A Distributed Control Method for Distributed Generators in Distribution Networks Based on Optimal Spectral Clustering

Abstract

This invention proposes a distributed control method for distributed generation in a distribution network based on optimal spectral clustering. It employs an active power flow-based spectral clustering method to divide the distribution network into small clusters according to distributed photovoltaic (PV) power, generating balanced clusters and identifying the error between the reference net load and the measured net load. Then, a controller is assigned to each cluster to manage the distributed generation within the cluster. Based on an error minimization method, the DER (Distributed Generation Controller) scheduling dynamically compensates for local net load changes. This controller manages the distributed generation within the cluster to reduce the intermittency of the net load. To address constantly changing grid conditions, the clusters dynamically change to determine a new optimal cluster configuration. Through these dynamically changing clusters, the network load balance is further improved.

Description

Technical Field

[0001] This invention relates to the field of distributed collaborative control technology for power distribution networks, and in particular to a distributed control method for distributed power sources in power distribution networks based on optimal spectral clustering. Background Technology

[0002] The increasing number of active devices in power distribution systems and their control make power grid management extremely complex. Furthermore, load variations in unbalanced power distribution systems are uneven.

[0003] One of the traditional methods for controlling and managing distributed generation (DG) units is a centralized approach. However, with a significant increase in the number of DG units, a centralized approach may become infeasible. The main factors contributing to the infeasibility of centralized control include: 1) the lack of a committed management unit; 2) computational stress due to the large number of controllable assets (such as DG units and loads); 3) frequent reconfiguration requirements, as even a change in a single entity can affect the main controller; 4) the sensitivity and reliability of the main controller as a common point of failure; 5) difficulties arising from a lack of methods or unwillingness to share data; and 6) the uneven distribution and variation of loads in the distribution network.

[0004] Therefore, centralized control modules for controllable elements would be complex, inefficient, and suboptimal when managing each local part of the power grid. However, if distributed energy clusters could be developed into clusters to distribute local control, it would be possible to effectively manage individual areas within the distribution system. Summary of the Invention

[0005] To address the shortcomings and deficiencies of existing technologies, this invention proposes a distributed control method for distributed power sources in power distribution networks based on optimal spectral clustering.

[0006] The present invention specifically adopts the following technical solution:

[0007] A distributed control method for distributed generation in a distribution network based on optimal spectral clustering is proposed. This method employs an active power flow-based spectral clustering approach to divide the distribution network into small clusters based on distributed photovoltaic (PV) power, generating balanced clusters and identifying the error between the reference net load and the measured net load. Then, a controller is assigned to each cluster to manage the distributed generation within the cluster. Based on an error minimization method, the DER (Distributed Generation Controller) scheduling dynamically compensates for local net load variations. This controller manages the distributed generation within the cluster to reduce the intermittency of the net load. To address constantly changing grid conditions, the clusters dynamically change to determine a new optimal cluster configuration. Through these dynamically changing clusters, the network load balance is further improved.

[0008] Furthermore, the specific steps include:

[0009] Step S1: Use an undirected graph G = (V; E) to represent the multiphase power distribution network. The vertices (V) of the graph represent lines, and the edges (E) represent the connections between vertices. Model the power distribution network as a graph G = (V; E). The vertices (V) of the graph represent the network bus, and the edges (E) represent the connections between vertices.

[0010] Step S2: Calculate the weights of edges based on the original active power flow of the distribution network lines, take the relative tilt angle caused by the power flow through the distribution network as weak connection points, partition the undirected graph according to the weights of the connections between graph nodes, and create a virtual cluster.

[0011] Step S3: As the distributed energy source operates or the load changes, the power flow of the distribution network is periodically calculated and partitioned, and then spectral clustering is periodically performed to dynamically re-cluster and identify new clusters.

[0012] Step S4: Calculate the ratio of the boundary to the volume of each cluster, measure the quality of each cluster, and evaluate the clustering quality.

[0013] Step S5: Integrated cluster formation. The net load of the branch and the critical load of the branch are measured by the intelligent inverter. A dynamic distribution model of the power distribution system caused by the change of distributed photovoltaic output is established. Based on the distribution model, the error between the reference net load and the measured net load of different clusters is calculated.

[0014] Step S6: Apply the linear least squares optimization method to optimize the error between the cluster's reference net load and the measured net load. Distributed energy scheduling dynamically compensates for local net load changes based on the error minimization method.

[0015] Furthermore, step S2 specifically includes the following steps:

[0016] Step S201, in a multiphase power distribution network, the weight of the edge or line of the x-th phase bus connecting the i-th and j-th vertices, and the power flowing between the buses, are expressed as:

[0017]

[0018] In the formula P ij Let x represent the power flowing between vertices i and j, where x = 1, 2, ..., φ;

[0019] From the above formula, we obtain the weighted adjacency matrix A of the generalized N-bus network. ij Represented as:

[0020]

[0021] In the formula, i = 1, 2, 3, ... B, j = 1, 2, 3, ... B, A ij It is a measure of the power flow between vertex i and vertex j;

[0022] Step S202: Use line weights as a penalty for cutting corresponding lines during clustering, and also as a measure of connection strength;

[0023] Let the degree matrix (D) be a diagonal matrix, where each diagonal element (D) ii D 22 ...D BB The degree is represented by ); the degree of a vertex or bus is calculated by summing the weight strengths of all edges attached to that particular vertex; then, the overall strength or degree of each B bus is expressed as:

[0024]

[0025] The above formula describes the injected power of each bus in the distribution network; j = 1, 2, 3…B;

[0026] Step S203: Establish the Laplace matrix L of the multiphase distribution network. The Laplace matrix L is a B×B matrix.

[0027]

[0028] L represents the Laplace matrix of the power distribution system, L ij The element in the i-th row and j-th column is represented; the edge set E represents the connection between vertices.

[0029] Normalized Laplace matrix

[0030]

[0031] The number of connected components is determined by observing the eigenvalues ​​of the Laplace matrix with zero values; to achieve clustering objectives, a normalized Laplace matrix is ​​used for the randomly weighted network in terms of cluster solution quality; in its simplest form, the normalized Laplace (L... n The matrix is ​​derived from the following formula:

[0032]

[0033] The branch power calculated by the power distribution management system is used as the weight for spectral clustering.

[0034] Furthermore, in step S3, when photovoltaic power generation experiences sudden changes or power fluctuations, the clustering algorithm identifies the changes and re-clusters the data; specifically, time-based periodic clustering is employed; each time a cluster is formed, a clustering quality check is performed based on the optimal dimension of the cluster embedding. The calculation method for the cluster embedding is as follows:

[0035]

[0036] in The maximum value determines the optimal number of clusters k;

[0037] Where k represents the optimal number of clusters, calculated based on the highest density index and location. The higher the relative characteristics, the lower the tie-line weight or power flow (inter-cluster power), and the better the partitioning;

[0038] Let's say there's a cluster C t =(V t E t As a sub-cluster of the original distributed network G=(V,E), C t ∈G,V t ∈V,E t ∈E, the vertices (V) of the graph represent the network bus, and the edges (E) represent the connections between vertices; the quality of a partition is measured by the volume of the cluster and the size of its boundary; cluster C t The volume is calculated as follows:

[0039]

[0040] Computing cluster C t Boundary:

[0041]

[0042] The overall quality of a cluster is measured by the ratio of its boundary to its volume.

[0043]

[0044] Furthermore, the power flow model of the distribution network in step S4 is expressed as follows:

[0045] V j (t) represents the voltage at node j at time t, Z ij P is the impedance between nodes i and j. ij (t), Q ij (t) represents the active power and reactive power flowing between nodes i and j. Let be the voltage value at node i at time t;

[0046] The power flowing through these lines is expressed as:

[0047]

[0048]

[0049] Therefore, the current flow model is simplified to:

[0050]

[0051] Where l represents the square of the current in the branch;

[0052] Considering the systematic distribution of all non-boundary nodes in the cluster: R >> X, it simplifies to:

[0053]

[0054] Where V j (t) represents the voltage at node j at time t, Z ij P is the impedance between nodes i and j. ij (t), Q ij (t) represents the active power and reactive power flowing between nodes i and j. Let be the voltage value at node i at time t; l represents the square of the current in the branch.

[0055] Furthermore, in step S5, the power flow model equations are expressed in Laplace form:

[0056]

[0057] In the formula, j∈v x(nb) , i∈v x(nb) ;v x(nb) It is the set of all non-boundary nodes in cluster K; for all boundary nodes connecting cluster K and cluster M, the voltage equation is:

[0058] Based on the voltage equation, the above expression can be represented by the cluster boundary as follows:

[0059]

[0060] In the formula, b∈v x(B) , and v x(B) Let K be the set of all boundary nodes in clusters K and m. The above equation can then be rewritten as:

[0061]

[0062]

[0063] Since voltage variations are based on line load, Where γ(k) represents the line load factor; further considering the formation of stable clusters; therefore Represented as α*P set (t), so as to minimize the deviation between clusters; therefore, the above formula can be expressed as:

[0064]

[0065] in

[0066]

[0067] in

[0068] A comprehensive cluster is formed, and a dynamic distribution model of the power distribution system caused by changes in distributed photovoltaic output is established.

[0069] Furthermore, step S6 specifically includes the following steps:

[0070] Step S601: Calculate the cluster scheduling point for the next time period by minimizing the squared error. The cluster scheduling point assigns a controller node to each cluster to manage the distributed power supply within the cluster, as shown in the following formula. In N measurements, the linear least squares optimization method is applied to minimize the net load of the k-th cluster. Compared with the measured net load (P) nl The error between )

[0071]

[0072] In the formula, Υ(k) is the optimization coefficient; For the net load of the k-th cluster, P nl This is the actual measured net load;

[0073] Step S602: The net load support provided by the BESS within the cluster is limited by the rated active power of the BESS within the cluster; for m BESS within the cluster, the total active power available to cluster k is... The upper and lower bounds of the optimization coefficient γ(k) are expressed as follows;

[0074]

[0075] Among them, P kW (k) represents the total active power support available for cluster k, P nl For net load, the upper and lower limits restrict the support capacity of each cluster based on the total kW rating of all BESSs within the cluster; for Y(k), the optimal active support of cluster k by m BESSs is calculated as follows:

[0076]

[0077] P BESS (k, n) represents the total active power allocated to the BESS of the k-th cluster, and T(k) is the optimization coefficient. For the net load of the k-th cluster, P nlThis represents the actual net load; the scheduling of a single BESS is allocated among the BESS cluster in descending order of SoC error and BESS capacity:

[0078] e soc (j, n) = SOC t (j, n)-SOC(j, n)

[0079] ekWh soc (j, n) = e soc (j, n)*BESS kwh

[0080] SOC t (j, n) represents the recovered value of the SoC, e soc (j, n) represents the error value of soc;

[0081] Step S603: Calculate the active power distribution point of a single BESS using a proportional-integral method. To improve the recovery of the target SoC (System-on-a-Chip). t (j, n)), and keep the BESS SoC within the upper and lower limits;

[0082] For each BESS within the cluster, based on e soc Calculate the slope based on the (j, n) values. As shown in the following formula; when the SoC error esoc is less than the error threshold When the SoC error exceeds the error threshold, the BESS scheduling is controlled based on the integral component of the error of the last N measurements. The Ki value is further dynamically changed according to the charging and discharging opportunities to improve the recovery of the SoC.

[0083]

[0084] in is the upper limit threshold of the SOC error value, kp is the proportional component, and the slope coefficient of the proportional component is based on kp to control sudden changes in BESS scheduling;

[0085] based on The active power scheduling value for each BESS within the cluster is calculated sequentially as shown in the following formula:

[0086]

[0087] in This represents the active power scheduling value for each BESS within the cluster. For e-based soc The (j,n) value is used to calculate the slope of sudden changes in the control BESS scheduling, P. BESS(k,n) represents the charging and discharging power of bess;

[0088] Subject to BESS kW limit The restrictions are shown in the following formula:

[0089]

[0090] And an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, when the processor executes the program, it implements the steps of the distributed control method for distributed power sources in a power distribution network based on optimal spectral clustering as described above.

[0091] A non-transitory computer-readable storage medium storing a computer program thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the distributed control method for distributed power sources in a power distribution network based on optimal spectral clustering as described above.

[0092] Compared to existing technologies, this invention and its preferred embodiment employ an improved spectral clustering method based on active power flow to divide the distribution network into small clusters according to distributed photovoltaic (PV) power. The proposed spectral clustering method can accurately generate balanced clusters and identify the error between the reference net load and the measured net load. Then, a controller is assigned to each cluster to manage the distributed power sources within the cluster. Based on an error minimization method, the DER scheduling dynamically compensates for local net load variations. This controller manages the distributed power sources within the cluster to reduce the intermittency of the net load. To address constantly changing grid conditions, the clusters dynamically change to determine new optimal cluster configurations. Through dynamically changing clusters, network load balancing is further improved. Attached Figure Description

[0093] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0094] Figure 1 This is a flowchart of a distributed control method for distributed power sources in a power distribution network based on optimal spectral clustering, according to an embodiment of the present invention.

[0095] Figure 2 This is a schematic diagram of the distributed control architecture according to an embodiment of the present invention. Detailed Implementation

[0096] In the following, specific embodiments of this application will be described in detail with reference to the accompanying drawings. Based on these detailed descriptions, those skilled in the art will be able to clearly understand and implement this application. Without departing from the principles of this application, features from various embodiments can be combined to obtain new implementations, or certain features from some embodiments can be substituted to obtain other preferred implementations.

[0097] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0098] To make the features and advantages of this patent more apparent and understandable, specific embodiments are provided below for detailed explanation:

[0099] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0100] like Figure 1 , Figure 2 As shown, this embodiment of the invention provides a distributed control system for distributed generation in a distribution network based on optimal spectral clustering, and provides a distributed dynamic model of a grid with stable and scalable measurement values. Based on an error minimization method, the controller dynamically compensates for local net load changes in DER scheduling, managing distributed generation within the cluster to reduce the intermittency of net load. Specific implementation steps include:

[0101] A. Use an undirected graph G = (V; E) to represent a multiphase power distribution network. The vertices (V) of the graph represent the network bus, and the edges (E) represent the connections between vertices.

[0102] B. Calculate the weight of the edges based on the original active power flow of the distribution network lines, take the "relative tilt angle" caused by the power flow through the distribution network as the weak connection point, partition the undirected graph according to the weight of the connection between the graph nodes, and create a virtual cluster.

[0103] C. As the DER (Distributed Energy) operates or the load changes, the power flow of the distribution network is periodically calculated and partitioned, and then spectral clustering is periodically performed to dynamically re-cluster and identify new clusters.

[0104] D. Calculate the ratio of the boundary to the volume of each cluster to measure the quality of each cluster and evaluate the clustering quality.

[0105] E. Integrated cluster formation: By measuring the net load of the branch and the critical load of the branch through intelligent inverters, a dynamic distribution model of the power distribution system caused by the changes in distributed photovoltaic output is established. Based on the distribution model, the error between the reference net load and the measured net load of different clusters is calculated.

[0106] F. The linear least squares optimization method is applied to optimize the error between the reference net load and the measured net load of the cluster. DER (Distributed Energy) scheduling dynamically compensates for local net load changes based on the error minimization method.

[0107] In this preferred embodiment, spectral clustering in step A is a partitioning algorithm based on an undirected graph. It treats data points as graph nodes and partitions the graph according to the weights of connections between these nodes. In this method, the power distribution network is modeled as a graph G = (V; E). Vertices (V) represent network buses, and edges (E) represent connections between vertices.

[0108] In step B, the weight of the φ-th phase busbar connecting two vertices i and j in the multiphase distribution network can be expressed as the power flowing between the two buses:

[0109]

[0110] In the formula P ij This represents the power flowing between vertices i and j. X = 1, 2, 3…φ

[0111] From the above equation, we can obtain the weighted adjacency matrix A of the generalized N-bus network. ij It can be represented as:

[0112]

[0113] In the formula, i = 1, 2, 3, ... B, j = 1, 2, 3, ... B, A ij It is a measure of the power flow between vertex i and vertex j.

[0114] In step B, in the graph depicting the electrical network, the edge weights relate to the strength of the electrical connection between two buses. The weight of a line connecting two vertices measures the importance of that line under given operating conditions. Line weights can be interpreted as a penalty for cutting off the corresponding line during clustering, but they can also serve as a measure of connection strength.

[0115] In step B, degree is one of the most important parameters for measuring cluster quality. According to graph theory, the degree matrix (D) is a diagonal matrix, where each diagonal element (D) ii D 22 ...D BB The degree is represented by ). The degree of a vertex or bus can be calculated by summing the weight strengths of all edges attached to that particular vertex. The overall strength or degree of each B-bus can then be expressed as:

[0116]

[0117] The above formula describes the injected power of each bus in the distribution network.

[0118] In step B, the Laplace matrix (L) of the power distribution system can be represented by L, which is a B×B matrix and can be calculated as follows:

[0119]

[0120] L represents the Laplace matrix of the power distribution system, L ij Let E represent the element in the i-th row and j-th column. The edge set E represents the connections between vertices.

[0121] By performing corresponding mathematical operations on the order and the Laplace matrix, the normalized Laplace matrix can be derived:

[0122]

[0123] The number of connected components can be determined by observing the eigenvalues ​​of the Laplace matrix with zero values. To achieve clustering, a normalized Laplace matrix is ​​used for the randomly weighted network to improve the quality of the clustering solution. In its simplest form, the normalized Laplace matrix (L...) n The matrix can be derived from the following formula:

[0124]

[0125] According to graph theory, L represents the Laplace matrix of the power distribution system, and the degree matrix (D) is one of its diagonal matrices.

[0126] Higher relative eigenvalues ​​and lower eigenvalues ​​define good clustering quality. Higher relative eigenvalues ​​or lower eigenvalues, along with smaller boundary values, indicate better clustering quality. Distributed generation grouping (clustering) is based on power flow in the distribution network, minimizing line power (load generation) between clusters. In power flow management based on Advanced Distribution Management Systems (ADMS), branch power calculated by ADMS can be used as weights for spectral clustering.

[0127] In step C, when PV (photovoltaic power generation) experiences sudden changes or power fluctuations, the clustering algorithm identifies the changes and re-clusters the data. This invention designs time-based clustering (periodically performed clustering). Each time a cluster is formed, a clustering quality check is performed based on the optimal dimension of the cluster embedding. The calculation method for the cluster embedding is as follows:

[0128]

[0129] in The maximum value determines the optimal number of clusters k.

[0130] Where k represents the optimal number of clusters, which is calculated based on the highest density index and location. The higher the relative characteristics, the lower the tie-line weight or power flow (inter-cluster power), and the better the partitioning.

[0131] In step D, assume there is a cluster C t =(V t E t It is also a sub-cluster (i.e., C) of the original distributed network G=(V,E). t ∈G,V t ∈V,E t In a graph (V ∈ E), vertices (V) represent the network bus, and edges (E) represent connections between vertices. The quality of a partition is measured by the cluster volume and its boundary size. Cluster C t Volume (vol(C) t This can be calculated as:

[0132]

[0133] Computing cluster C t Boundary:

[0134]

[0135] The overall quality of the cluster is measured by the ratio of its boundary to its volume, known as the expansion value.

[0136]

[0137] If the extended value C t A smaller value indicates better cluster quality. This means minimal distribution power flow, suggesting a weak connection between the cluster and the remaining grid. It's important to note that providing good cluster capacity and low power consumption between clusters is crucial for a high-quality cluster.

[0138] The power flow model of the distribution network in step E can be expressed as:

[0139]

[0140] V j (t) represents the voltage at node j at time t, Z ij P is the impedance between nodes i and j. ij (t), Q ij (t) represents the active power and reactive power flowing between nodes i and j. Let be the voltage value at node i at time t.

[0141] Similarly, the power flowing through these lines can be expressed as:

[0142]

[0143]

[0144] The above power flow model can be simplified as follows:

[0145]

[0146] Where l represents the square of the current in the branch.

[0147] Considering the distributed system properties (R >> X) of all non-boundary nodes in the cluster, the above equation simplifies to:

[0148]

[0149] Where V j (t) represents the voltage at node j at time t, Z ij P is the impedance between nodes i and j. ij (t), Q ij (t) represents the active power and reactive power flowing between nodes i and j. Let be the voltage value at node i at time t. Let l represent the square of the current in the branch.

[0150] In step E, the power flow model equations can be expressed using the Laplace equation:

[0151]

[0152] In the formula, j∈v x(nb) , i∈v x(nb) ;v x(nb) It is the set of all non-boundary nodes in cluster K.

[0153] For all boundary nodes connecting cluster K and cluster M, the voltage equation is:

[0154]

[0155] According to the voltage equation, the above equation can be expressed using cluster boundaries as follows:

[0156]

[0157] In the formula, b∈v x(B) , and v x(B) Let K be the set of all boundary nodes in clusters K and m. Considering the subsequent time step t+1 and the difference in cluster quality (low inter-cluster power), the above formula can be written as:

[0158]

[0159]

[0160] Voltage variations are based on line load. Where γ(k) represents the line load factor. Furthermore, considering the formation of stable clusters, the expansion coefficient is relatively small. Therefore... It can be represented as α*P set (t) is used to minimize the inter-cluster bias. Therefore, the above equation can be expressed as:

[0161]

[0162] in

[0163]

[0164] in

[0165] Therefore, by taking into account the formation of clusters, a dynamic distribution model of the power distribution system caused by changes in distributed photovoltaic output can be established.

[0166] In step F, the goal of the spectral clustering-based distributed control method is to utilize distributed power sources to support the grid and maximize photovoltaic penetration. Therefore, a distributed control method is adopted, leveraging the control capability characteristics of the DERS based on smart inverters to monitor and control the operation and output of the DERS. The goal of the cluster control module is to support locally connected loads and reduce the possibility of interference caused by PV interruptions and load fluctuations within and across clusters.

[0167] In step F, this invention calculates the cluster scheduling point for the next time period by minimizing the squared error. The cluster scheduling point assigns a controller node to each cluster to manage the distributed power supply within the cluster, as shown in the following formula. In N measurements, the linear least squares optimization method is applied to minimize the net load of the k-th cluster. The error between the measured net load (Pnl) and the actual net load.

[0168]

[0169] In the formula, γ(k) is the optimization coefficient. For the net load of the k-th cluster, P nl This represents the actual measured net load.

[0170] In step F, the net load support provided by the BESS within the cluster is limited by the active power rating of the BESS within the cluster. For m BESS within the cluster, the total active power support available to cluster k is... The upper and lower bounds of the optimization coefficient Υ(k) can be expressed as follows.

[0171]

[0172] Among them, P kW (k) represents the total active power support available for cluster k, P nl For net load, the upper and lower limits restrict the support capacity of each cluster based on the total kW rating of all bess within the cluster. For γ(k), the optimal active support of cluster k by m bess can be calculated as:

[0173]

[0174] P BESS (k, n) represents the total active power allocated to the BESS of the k-th cluster, and Υ(k) is the optimization coefficient. For the net load of the k-th cluster, P nl This represents the actual net load. The scheduling of a single BESS is allocated within the cluster of BESS instances in descending order of SoC error and BESS capacity.

[0175] e soc (j, n) = SOC t (j, n)-SOC(j, n)

[0176] ekWh soc (j, n) = e soc (j, n)*BESS kwh

[0177] SOC t (j, n) represents the recovered value of the SoC, e soc (j, n) represents the error value of soc.

[0178] In step F, the active power distribution point of a single BESS is calculated using a proportional-integral method. This method aims to improve the recovery of the target SoC (System-on-a-Chip). t (j, n)), and effectively keep the BESS SoC within the upper and lower limits.

[0179] For each BESS within the cluster, based on e soc Calculate the slope based on the (j, n) values. As shown in the following formula. When the SoC error (esoc) is less than the error threshold... At this time, the proportional component slope coefficient is used to control sudden changes in BESS scheduling based on the proportional component (kp). When the SoC error exceeds the error threshold, BESS scheduling is controlled based on the integral component of the error from the last N measurements. The Ki value can be further dynamically changed according to charging and discharging opportunities to improve SoC recovery.

[0180]

[0181] in is the upper limit threshold of the SOC error value, kp is the proportional component, and the slope coefficient of the proportional component is based on kp to control sudden changes in BESS scheduling.

[0182] based on The active power scheduling value for each BESS within the cluster is calculated sequentially as shown in the following formula.

[0183]

[0184] in This represents the active power scheduling value for each BESS within the cluster. For e-based soc The (j, n) value is used to calculate the slope of sudden changes in the control BESS scheduling, P. BESS (k, n) represents the charging and discharging power of the bess.

[0185] Subject to BESS kW limit The restrictions are shown in the following formula:

[0186]

[0187] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and these variations still fall within the protection scope of the present invention.

[0188] The system and method provided in this embodiment can be stored in a computer-readable storage medium in the form of code, implemented as a computer program, and the basic parameter information required for calculation can be input through computer hardware, and the calculation results can be output.

[0189] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0190] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0191] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0192] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0193] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

[0194] This patent is not limited to the above-described preferred embodiment. Anyone can derive other forms of distributed control methods for distributed power sources in power distribution networks based on optimal spectral clustering under the guidance of this patent. All equivalent changes and modifications made within the scope of this patent application shall fall within the scope of this patent.

Claims

1. A distributed control method for distributed generation in a distribution network based on optimal spectral clustering, characterized in that: Includes the following steps: Step S1: Represent the multiphase power distribution network using an undirected graph G = (V; E), where the vertices V of the graph represent lines and the edges E represent connections between vertices; Model the power distribution network as a graph G = (V; E); where the vertices V of the graph represent network buses and the edges E represent connections between vertices. Step S2: Calculate the weights of edges based on the original active power flow of the distribution network lines, take the relative tilt angle caused by the power flow through the distribution network as weak connection points, partition the undirected graph according to the weights of the connections between graph nodes, and create a virtual cluster. Step S3: As the distributed energy source operates or the load changes, the power flow of the distribution network is periodically calculated and partitioned, and then spectral clustering is periodically performed to dynamically re-cluster and identify new clusters. Step S4: Calculate the ratio of the boundary to the volume of each cluster, measure the quality of each cluster, and evaluate the clustering quality. Step S5: Integrated cluster formation. The net load of the branch and the critical load of the branch are measured by the intelligent inverter. A dynamic distribution model of the power distribution system caused by the change of distributed photovoltaic output is established. Based on the distribution model, the error between the reference net load and the measured net load of different clusters is calculated. Step S6: Apply the linear least squares optimization method to optimize the error between the cluster's reference net load and the measured net load. Distributed energy scheduling dynamically compensates for local net load changes based on the error minimization method. In step S3, when photovoltaic power generation experiences sudden changes or power fluctuations, the clustering algorithm identifies the changes and re-clusters the data. Specifically, time-based periodic clustering is employed. Each time a cluster is formed, a clustering quality check is performed based on the optimal dimension of the cluster embedding. The calculation method for the cluster embedding is as follows: in The maximum value determines the optimal number of clusters k; Where k represents the optimal number of clusters, calculated based on the highest density index and location. The higher the relative characteristics, the lower the tie-line weight or inter-cluster power flow, and the better the partitioning; Let's say there's a cluster C t =(V t E t As a sub-cluster of the original distributed network G=(V,E), C t ∈G,V t ∈V,E t ∈E, where vertices V of the graph represent the network bus and edges E represent connections between vertices; the quality of a partition is measured by the volume of the cluster and the size of its boundary; cluster C t The volume is calculated as follows: Computing cluster C t Boundary: The overall quality of a cluster is measured by the ratio of its boundary to its volume. Among them, W ij D represents the weight of the edge or line connecting the i-th and j-th phase bus in a multiphase distribution network. ii This indicates the overall strength or degree of each B bus.

2. The distributed control method for distributed generation in a distribution network based on optimal spectral clustering according to claim 1, characterized in that: Step S2 specifically includes the following steps: Step S201, in a multiphase power distribution network, the weight of the edge or line of the x-th phase bus connecting the i-th and j-th vertices, and the power flowing between the buses, are expressed as: In the formula P ij Let x represent the power flowing between vertices i and j, where x = 1, 2, ..., φ; From the above formula, we obtain the weighted adjacency matrix A of the generalized N-bus network. ij Represented as: In the formula, i = 1, 2, 3, ... B, j = 1, 2, 3, ... B, A ij It is a measure of the power flow between vertex i and vertex j; Step S202: Use line weights as a penalty for cutting corresponding lines during clustering, and also as a measure of connection strength; Let the degree matrix D be a diagonal matrix, where each diagonal element D ii D 22 ...D BB Degree; the degree of a vertex or bus is calculated by summing the weight strengths of all edges attached to that particular vertex; then, the overall strength or degree of each B bus is expressed as: The above formula describes the injected power of each bus in the distribution network; j = 1, 2, 3…B; Step S203: Establish the Laplace matrix L of the multiphase distribution network. The Laplace matrix L is a B×B matrix. L represents the Laplace matrix of the power distribution system, L ij The element in the i-th row and j-th column is represented; the edge set E represents the connection between vertices. Normalized Laplace matrix The number of connected components is determined by observing the eigenvalues ​​of the Laplace matrix with zero values; to achieve clustering objectives, a normalized Laplace matrix is ​​used for the randomly weighted network in terms of cluster solution quality; in its simplest form, the normalized Laplace matrix L... n The matrix is ​​derived from the following formula: The branch power calculated by the power distribution management system is used as the weight for spectral clustering.

3. The distributed control method for distributed generation in a distribution network based on optimal spectral clustering according to claim 1, characterized in that: The power flow model of the distribution network in step S4 is represented as follows: V j (t) represents the voltage at node j at time t, Z ij P is the impedance between nodes i and j. ij (t), Q ij (t) represents the active power and reactive power flowing between nodes i and j. Let be the voltage value at node i at time t; The power flowing through these lines is expressed as: Therefore, the current flow model is simplified to: in Represents the square of the current in the branch; Considering the systematic distribution of all non-boundary nodes in the cluster: R >> X, it simplifies to:

4. The distributed control method for distributed generation in a distribution network based on optimal spectral clustering according to claim 3, characterized in that: In step S5, the power flow model equations are expressed in Laplace form: In the formula, j∈v x(nb) ,i∈v x(nb) ;v x(nb) It is the set of all non-boundary nodes in cluster K; for all boundary nodes connecting cluster K and cluster M, the voltage equation is: Based on the voltage equation, the above expression can be represented by the cluster boundary as follows: In the formula, b∈v x(B) , v x(B) Let K be the set of all boundary nodes in clusters K and m. The above equation can then be rewritten as: Since voltage variations are based on line load, Where Υ(k) represents the line load factor; further considering the formation of stable clusters; therefore Represented as α*P set (t), so as to minimize the deviation between clusters; therefore, the above formula can be expressed as: in in A comprehensive cluster is formed, and a dynamic distribution model of the power distribution system caused by changes in distributed photovoltaic output is established.

5. The distributed control method for distributed generation in a distribution network based on optimal spectral clustering according to claim 4, characterized in that: Step S6 specifically includes the following steps: Step S601: Calculate the cluster scheduling point for the next time period by minimizing the squared error. The cluster scheduling point assigns a controller node to each cluster to manage the distributed power supply within the cluster, as shown in the following formula. In N measurements, the linear least squares optimization method is applied to minimize the net load of the k-th cluster. Compared with the measured net load P nl The error between; Step S602: The net load support provided by the BESS within the cluster is limited by the rated active power of the BESS within the cluster; for m BESS within the cluster, the total active power available to cluster k is... The upper and lower bounds of the optimization coefficient Υ(k) are expressed as follows; The upper and lower limits restrict the support capacity of each cluster based on the total kW rating of all BESSs within the cluster; for Υ(k), the optimal active support of m BESSs for cluster k is calculated as follows: P BESS (k,n) represents the total active power allocated to the BESS of the k-th cluster, and γ(k) is the optimization coefficient; the scheduling of a single BESS is allocated among the BESS in the cluster in descending order of SoC error and BESS capacity: in soc (j,n)=SOC t (j,n)-SOC(j,n) ekWh soc (j,n)=e soc (j,n)*BESS kWh SOC t (j,n) represents the recovered value of the SoC, e soc (j,n) represents the error value of soc; Step S603: Calculate the active power distribution point of a single BESS using a proportional-integral method. To improve the recovery of the target SoC t (j,n), and keep the BESS SoC within the upper and lower limits; For each BESS within the cluster, based on e soc Calculate the slope based on the (j,n) value. As shown in the following formula; when the SoC error esoc is less than the error threshold When the SoC error exceeds the error threshold, the BESS scheduling is controlled based on the integral component of the error of the last N measurements. The Ki value is further dynamically changed according to the charging and discharging opportunities to improve the recovery of the SoC. in is the upper limit threshold of the SOC error value, kp is the proportional component, and the slope coefficient of the proportional component is based on kp to control sudden changes in BESS scheduling; based on The active power scheduling value for each BESS within the cluster is calculated sequentially as shown in the following formula: in This represents the active power scheduling value for each BESS within the cluster. For e-based soc The (j,n) value is used to calculate the slope of sudden changes in the control BESS scheduling, P. BESS (k,n) represents the charging and discharging power of bess; Subject to BESS kW limit The restrictions are shown in the following formula:

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that, When the processor executes the program, it implements the steps of the distributed control method for distributed power sources in a power distribution network based on optimal spectral clustering as described in any one of claims 1-5.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the distributed control method for distributed power sources in a distribution network based on optimal spectral clustering as described in any one of claims 1-5.

Images