A Reactive Power Consistency Allocation Method for Distributed Photovoltaic Clusters Considering Special Nodes

Abstract

The present invention discloses a method for reactive power consistency allocation of a distributed photovoltaic group considering special nodes. This method takes special nodes into account and, without modifying the structure, achieves the effect of special treatment for special nodes and consistent processing of the power reference mean values of non-special nodes. There is no need to recalculate the weight matrix W because the weight W is not changed during the processing, so the convergence does not change; at special points, virtual values are used to participate in the distributed mean value consistency process. Finally, except for special nodes, the reactive power reference values are evenly distributed.

Description

Technical Field

[0001] The present invention relates to the field of autonomous operation within a distributed photovoltaic cluster, and particularly to a method for reactive power consistency allocation in a distributed photovoltaic cluster considering special nodes. Background Art

[0002] Distributed photovoltaics have characteristics such as being dispersed and numerous. Autonomous operation within a distributed photovoltaic cluster is very important. The distributed consensus algorithm is a commonly used consensus algorithm and has great prospects for power distribution consistency in photovoltaics. However, directly applying the consensus algorithm to the distributed photovoltaic power reference still has some problems: for example, the individual states within the photovoltaic cluster are different, and the power factor and capacity constraints are also different, so the average requirement cannot be fully achieved.

[0003] Distributed average consensus technology: Neighboring nodes can communicate. The reference values of N nodes are represented by an N×1 vector X, and the neighboring weight network is represented by a matrix W, where W(i,j) represents the weight of the connected nodes. Through a finite number of iterations, it can be achieved

[0004] The constraints of the distributed average consensus algorithm are as follows: If node i and node j are not directly connected, then w(i,j)=0; W×1 = 1; spectral radius where the bold 1 represents an N×1 dimensional identity matrix. To achieve the best convergence, the spectral radius should be minimized by setting parameters. The weight parameter of W is related to the network structure and has nothing to do with X. If the network structure changes, W also needs to change accordingly. Summary of the Invention

[0005] To enable some nodes to maintain their own power reference values unchanged under special circumstances and at the same time utilize the weight matrix W designed for the overall framework, the method of the present invention proposes a method for reactive power consistency allocation in a distributed photovoltaic cluster considering special nodes. Using this method, the power reference value of the special node remains a certain fixed value, and at the same time, this node participates in the distributed communication and average consensus process.

[0006] The technical solution adopted by the present invention to solve its technical problems is as follows: Design a method for reactive power consistency allocation in a distributed photovoltaic cluster considering special nodes, characterized in that the method comprises the following steps:

[0007] Step 1: Obtain the weight matrix W through the network structure and constraints within the distributed photovoltaic cluster. The mathematical expression for obtaining W is as follows:

[0008]

[0009] where the bold 1 represents an N×1 dimensional identity matrix, N is the number of nodes in the distributed photovoltaic cluster network, and i, j are two of the N nodes; It is indicated that nodes i and j are not directly connected, and ε represents the set of node connections; s.t. is the abbreviation of "subject to", meaning constraint;

[0010] To obtain the accurate value of W in Equation (1), it is necessary to perform an equivalent transformation on Equation (1):

[0011]

[0012] where η is a scalar and I is the identity matrix. By transformation, the problem is changed into a semi - definite optimization problem, and the specific solution is implemented by semi - definite optimization software;

[0013] Step 2: Use the distributed consensus algorithm to obtain the sequence of average reactive power references Qref1, and the formula is as follows:

[0014]

[0015] where Q is the transposed matrix of the one - dimensional matrix of the initial reactive power of the distributed photovoltaic group network nodes, and each value in the sequence Qref1 is equal, that is, Qref1(1) = Qref1(i) = Qref1(N);

[0016] Step 3: Each node judges whether it can reach the value in the sequence Qref1 obtained in Step 2 according to its own actual state, power factor, and capacity information, and records its own state as Q(i)_state: If the maximum reactive power reference value Q(i)set that node i can reach is less than Qref1(i), then node i is marked as a special node, and the constraint is denoted as Constrain;

[0017] Special nodes first execute a fixed protocol, that is, the reactive power quantity Q(i) is fixed, and the set reactive power reference value Q(i)set makes the node state Q(i)_state within the constraint. The expression is as follows:

[0018]

[0019] At this node, a virtual reactive power reference value V_Q(i) is established:

[0020]

[0021] N represents the total number of all nodes;

[0022] Step 4: Replace the value Qref1(i) at node i in the sequence Qref1 with the virtual reactive power reference value V_Q(i) of node i obtained in Step 3, and use the distributed consensus algorithm again to update the sequence of average reactive power references Qref1 to Qref2:

[0023]

[0024] The sequence of reactive power reference values Q1_final after considering a special point is as follows:

[0025]

[0026] Step 5: If the next special node p appears in the distributed PV cluster network, that is, Q(p)set < Qref2(p), repeat the process of Steps 3 to 4, and establish the virtual reactive power reference value V_Q(p) at this node:

[0027]

[0028]

[0029] The sequence of reactive power reference values Q2_final after considering two special nodes is as follows:

[0030] Q2_final = [Qref3(1),..., Q(i)set,..., Q(p)set,..., Qref3(N)] T (10)

[0031] Step 6: Refer to Steps 4 to 5, and adjust the reactive power reference value each time a special node appears. If the Kth special node s appears in the distributed PV cluster network, that is, Q(s)set < Qrefk(s), establish the virtual reactive power reference value V_Q(s) at this node:

[0032]

[0033] N represents the total number of all nodes, and K represents the Kth special node in the distributed PV cluster network;

[0034] Replace Qrefk(s) in the QrefK sequence with V_Q(s), and use the distributed consensus algorithm for the sequence after replacement to obtain a new sequence of reactive power reference mean values Qref(K + 1). Finally, the sequence of reactive power reference values QK_final after considering K special nodes is as follows:

[0035] QK_final = [Qref(K + 1)(1),..., Q(i)set,..., Q(p)set,..., Q(s)set,..., Qref(K + 1)(N)] T (12)

[0036] Compared with the prior art, the beneficial effects of the present invention are as follows: The method of the present invention takes into account special nodes. Without modifying the structure, it realizes the special processing of special nodes and the consistent processing of the power reference mean value of non-special nodes. There is no need to recalculate the weight matrix W because the weight W is not changed during the processing, so the convergence does not change; at special points, virtual values are used to participate in the distributed mean consensus process. Finally, except for special nodes, the reactive power reference values are evenly distributed. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 FIG. is a flowchart of the steps of an embodiment of the method for reactive power consensus distribution of a distributed photovoltaic group considering special nodes according to the present invention;

[0038] Figure 2 FIG. is a schematic diagram of the connection of photovoltaic group network nodes in Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] The present invention will be further explained below in conjunction with embodiments and the accompanying drawings, but this is not used as a limitation to the protection scope of the present application.

[0040] The present invention provides a method for reactive power consensus distribution of a distributed photovoltaic group considering special nodes, and the method includes the following steps:

[0041] Step 1: Obtain the weight matrix W through the network structure and constraints within the distributed photovoltaic group. The mathematical expression for obtaining W is as follows:

[0042]

[0043] Wherein, the bold 1 represents an N×1 dimensional identity matrix, N is the number of nodes in the distributed photovoltaic group network, and i, j are two nodes among the N nodes; indicates that nodes i and j are not directly connected, ε represents the set of node connections; s.t is the abbreviation of subject to, meaning constraint;

[0044] To obtain the accurate value of W in formula (1), it is necessary to perform an equivalent transformation on formula (1):

[0045]

[0046] Where η is a scalar and I is an identity matrix. By transformation, the problem is changed into a semi-definite optimization problem, and the specific solution is realized by a semi-definite optimization software;

[0047] Step 2: Use the distributed consensus algorithm to obtain the sequence of reactive power reference means Qref1, and the formula is as follows:

[0048]

[0049] Among them, Q is the transposed matrix of the one-dimensional matrix of the initial reactive power of the distributed photovoltaic cluster network nodes. Each value in the sequence Qref1 is equal, that is, Qref1(1) = Qref1(i) = Qref1(N).

[0050] Step 3: Each node determines whether it can reach the value in the sequence Qref1 in Step 2 according to its own actual state, power factor, and capacity information, and records its own state as Q(i)_state: If the maximum reactive power reference value Q(i)set that node i can reach is less than Qref1(i) due to constraints such as voltage constraints, capacity constraints, or power factor at the node, then node i is marked as a special node, and the constraint is recorded as Constrain;

[0051] The special node first executes a fixed protocol, that is, the reactive power quantity Q(i) is fixed, and the set reactive power reference value Q(i)set makes the node state Q(i)_state within the constraint. The expression is as follows:

[0052]

[0053] At this node, a virtual reactive power reference value V_Q(i) is established:

[0054]

[0055] N represents the total number of all nodes;

[0056] Step 4: Replace the value Qref1(i) at node i in the sequence Qref1 with the virtual reactive power reference value V_Q(i) of node i obtained in Step 3, and use the distributed consensus algorithm again to update the reactive power reference mean sequence Qref1 to Qref2:

[0057]

[0058] The reactive power reference value sequence Q1_final considering one special point is:

[0059]

[0060] Step 5: If there is a next special node p in the distributed photovoltaic cluster network, that is, Q(p)set < Qref2(p), repeat the process of Step 3 to Step 4, and establish a virtual reactive power reference value V_Q(p) at this node:

[0061]

[0062]

[0063] The reactive power reference value sequence Q2_final after considering two special nodes is as follows:

[0064] Q2_final = [Qref3(1),..., Q(i)set,..., Q(p)set,..., Qref3(N)] T (10)

[0065] Step 6: Referring to Steps 4 to 5, adjust the reactive power reference value once for each occurrence of a special node. If the Kth special node s appears in the distributed photovoltaic cluster network, i.e., Q(s)set < Qrefk(s), establish a virtual reactive power reference value V_Q(s) at this node:

[0066]

[0067] N represents the total number of all nodes, and K represents the Kth special node in the distributed photovoltaic cluster network;

[0068] Replace Qrefk(s) in the QrefK sequence with V_Q(s), and use the distributed consensus algorithm for the sequence after replacement to obtain a new reactive power reference mean sequence Qref(K + 1). Finally, the reactive power reference value sequence QK_final after considering K special nodes is as follows:

[0069] QK_final = [Qref(K + 1)(1),..., Q(i)set,..., Q(p)set,..., Q(s)set,..., Qref(K + 1)(N)] T (12)

[0070] Embodiment 1

[0071] This embodiment provides a method for reactive power consistency allocation in a distributed photovoltaic cluster considering special nodes. The method includes the following steps:

[0072] Step 1: Obtain the weight matrix W through the distributed photovoltaic cluster network structure and constraints, and the mathematical expression for obtaining W is as follows:

[0073]

[0074] Among them, bold 1 represents an N * 1 dimensional identity matrix, N is the number of nodes in the distributed photovoltaic cluster network, and i, j are two nodes among the N nodes; indicates that nodes i and j are not directly connected, ε represents the set of node connections; s.t is the abbreviation of subject to, meaning constraint;

[0075] To obtain the accurate value of W in Equation (1), it is necessary to perform an equivalent transformation on Equation (1):

[0076]

[0077] where η is a scalar and I is the identity matrix. By transformation, the solution of W is transformed into a semi - definite optimization problem, and the specific solution is implemented by semi - definite optimization software;

[0078] Figure 2 is a simplified schematic diagram of the node connection of the photovoltaic group network in this embodiment. Through Figure 2 the network structure and constraints, according to Equations (1) and (2), using a semi - definite optimization software package, the minimum value that η can obtain is 0.707. At this time, the weight matrix W:

[0079]

[0080] Step 2: Use the distributed consensus algorithm to solve the mean value of the reactive power reference mean sequence Qref1. The formula is as follows:

[0081]

[0082] where Q is the transposed matrix of the one - dimensional matrix of the initial reactive power state of the distributed photovoltaic group network. In this embodiment, it is assumed that the initial reactive power Q = [3, 5, 8, 1] T , with the unit of kvar. For the convenience of description, the unit will not be mentioned hereinafter.

[0083] Use the distributed consensus algorithm to solve the mean value of the reactive power reference Qref1. After 28 iterations of calculation, the reference values reach consistency. According to Equation (3), we get:

[0084] Qref1 = W 28 *Q = [4.25, 4.25, 4.25, 4.25] T

[0085] Step 3: If Node 1 can only reach a reactive power reference value with a deviation from Qref1 due to voltage constraints, capacity constraints, or power factor at the node, it is marked as a special node. Assume that the limit that Node 1 can withstand is 3.5;

[0086] The special node 1 first executes the fixed protocol. According to formula (4),

[0087] Q(1) = Q(1)set = 3.5

[0088] At this node, a virtual reactive power reference value V_Q(1) is established. According to formula (5),

[0089]

[0090] Where N = 4 and K = 1;

[0091] Step 4: Use the virtual reactive power reference value V_Q(1) to replace the value Qref1(1), and use the distributed consensus algorithm again to update the reactive power reference value to Qref2. According to formula (6),

[0092] Qref2 = W 10 *[5.25, 4.25, 4.25, 4.25] T = [4.5, 4.5, 4.5, 4.5] T

[0093] According to formula (7), we get:

[0094] Q1_final = [3.5, 4.5, 4.5, 4.5] T

[0095] At this time, the reactive power output reference values of the four nodes are 3.5, 4.5, 4.5, and 4.5 respectively, achieving the reactive power averaging except for the special node;

[0096] Step 5: Assume that the upper limit of the reactive power of node 4 is 4.3 and it cannot reach 4.5. According to formula (4), fix the reactive power of node 4:

[0097] Q(4) = Q(4)set = 4.3

[0098] Use the virtual reactive power reference value and utilize formula (8):

[0099]

[0100] Participate in the averaging process of other nodes and utilize formula (9):

[0101] W 21 *[4.5, 4.5, 4.5, 4.9] T = [4.6, 4.6, 4.6, 4.6] T

[0102] Utilize formula (10) to get:

[0103] Q2_final = [3.5, 4.6, 4.6, 4.3] T

[0104] At this time, the reactive power output reference values of the four nodes are 3.5, 4.6, 4.6, and 4.3 respectively, achieving the average distribution of reactive power except for the special node.

[0105] The parts not described in this invention are applicable to the prior art.

Claims

1. A reactive power consistency allocation method for distributed photovoltaic groups considering special nodes, characterized in that, the method includes the following steps: Step 1: Obtain the weight matrix W through the network structure and constraints within the distributed photovoltaic group. The mathematical expression for obtaining W is as follows: Among them, bold 1 represents an N×1 dimensional identity matrix, where N is the number of nodes in the distributed photovoltaic cluster network, and i, j are two nodes among the N nodes; indicates that nodes i and j are not directly connected, ε represents the set of node connections; s.t is the abbreviation of subject to, meaning constraint; To obtain the accurate value of W in Equation (1), it is necessary to perform an equivalent transformation on Equation (1): where η is a scalar and I is the identity matrix. By transformation, the problem is changed into a semi - definite optimization problem, and the specific solution is implemented by semi - definite optimization software; Step 2: Use the distributed consistency algorithm to obtain the reactive power reference mean sequence Qref1, and the formula is as follows: where Q is the transposed matrix of the one - dimensional matrix of the initial reactive power of the network nodes in the distributed photovoltaic group. Each value in the sequence Qref1 is equal, that is, Qref1(1)=Qref1(i)=Qref1(N); Step 3: Each node judges whether it can reach the value in the sequence Qref1 in Step 2 according to its own actual state, power factor, and capacity information, and records its own state as Q(i)_state: If the maximum reactive power reference value Q(i)set that node i can reach is less than Qref1(i), then node i is marked as a special node, and the constraint is denoted as Constrain; The special node first executes a fixed protocol, that is, the reactive power quantity Q(i) is fixed, and the set reactive power reference value Q(i)set makes the node state Q(i)_state within the constraint. The expression is as follows: At this node, establish a virtual reactive power reference value V_Q(i): N represents the total number of all nodes; Step 4: Replace the value Qref1(i) at node i in the sequence Qref1 with the virtual reactive power reference value V_Q(i) of node i obtained in Step 3, and use the distributed consistency algorithm again to update the reactive power reference mean sequence Qref1 to Qref2: The reactive power reference value sequence Q1_final after considering one special point is: Step 5: If there is a next special node p in the distributed photovoltaic group network, that is, Q(p)set<Qref2(p), repeat the process from Step 3 to Step 4, and establish a virtual reactive power reference value V_Q(p) at this node: The reactive power reference value sequence Q2_final after considering two special nodes is: Q2_final = [Qref3(1), …, Q(i)set, …, Q(p)set, …, Qref3(N)] T (10) Step 6: Refer to Step 4 to Step 5. Each time a special node appears, the reactive power reference value is adjusted. If there is a K - th special node s in the distributed photovoltaic group network, that is, Q(s)set<Qrefk(s), establish a virtual reactive power reference value V_Q(s) at this node: N represents the total number of all nodes, and K represents the K - th special node in the distributed photovoltaic group network; Replace Qrefk(s) in the QrefK sequence with V_Q(s), and use the distributed consistency algorithm on the sequence after replacement values again to obtain a new reactive power reference mean sequence Qref(K + 1). Finally, the reactive power reference value sequence QK_final after considering K special nodes is: QK_final = [Qref(K+1)(1), …, Q(i)set, …, Q(p)set, …, Q(s)set, …, Qref(K+1)(N)] T (12).

Images