such as tie-line power, boundary node voltage, frequency regulation reference command, and distributed PV frequency regulation participation related to the sub-network , which is called the coupling variable. Similarly, represents the vector composed of variables such as tie-line power, boundary node voltage, frequency regulation reference command, and distributed PV frequency regulation participation related to the sub-network ; and are respectively the values of the coupling variables and calculated for the sub-networks and .

[0169] The objective functions of each sub-model are as follows:

[0170] ,

[0171] In the formula: represents the objective function of the sub-model corresponding to sub-network m; , and are respectively the non-negative weight factors for active power reserve, reactive power, and network loss of distributed PV in sub-network , and satisfy ; represents the set of lines in sub-network m; Specifically, it is obtained by solving the sub-model corresponding to sub-network ; is the Lagrange multiplier vector; represents the penalty coefficient. Except for the equality constraints corresponding to the active / frequency droop control gain, the constraints in the active / frequency and reactive / voltage local droop control parameter optimization model are all included in the constraints of each sub-model. In addition, the constraints of each sub-model also include the above-mentioned consistency constraints.

[0172] Based on the alternating direction multiplier method, coordinate optimization is performed on all sub-models. The specific iterative process is as follows:

[0173] ,

[0174] In the formula: and respectively represent sub-model m and sub-model The feasible decision space is composed of the constraints of each sub-model; and represent the optimization variables of sub-model and sub-model ; the subscript represents the current time; and respectively represent the th iteration values of the coupling variables calculated by sub-model and sub-model ; the superscript represents the iteration number; and respectively represent the optimization results of sub-model and at the th iteration; and respectively represent the Lagrange multiplier vectors of sub-model and sub-model during the th iteration calculation process; represents the penalty coefficient of sub-model during the th iteration calculation process.

[0175] Define the original residual and the dual residual of sub-model after the th iteration calculation respectively as:

[0176] ,

[0177] where: is the coupling variable related to sub-network ; represents the value of the coupling variable calculated by sub-network and sent to sub-network ; and are the Lagrange multipliers calculated at the current adjacent iteration step and the previous iteration step respectively.

[0178] Then the judgment condition for whether the algorithm converges after the th iteration is expressed as:

[0179] ,

[0180] ,

[0181] ,

[0182] where: and respectively represent the residual vectors composed of the original residuals and dual residuals of all sub-models after the -th iteration; is the maximum allowable residual of the alternating direction multiplier method and can be set to 1×10 -3 ; represents the infinity norm; represents the number of sub-models, that is, the number of distributed photovoltaic clusters.

[0183] Introduce a dynamic update strategy for the penalty coefficient to reduce the number of iterations. Taking the cluster as an example, its corresponding dynamic update strategy for the penalty coefficient is specifically expressed as:

[0184] ,

[0185] In the formula: is the adjustment factor; is the balance factor; represents the two-norm.

[0186] In the global optimization layer, by combining the alternating direction multiplier method for coordinated optimization of multiple sub-models, multiple distributed photovoltaic active / frequency droop control related parameters (upward reserve, downward reserve, droop gain) and reactive power / voltage droop control related parameters (droop gain, voltage dead zone, voltage saturation threshold, reactive power reference value, etc.) can be obtained, and these parameters are sent to each distributed photovoltaic.

[0187] (5)Execute the in-cluster voltage self-correction control strategy in the local coordination layer to perform fast coordinated control on some distributed photovoltaics in the cluster with voltage over-limit. By modifying the reactive power reference value of the reactive power / voltage droop control of the distributed photovoltaics in the cluster, the reactive power output of the distributed photovoltaics is changed, thereby alleviating the problem of distribution network voltage over-limit caused by rapid source-load fluctuations.

[0188] Among them, the correction amount of the reactive power reference value of the reactive power / voltage droop control of the distributed photovoltaics in the cluster with voltage over-limit is calculated by the following formula:

[0189] ,

[0190] ,

[0191] ,

[0192] In the formula: represents the vector composed of the reactive power correction amounts of each distributed photovoltaic in the cluster; represents the steady-state voltage at node after reactive power / voltage droop control; represents the vector composed of the voltage correction amounts of each node, Represents a distributed photovoltaic cluster The number of internal nodes; Represents a node The voltage correction amount of; And Respectively represent the upper and lower limits of the distribution network voltage safety; Represents the droop control gain of each distributed photovoltaic in the cluster The vector composed of; Represents the diagonal element is The diagonal matrix of.

[0193] By sending the correction amount of the reactive power reference value of the reactive power / voltage droop control to each distributed photovoltaic in the cluster, the distributed photovoltaic will correspondingly change its reactive power output, thereby adjusting the voltage of each node in the distributed photovoltaic cluster to within the safe range.

[0194] (6) In the local control layer, based on the calculation results of the global optimization layer, configure the active / frequency and reactive / voltage droop control parameters of each distributed photovoltaic. When a frequency or voltage deviation is detected, the distributed photovoltaic will adjust its active and reactive power outputs in real time according to the droop control, thereby providing immediate frequency and voltage support. Among them, the active / frequency and reactive / voltage droop control are specifically expressed as:

[0195]

[0196]

[0197] In the formula: And Respectively represent the distributed photovoltaic At The corresponding real-time active and reactive power adjustment amounts in response to frequency and voltage deviations; Represents The system frequency measurement value at the moment; Represents The moment of node Voltage measurement value.

[0198] On the IEEE 33-node distribution network, the control scheme of the present invention and another three control schemes are selected for comparative analysis to illustrate the effectiveness of the proposed control scheme in frequency-voltage coordinated support. Among them, Schemes 1-3 are respectively:

[0199] Scheme 1: No control is applied;

[0200] Scheme 2: Update the active and reactive power set points of the distributed photovoltaic every 5 s;

[0201] Solution 3: Adopt the global optimization layer and the local control layer in the proposed control solution, but do not include the local coordination layer.

[0202] Table 1 compares the control performance of the method of the present invention with three other solutions.

[0203] Table 1 Performance comparison of different control solutions (24 hours)

[0204]

[0205] To intuitively compare the frequency support effects of different methods, the aggregated support power deficit is defined here, which represents the difference between the actual power regulation amount at the outlet of the distribution network substation and the power regulation amount expected by the transmission network. In terms of the frequency support effect, the smaller the aggregated support power deficit, the better. It can be seen from the table that there are voltage over-limit situations for a long time in Solutions 1, 2, and 3, and the method of the present invention can fully ensure the voltage safety of the distribution network. Solution 3 and the method of the present invention are optimized with a period of 1 h, while Solution 2 is optimized every 5 s. Due to the large change in the system operation state, the reactive power compensation amount and network loss of Solution 3 and the method of the present invention are higher than those of Solution 2. Solution 3 and the method of the present invention have similar effects in terms of frequency support, and the aggregated support power deficit is much smaller than that of Solution 2. In summary, in terms of the distribution network voltage regulation and the system frequency support effect, the method of the present invention performs well under various operating conditions, and the control effect is better than that of the traditional method.

[0206] Figure 2 shows the dynamic performance of different control methods. By comparing with three methods of not applying control, updating the active and reactive power set points of distributed photovoltaics every 5 s, and adopting the global optimization layer and the local control layer in the proposed control solution but not including the local coordination layer, the frequency ( Figure 2 in (a) of Figure 2 and the voltage control results ( in (b) of

[0207] are compared. The method of the present invention has better dynamic response characteristics. By further applying the in-group voltage self-correction control on the basis of Solution 3, the voltage over-limit is significantly improved, and the frequency support effect is better than those of Solutions 1 and 2, and the maximum frequency deviation is 0.1860 Hz. In summary, the method of the present invention performs more balancedly in terms of frequency and voltage support, and the overall performance is better than other solutions. The specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, which is not a limitation on the protection scope of the present invention. All equivalent models or equivalent algorithm processes made by using the content of the specification and drawings of the present invention and directly or indirectly applied to other related technical fields fall within the scope of the patent protection of the present invention.

Claims

1. A distributed photovoltaic hierarchical cluster frequency-voltage coordinated support control method, characterized in that It includes the following steps: 1) Considering the influence of distributed photovoltaic (PV) participating in system frequency regulation on the distribution network voltage, improving the electrical distance index, and clustering the distributed PV based on non-negative matrix factorization technology; 2) Constructing a hierarchical cluster coordination control scheme including a global optimization layer, a local coordination layer, and a local control layer to achieve multi-time scale coordinated control of distributed PV, so as to provide fast frequency support and alleviate the problem of distribution network voltage over-limit; 3) In the global optimization layer, aiming at minimizing the loss of distributed PV power generation benefits, the use of reactive power, and network losses, and taking the distribution network power flow equation, the operating characteristics of distributed PV, and the local droop control function as constraints, constructing an optimization model for the local droop control parameters of active power / frequency and reactive power / voltage; 4) Based on the clustering result, reconstructing the constraints related to frequency regulation in the optimization model of the local droop control parameters of active power / frequency and reactive power / voltage, and adding consistency constraints, so as to decompose the optimization model of the local droop control parameters of active power / frequency and reactive power / voltage into multiple sub-models, and coordinately optimizing all sub-models based on the alternating direction method of multipliers, thereby realizing the optimal design of the local droop control parameters of active power / frequency and reactive power / voltage for each distributed PV; 5) In the local coordination layer, for the distributed PV cluster with voltage over-limit, by modifying the reactive power reference value of the reactive power / voltage droop control of the distributed PV within the cluster, realizing the voltage self-correction control within the distributed PV cluster with voltage over-limit; 6) In the local control layer, based on the parameter optimization design result obtained in the global optimization layer, configuring the local droop control parameters of active power / frequency and reactive power / voltage for each distributed PV, and each distributed PV combines the local real-time frequency and voltage measurement values to quickly respond to frequency and voltage deviations, realizing instant frequency and voltage collaborative support.

2. A method for hierarchical cluster frequency-voltage collaborative support control of distributed PV according to claim 1, wherein in step 1), the method of considering the influence of distributed PV participating in system frequency regulation on the distribution network voltage, improving the electrical distance index, and clustering the distributed PV based on non-negative matrix factorization technology is as follows: ， ， ， ， ， ， ， ， Wherein: represents the node-edge adjacency matrix of the network, and its elements are , representing node and node the electrical connection between; represents the number of distributed photovoltaic clusters, represents the number of nodes in the distribution network; represents the set composed of the remaining nodes in the distribution network except the reference node; and are two non-negative basic matrices; node-cluster membership matrix, the row number corresponding to the maximum value of each column element in matrix is the number of the cluster to which each node belongs; represents the node similarity matrix, and its elements are , representing node and node the similarity between; represents the Frobenius norm; is a positive real parameter used to balance the contributions of topological and attribute information; The third term of the objective function is a regularization term used to control the number of nodes within each cluster. is the regularization constant; represents the matrix the th element of represents the total membership degree of each cluster, is a constant vector with all elements equal to 1; represents the distribution network node attribute matrix; and are respectively the th and the th column vectors of the distribution network node attribute matrix is a constant coefficient vector, and its element is used to quantify the impact of the injection power adjustment of node on the power at the substation outlet; represents the active-voltage sensitivity matrix, is its th row and th column element; represents the reactive-voltage sensitivity; represents taking the diagonal elements of the matrix and forming a column vector; represents the electrical distance index between node and ; represents the improved electrical distance index; and respectively represent the impact of injecting a unit power of distributed PV located at node on the voltages of node and node ; and respectively represent the impact of injecting a unit power of distributed PV located at node on the voltages of node and node ; represents the installed capacity of distributed PV at node ; and respectively represent the th and the th diagonal elements; represents the th row and th column element of ; 3. A method for hierarchical cluster frequency-voltage collaborative support control of distributed PV according to claim 1, wherein in step 2), the specific method for realizing multi-time scale coordinated control of distributed PV based on the hierarchical cluster coordination control scheme is as follows: The hierarchical cluster coordinated control scheme includes three control levels: the global optimization layer, the local coordination layer, and the local control layer. The global optimization layer performs distributed optimization once per hour to optimize the active / frequency and reactive / voltage droop control parameters of distributed photovoltaics. The local coordination layer executes voltage self-correction control within the cluster on a minute-scale time horizon by modifying the reactive power reference value of the reactive / voltage droop control of distributed photovoltaics within the cluster to regulate the reactive power output of distributed photovoltaics within the cluster, thereby alleviating the over-limit of the distribution network voltage. In the local control layer, based on the optimization design results of the global optimization layer, the active / frequency and reactive / voltage droop control parameters of each distributed photovoltaic are configured. The distributed photovoltaic combines local frequency and voltage measurement values and responds to system frequency and voltage deviations in real time. Through the mutual cooperation of the three control levels, coordinated control of distributed photovoltaics at multiple time scales is achieved.

4. A method for hierarchical cluster frequency-voltage coordinated support control of distributed photovoltaics according to claim 1, wherein In step 3), in the optimization model of the local active / frequency and reactive / voltage droop control parameters, the objective function is specifically expressed as: ， In the formula: is the objective function of the active / frequency and reactive / voltage local droop control parameter optimization model; , and are all non - negative weight factors regarding the active reserve, reactive power and network loss of distributed photovoltaic, and satisfy ; The subscript represents the time; and respectively represent the upward and downward reserve powers of node at time is the inverter reactive power adjustment amount of node at time represents the square of the current amplitude on line ; represents the resistance of line ; represents the set of distribution network lines; represents the set of nodes with distributed photovoltaic installed in the entire distribution network; is the optimization time domain; The constraints of the distribution network power flow equation are specifically: ， ， ， ， ， ， Wherein: and respectively represent the active and reactive power flows on each line ; represents the set of downstream nodes that are connected to node by lines; and respectively represent the active and reactive powers injected into node ; represents the square of the voltage magnitude of node , where is the voltage magnitude of node represents the reactance of line ; and respectively represent the upper and lower limits of the distribution network voltage safety; represents the set composed of all nodes in the distribution network except the reference node.

5. A method for hierarchical cluster frequency-voltage coordinated support control of distributed photovoltaics according to claim 4, wherein In step 4), based on the cluster division result, the constraints related to frequency regulation in the optimization model of the local active / frequency and reactive / voltage droop control parameters are reconstructed, and a consistency constraint is added. The specific method is: If there is a common line between two clusters and , then a virtual node is introduced in the middle and the line is disconnected at this virtual node; the entire network is divided into subnets , and ; each subnet has a connection line and a boundary node; in distributed optimization, each of the subnets and has an optimization model, called a sub-model; and each have an optimization model, called a sub-model The constraints related to frequency regulation obtained after reconstruction are specifically expressed as: ， ， Wherein: and respectively represent the downward and upward frequency modulation reference commands issued by the transmission grid operator at time and respectively represent the downward and upward frequency modulation reference commands sent from subnet to subnet at time and respectively represent the downward and upward frequency modulation reference commands sent from subnet to subnet at time represents the linearization constant coefficient corresponding to subnet which is calculated separately by each subnet; and respectively represent the downward and upward active / power frequency droop control gains of distributed PVs in the cluster; and respectively represent the upstream and downstream subnet sets of subnet is a 0-1 indicator variable, 1 indicates that subnet is connected to the substation and receives the frequency modulation command issued by the transmission grid operator, otherwise subnet only receives the frequency modulation commands sent by adjacent subnets; represents the set of nodes installed with distributed PVs in subnet The added consistency constraint specifically includes: ， ， ， Wherein: and respectively represent at time the active power and reactive power flowing through the tie line between the subnets and calculated by the subnets; and respectively represent at time the active power and reactive power flowing through the tie line between the subnets and calculated by the subnets; and are respectively at time the voltages at the boundary nodes calculated by the subnets and the subnets; and respectively represent at time the down - and up - frequency modulation reference commands calculated by the subnets and sent to the subnets; and respectively represent at time the down - and up - frequency modulation reference commands calculated by the subnets and sent to the subnets; and respectively represent the up - and down - frequency modulation participation degrees of distributed photovoltaics calculated by the subnet m; and respectively represent the up - and down - frequency modulation participation degrees of distributed photovoltaics calculated by the subnets ; The above consistency constraint is further expressed in the following vector form: ， Wherein: is a coupling variable related to the subnet ; similarly, is a coupling variable related to the subnet ; and are respectively the values of the coupling variables and calculated by the subnet and .

6. A method for hierarchical cluster frequency-voltage coordinated support control of distributed photovoltaics according to claim 5, wherein In step 4), the objective function of each sub-model is: ， In the formula: represents a subnet the objective function of the corresponding sub-model; , and are both non-negative weight factors of the subnet with respect to distributed photovoltaic active reserve, reactive power and network loss, and satisfy ; represents the set of lines within subnet m; is the Lagrange multiplier vector; represents the penalty coefficient.

7. A method for hierarchical cluster frequency-voltage coordinated support control of distributed photovoltaics according to claim 6, wherein The coordinated optimization of all sub-models based on the alternating direction multiplier method has the following specific iterative process: ， Where: and respectively represent the feasible decision spaces of sub - model m and sub - model ; and represent the optimization variables of sub - model and sub - model ; The subscript represents the current moment; and respectively represent the coupled variable values calculated by the -th iteration of sub - model and sub - model ; and respectively represent the optimization results obtained by sub - model and in the -th iteration; and respectively represent the Lagrange multiplier vectors of sub - model and sub - model and sub - model in the calculation process of the -th iteration; represents the penalty coefficient of sub - model in the calculation process of the Define the original residuals and dual residuals of the sub-model after the -th iteration calculation as follows: ， In the formula: is the coupling variable related to the subnet ; represents the value of the coupling variable calculated by the subnet and sent to the subnet ; and are the Lagrange multipliers calculated in the current iteration step and the previous iteration step respectively; Then the judgment condition for whether the algorithm converges after the th iteration is expressed as: ， ， ， In the formula: and respectively represent the residual vectors composed of the original residuals and dual residuals of all sub-models after the -th iteration; is the maximum allowable residual of the alternating direction multiplier method; represents the infinity norm; represents the number of sub-models, that is, the number of distributed photovoltaic clusters; A penalty coefficient dynamic update strategy is introduced, which is specifically expressed as: ， Wherein: is an adjustment factor; is a balance factor; represents the two-norm.

8. A distributed photovoltaic hierarchical cluster frequency-voltage coordinated support control method according to claim 1, characterized in that In step 5), in the local coordination layer, the method for implementing voltage self-correction control by modifying the reactive power reference value of the reactive / voltage droop control of some distributed photovoltaics within the cluster is specifically: If a distributed photovoltaic cluster detects that there is a voltage over-limit in the internal node, the voltage correction amount is calculated by the following formula: ， ， Wherein: represents the vector formed by the voltage correction amounts of each node ; and respectively represent the upper and lower limits of the voltage safety of the distribution network; represents the steady-state voltage at node after reactive power / voltage droop control; represents the number of nodes in the distributed photovoltaic cluster ; Calculate the correction amount of the reactive power reference value of the reactive / voltage droop control of each distributed photovoltaic according to the following formula: ， In the formula: represents the vector composed of the distributed photovoltaic reactive power correction amounts within the cluster; represents the reactive power-voltage sensitivity; represents the distributed photovoltaic droop control gains within the cluster that form a vector; represents a diagonal matrix with diagonal elements of ; By calculating the correction amount of the reactive power reference value for the distributed PV reactive power / voltage droop control and sending it to each distributed PV within the cluster to modify the reactive power reference value of the distributed PV reactive power / voltage droop control.

9. A computer-readable storage medium having computer instructions stored thereon, characterized in that, The computer instructions are used to cause a computer to execute the steps of the method according to any one of claims 1-8.

10. An electronic device, characterized in that, Including: One or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1-8.

Images