Graph computing-based multi-stage topology reconstruction method and system for power distribution network

Abstract

A graph computing-based multi-stage topology reconstruction method and system for a power distribution network. The method comprises: providing a power distribution network reconstruction method based on an optimal flow method and a branch exchange method (S110); constructing a power distribution network multi-stage reconstruction model with optimal power loss (S120); providing a graph computing-based multi-stage reconstruction computing method for the power distribution network (S130); and solving the graph computing-based optimal power loss model to obtain a power distribution network reconstruction technology (S140).

Description

A Multi-Stage Topology Reconfiguration Method and System for Distribution Networks Based on Graph Computation

[0001] This application claims priority to Chinese Patent Application No. 202410985204.9, filed on July 23, 2024, the entire contents of which are incorporated herein by reference. Technical Field

[0002] This application relates to the field of distribution network energy storage technology, such as a method and system for multi-stage topology reconfiguration of distribution networks based on graph computing. Background Technology

[0003] The distribution network is one of the basic components of the power grid. It involves all aspects of users' electricity consumption, has a large structure, and is complex in operation. The quality of the distribution network's operation directly affects the quality of electricity consumption for users. The distribution network mainly transmits electrical energy from generators at nodes and the upper-level grid at the root node to other load nodes. However, as the number of branches and nodes in the distribution network gradually increases, the system scale continues to expand, and the calculations become increasingly complex. Traditional centralized computing has poor scalability and high communication costs, making it difficult to meet the computing needs of distribution networks.

[0004] In the field of distribution network research, reducing network losses has always been a problem to be solved because the impedance of a branch is much greater than its resistance. Distribution network reconfiguration reduces power losses by changing the topology of the distribution network, that is, changing the open and closed states of sectionalizing switches and tie switches between nodes, thereby transferring load nodes.

[0005] Current research on distribution network reconfiguration mainly focuses on three methods: mathematical optimization, heuristic methods, and intelligent optimization algorithms. While mathematical optimization can quickly find the optimal solution, it cannot be directly applied to non-convex and nonlinear problems and relies excessively on initial values. Heuristic methods, although simple and capable of finding the optimal solution with high quality, become computationally inefficient as the number of nodes increases. Intelligent optimization algorithms, while suitable for small systems, struggle to converge to the optimal solution in large-node systems. Furthermore, with increasing communication costs and growing user privacy requirements, efficiently and securely reducing network losses is a significant challenge for distribution network reconfiguration.

[0006] For example, invention patent application number 202310097347.1 discloses a distributed photovoltaic capacity configuration method considering distribution network reconfiguration. It employs the Bird's Nest algorithm to make the distribution network capacity configuration more reasonable and uses the optimal flow pattern method for reconfiguration. Invention patent application number 202111106696.2 discloses a method for handling distribution network faults and reconfiguring networks in the face of strong uncertainty, which achieves post-fault network optimization and reconfiguration through branch switching. However, these methods have the following shortcomings: 1) As the number of nodes increases, the workload becomes excessive and the computational efficiency is low. 2) They cannot accurately and effectively reconfigure the distribution network, and are affected by the original topology. Specifically, the optimal flow pattern method has a slow solution speed and difficulty in searching for the global optimum, and is affected by the mutual influence of currents. The branch switching method has a large workload and low computational efficiency as the number of nodes increases, and is affected by the power flow calculation in non-independent topologies. 3) In distribution network reconfiguration, it cannot safely guarantee user information and effectively reduce network losses. Summary of the Invention

[0007] This application provides a graph-based method for multi-stage topology reconfiguration of distribution networks, and a graph-based system for multi-stage topology reconfiguration of distribution networks.

[0008] A first aspect of this application provides a multi-stage topology reconfiguration method for distribution networks based on graph computation, the method comprising the following steps:

[0009] The distribution network to be restructured is divided into multiple power grid areas by utilizing the space or structure of the distribution network;

[0010] For each power grid region, a multi-stage reconfiguration method based on the improved optimal flow pattern method and branch switching method is used to obtain different topologies and corresponding power loss values ​​for each topology.

[0011] Determine whether each topology and its corresponding power loss value meet the preset conditions of the distribution network reconfiguration model. The preset conditions include: an objective function with the goal of minimizing power loss and inherent constraints. In response to the power loss value corresponding to the topology meeting the preset conditions, save the corresponding topology of each power grid region.

[0012] The topology of each divided power grid area that meets the preset conditions is input into the graph calculation library. The graph calculation iterative training is performed using the synchronous alternating direction multiplier method until the switches in all power grid areas of the distribution network complete the reconfiguration process. The result of the multi-stage topology reconfiguration of the distribution network is then output.

[0013] The multi-stage reconstruction method based on the improved optimal flow pattern method and branch switching method includes:

[0014] Close all switches in the initial topology of the distribution network to form a ring topology;

[0015] The iteration number is set to be equal to the number of branches in the ring topology. In each iteration, the switches in the ring topology are disconnected in turn, the power loss value is calculated in parallel and distributed, and the switch with the smallest power loss value is selected as the connecting switch. After the iteration is completed, all the connecting switches are disconnected to obtain a radial network topology.

[0016] The obtained radial network topology is processed by a switch pre-classification method to perform state transitions of segmented switches and tie switches, obtaining the topology and power loss values ​​under each state transition. The optimal topology and power loss values ​​are then obtained through parallel distributed computing.

[0017] On the other hand, this application also provides a multi-stage topology reconfiguration system for distribution networks based on graph computation, including:

[0018] The partitioning module is configured to divide the distribution network to be reconfigured into multiple power grid regions using the space or structure of the distribution network.

[0019] The topology setting module is configured to use a multi-stage reconfiguration unit based on an improved optimal flow pattern method and a branch exchange method to obtain different topologies and corresponding power loss values ​​for each topology for each power grid region.

[0020] The graph calculation and judgment module is configured to determine whether each topology and the power loss value corresponding to each topology meet the preset conditions of the distribution network reconfiguration model. The preset conditions include: an objective function with the goal of minimizing power loss and inherent constraints; if the power loss value corresponding to the topology meets the preset conditions, the corresponding topology of each power grid region is saved.

[0021] The multi-stage reconfiguration module is configured to input the topology of each divided power grid area that meets the preset conditions into the graph calculation library, and use the synchronous alternating direction multiplier method to perform graph calculation iterative training until the switches in all power grid areas of the distribution network complete the reconfiguration process, and output the result of the multi-stage topology reconfiguration of the distribution network.

[0022] The multi-stage reconstruction unit based on the improved optimal flow pattern method and branch switching method includes:

[0023] Loop topology area: This is set to close all switches in the initial topology of the distribution network to form a loop topology.

[0024] The graph calculation iteration area is set to have the number of iterations equal to the number of branches in the ring topology. In each iteration, the switches in the ring topology are disconnected sequentially, the power loss value is calculated in parallel and distributed, and the switch with the smallest power loss value is selected as the connecting switch. After the iteration ends, all the connecting switches are disconnected to obtain a radial network topology.

[0025] The graph computation state transition region is set up to perform segmented switching and tie-switching state transitions on the obtained radial network topology using a switch pre-classification method, obtaining the topology and power loss value under each state transition, and obtaining the optimal topology and power loss value through parallel distributed computation. Attached Figure Description

[0026] Figure 1A is a flowchart illustrating a graph-based multi-stage topology reconfiguration technique for distribution networks according to an exemplary embodiment;

[0027] Figure 1B is an exemplary embodiment illustrating a multi-stage topology reconfiguration system method for a distribution network based on graph computation;

[0028] Figure 2 is a schematic diagram of a power distribution network structure according to an exemplary embodiment;

[0029] Figure 3 is a schematic diagram of a power distribution network based on graph computing principles according to an exemplary embodiment;

[0030] Figure 4 is a schematic diagram illustrating the convergence of an alternating direction multiplier method based on graph computation according to an exemplary embodiment;

[0031] Figure 5 is a schematic diagram of the voltage of multiple nodes in a distribution network according to an exemplary embodiment;

[0032] Figure 6 is a power diagram of multiple nodes in a distribution network according to an exemplary embodiment;

[0033] Figure 7 is a flowchart illustrating a graph-based computation according to an exemplary embodiment;

[0034] Figure 8 is a schematic diagram of a multi-stage topology reconfiguration system for a distribution network based on graph computation, according to an exemplary embodiment. Detailed Implementation

[0035] The technical solutions of the embodiments of this application will be described below with reference to the accompanying drawings. The described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0036] 1) The optimal flow pattern method includes:

[0037] The optimal flow pattern method assumes the distribution network is composed of pure resistance, and the load at each node is represented by the injected current. Therefore, the relationship between the node voltage matrix U and the node current matrix I can be obtained as U = I / Y, where Y is the node admittance matrix. The formula for calculating the network loss of the distribution network is then:

[0038] In the formula: v i i i The voltage and current at node i are respectively, v n Let be the voltage at node n, where n is the number of nodes in the distribution network.

[0039] This application modifies the switching states of nodes i and j in the original network N0 by turning off the switches of nodes i and j. Assume the conductance of the branch is g. ij The remaining conductance is Δg ij According to the optimal flow pattern method, when the switch between nodes is open, it is equivalent to a parallel conductance -g. ij This forms a new power network N.

[0040] Therefore, the change in power loss of the node switch before and after opening and closing can be expressed as:

[0041] The optimal flow mode method has good convergence in distribution network reconfiguration, but it is slow to solve and difficult to find the optimal solution, and the currents affect each other.

[0042] 2) Branch exchange method, including:

[0043] Branch-switching involves changing the state of tie switches and sectionalizing switches to transfer load demand to other nodes, thereby reducing power loss. Each opening and closing of a tie switch results in changes in active power loss; however, the voltage along the distribution network generally varies little, so reactive power flow can be ignored.

[0044] Branch switching can increase the number of feasible solutions in distribution network reconfiguration, but it reduces the convergence speed and introduces the impact of independent topologies on power flow calculation.

[0045] Therefore, based on the above, this application constructs a multi-stage distribution network reconfiguration technology that can accurately and effectively open and close sectionalizing switches and tie switches. It also considers the computationally complex process, aggregates distribution network areas, introduces the basic theory of graph computation, and calculates the optimal power loss of multiple regional power grids in parallel, while protecting user privacy information, to achieve optimal distribution network reconfiguration. That is:

[0046] 3) Multi-stage distribution network reconfiguration based on optimal flow pattern method and branch exchange method

[0047] This application provides a graph-based multi-stage topology reconfiguration technology for distribution networks, which can effectively reduce power loss and minimize costs. At the same time, it greatly improves the solution efficiency by utilizing the principle of graph computing. The reconfiguration technology proposed in this application can further reduce the impact of ring network currents and has low requirements for the initial topology structure. It also increases the solution efficiency and privacy. On this basis, the cost of distribution networks is reduced, providing a way to alleviate the power loss of distribution networks.

[0048] To improve upon the problems of mutual current influence in the ring network and the islanding effect of nodes in the optimal flow pattern method, as well as the impact of non-independent topology on power flow calculation in the branch exchange method, this application proposes a multi-stage distribution network reconfiguration method based on the optimal flow pattern method and the branch exchange method.

[0049] S100. Obtain the initial optimal solution for optimal power loss. Close all switches in the initial distribution network topology to form a ring topology. Set the number of iterations to equal the number of branches in the ring topology. In each iteration, sequentially disconnect the switches in the ring topology and calculate the power loss. Select the switch with the minimum power loss as the tie switch. After the iteration is complete, disconnect all the tie switches to obtain a radial network topology.

[0050] S200: Optimize the obtained radial network topology again, change the segmented switch to a tie switch, and use S100 to obtain a new radial network topology. Obtain the power loss of different radial network topologies, and thus obtain the optimal reconfiguration scheme.

[0051] To improve solution efficiency, a switch pre-classification method is used in S200 to exclude the following types of segmented switches:

[0052] Type I: In the initial network topology, the shortest path from a node to a substation node consists of n1 or fewer segmented switches.

[0053] Type II: In the initial network topology, the shortest path from a node to a substation node consists of n² or fewer segmented switches.

[0054] Type III: Section switches not in the closed circuit of the switch.

[0055] Because the state of the node to the substation node of type I is relatively easy to determine, the sectionalizing switches of type I can be excluded in advance. In order to avoid islanding effect, the sectionalizing switches of type II do not change their state, and the sectionalizing switches of type III cannot change their state either.

[0056] After the switch pre-classification in S300 and S200, and then the disconnection of the sectionalizing switches of types I, II, and III in S100, the optimal topology is obtained and the power loss is minimized. Then, the sectionalizing switches of type II are included. This application uses a simultaneous on-off method, that is, for each type II sectionalizing switch disconnected, a tie switch is simultaneously closed.

[0057] As shown in Figure 1A, the multi-stage topology reconfiguration method for the distribution network in this embodiment includes:

[0058] S110. A distribution network reconfiguration method based on the optimal flow pattern method and the branch exchange method is proposed.

[0059] S120. A multi-stage reconfiguration model for distribution networks that constructs an optimal power loss model;

[0060] S130. A calculation method for multi-stage reconfiguration of distribution networks based on graph computation is proposed.

[0061] S140. Solve the optimal power loss model to obtain the power distribution network reconfiguration technology.

[0062] As shown in Figure 1B, as a possible implementation of this embodiment, this embodiment provides a multi-stage topology reconfiguration method for distribution networks based on graph computing, which includes the following steps.

[0063] S1. Divide the distribution network to be reconfigured into multiple power grid areas by utilizing the space or structure of the distribution network;

[0064] To improve the efficiency of distribution network reconfiguration, this application reconfigures the distribution network to be reconfigured by dividing it into multiple regional power grids using indicators such as spatial or structural features.

[0065] S2. For each power grid region, a multi-stage reconfiguration method based on the improved optimal flow pattern method and branch exchange method is adopted to obtain different topologies and the corresponding power loss values ​​for each topology.

[0066] The multi-stage reconstruction method based on the improved optimal flow pattern method and branch switching method includes:

[0067] S21. Close all switches in the initial topology of the distribution network to form a ring topology.

[0068] S22. Set the number of iterations to equal the number of branches in the ring topology. In each iteration, disconnect the switches in the ring topology in turn, calculate the power loss value in parallel and distributed, and select the switch with the smallest power loss value as the connecting switch. After the iteration ends, disconnect all the connecting switches to obtain the radial network topology.

[0069] S23. The obtained radial network topology is subjected to a switch pre-classification method to perform state transitions of segmented switches and tie switches, and the topology and power loss value under each state transition are obtained. The optimal topology and power loss value are obtained through parallel distributed computing.

[0070] In one embodiment, step S22 includes:

[0071] S221. Input the ring topology corresponding to the current power grid area, find all switches K, and initialize the current switch K;

[0072] S222. Disconnect the k-th switch in the ring topology, calculate the power loss value, and determine whether each node in the ring topology has a path to the substation and whether it satisfies the inherent constraints. If each node has a path to the substation and satisfies the inherent constraints, proceed to step S223. If there are nodes that do not have a path to the substation or do not satisfy the inherent constraints, set the objective function value to a large positive value (10). 9 );

[0073] S223. Save the current power loss value and disconnect the (k+1)th switch;

[0074] S224. Determine whether k > K is satisfied. If k > K is satisfied, find the switch with the minimum power loss value as the tie switch. When the number of iterations reaches the number of branches in the ring topology, set all the tie switches obtained to open to form a radial network topology. If k > K is not satisfied, go to step S222.

[0075] In one embodiment, step S23 includes:

[0076] S231. Input the ring topology, convert all segment switches to tie switches, and use a multi-stage reconstruction method based on the improved optimal flow pattern method and branch exchange method to obtain the latest topology. Record the segment switches other than types I-III in the current topology as set Ω1, and the segment switches of type II as set Ω2. Type I is where the shortest path from the current node to the substation node in the ring topology consists of n1 or less segment switches; Type II is where the shortest path from the current node to the substation node in the ring topology consists of n2 or less segment switches; Type III is segment switches not in the switch closed loop.

[0077] S232. For the segmented switches excluding set Ω1, open one segmented switch at a time in sequence and close all other switches, execute the steps of step S22, and keep the segmented switch open until a radial network topology is formed. Save the radial network topology and the corresponding power loss value at this time.

[0078] S233. Perform on-off action on the segmented switches within set Ω2. The on-off action is to close a connecting switch for each type II segmented switch that is opened to maintain connectivity. For example, when the switches at nodes 5 and 6 in Figure 2 are opened, connecting switches 6-7 need to be closed to ensure connectivity.

[0079] S234. Execute the available switching action, solve the optimal power loss function, record the topology and power loss value after this switching action, determine whether this switching action reduces power loss, and if it reduces power loss, save the topology and corresponding power loss value obtained after this switching action.

[0080] S235. Perform step S234 on all segmented switches in set Ω2;

[0081] S236. Save the different topologies obtained after executing different on-off actions and the power loss value corresponding to each topology, and close the open / closed segment switches.

[0082] S237. Repeat steps S232-S236 for the segmented switches in the ring topology.

[0083] S3. Determine whether each topology and its corresponding power loss value meet the preset conditions of the distribution network reconfiguration model. The preset conditions include: an objective function with the goal of minimizing power loss and inherent constraints. If the preset conditions are not met, the corresponding topology will no longer be considered. If the preset conditions are met, the corresponding topology of each power grid region will be saved.

[0084] As one possible implementation of this embodiment, the process of constructing the multi-stage reconfiguration model of the distribution network with the optimal power loss model is as follows:

[0085] The safe and economical operation of a power distribution network is related to many parameters, and the network loss of the distribution network accounts for a large proportion of the total network loss in the power system. Therefore, this application constructs an objective function with the goal of minimizing power loss:

[0086] In the formula: p ij q ij r ij Let v represent the active power and reactive power from node i to node j, respectively, and the branch resistance between node i and node j. j N and N are the voltage at node j and the number of load nodes, respectively.

[0087] To ensure the safe and stable operation of the power distribution network, certain inherent constraints must be met, mainly including equality constraints and inequality constraints.

[0088] (1) Current constraints

[0089] In the formula: Formula (11a) represents the active power balance of the nodes, which is a prerequisite for ensuring the frequency stability of the distribution network, p demand,i v i G represents the active power and voltage at node i, respectively. ij B ij Ω represents the real and imaginary parts of the admittance matrix in the i-th row and j-th column, respectively. b It is the set of all branches. Formula (11b) represents the reactive power balance of the nodes, which is a prerequisite for ensuring the voltage stability of the distribution network. q demand,i This represents the reactive power of node i. Formula (11c) is the maximum transmission capacity limit between node i and node j, s ij It is the maximum apparent power.

[0090] (2) Distribution network current constraints, voltage constraints, and micro-turbine constraints:

[0091] In the formula: (12a) is the voltage constraint of the distribution network, which ensures the safe operation of the distribution network and electrical equipment, v i Let v be the voltage at node i. i,max v i,min These are the upper and lower voltage limits for node i, respectively. (12b) represents the branch current constraint from node i to node j, ensuring line safety. ij Let I be the branch current from node i to node j. ij,min I ij,max These are the upper and lower limits of the branch current, respectively. (12c) represents the active power output constraint of the micro-turbine, p g(i) Let p be the active power of the micro gas turbine at node i. g(i),max p g(i),min Let q be the upper and lower limits of the active power of the micro gas turbine at node i, respectively. (12d) represents the reactive power output constraint of the micro gas turbine. g(i) Let q be the reactive power of the micro gas turbine at node i. g(i),max q g(i),min These are the upper and lower limits of the reactive power of the micro gas turbine at node i, respectively.

[0092] (3) Distribution network topology constraints

[0093] During the reconfiguration of the distribution network, radial and connectivity constraints must be satisfied, which can be expressed as:

[0094] Where: n b Let dis(i,i) be the number of branches. ge ) represents the distance from node i to substation node i geThe distance.

[0095] S4. Iteratively calculate the corresponding topology in each power grid area using graph computing methods. When the convergence condition of parallel distributed computing is met, the distribution network reconfiguration ends. Alternatively, input the topology of each divided power grid area that meets the preset conditions into the graph computing library, and use the synchronous alternating direction multiplier method to perform graph computing iterative training until the switches in all power grid areas of the distribution network complete the reconfiguration process, and output the result of the multi-stage topology reconfiguration of the distribution network.

[0096] As one possible implementation of this embodiment, this application introduces graph computing for distribution network reconfiguration calculation. Graph computing utilizes the principles of graph theory to describe the nodes and edges of the distribution network in matrix form. Nodes are connected through edges, and communication information is exchanged through edges.

[0097] A power distribution network consists of nodes and branches. Multiple nodes are connected by branches to transmit information, forming a basic concept for graph computation. Utilizing the fundamental principles of graph computation, the structure of the power distribution network is transformed into a graph relationship. Multiple nodes are designated as vertices, and optimized data is stored in a graph database for parallel computation, improving computational efficiency. Each vertex is configured to perform parallel optimization calculations of power loss in the power distribution network based on information transmitted from adjacent vertices, and the boundary information is updated according to the optimization results, as shown in Figure 7.

[0098] This application derives a graph-based computational method for distribution network reconfiguration, expressing the power loss during reconfiguration in a compact form as follows:

[0099] In the formula: f is the objective function expression, x n Let be the decision variable for the nth power grid region.

[0100] By introducing the alternating direction multiplier method into graph computation, a graph computation method for distribution network reconfiguration based on the alternating direction multiplier method is derived.

[0101] Define an auxiliary function, namely the node indicator function:

[0102] In the formula: z n Let n be the state vector of node n.

[0103] Using the alternating direction multiplier method, formula (13) can be transformed into:

[0104] Then, transform formula (15) into a Lagrange function:

[0105] In the formula: w = l / ρ, l, and ρ are the Lagrange multiplier and the penalty factor, respectively.

[0106] Distributed computation is performed on the optimized data of the graph database. To this end, the synchronous alternating direction multiplier method is introduced into the distribution network reconfiguration, and formula (16) is transformed into:

[0107] In the formula: Let n be the exchange power of the nth power grid region. Let be the exchange power of the nth power grid region in the kth iteration.

[0108] According to the function h(z) n Given the definition of ), the final iterative form of the synchronous alternating direction multiplier method is:

[0109] In the formula: Let be the average exchange power in the k-th iteration.

[0110] To establish fully distributed graph computation, graph theory is introduced. This application describes the distribution network nodes as an undirected graph, where each node is a vertex and branches are edges. In summary, the Laplace matrix of the distribution network is described as:

[0111] In the formula: L1 and L2 are the adjacency matrix and degree matrix, respectively, and D i |D i | represents the set of nodes connected to node i and the number of nodes in that set, respectively.

[0112] Based on the principle of discrete consistency theory, if a Laplace matrix has L non-zero eigenvalues, the average power exchange of the nodes can be obtained after L iterations:

[0113] In the formula: x i (L) is the updated value of node i after L iterations, s ii (d), s ij (d) is the weight factor at the d-th iteration. It is the (d+1)th non-zero eigenvalue in the Laplace matrix.

[0114] In one example, the residual is defined for the (k+1)th iteration. When all areas of the distribution network converge, the distribution network is considered converged, and the result of the multi-stage topology reconfiguration of the distribution network is output.

[0115] In summary, the multi-stage topology reconfiguration model for distribution networks based on graph computing proposed in this application consists of equations (11)-(21). By solving this mathematical model, the safety of the distribution network can be improved while ensuring its economic operation. Furthermore, since the distribution network reconfiguration technology proposed in this application is based on graph computing, which is a novel parallel computing method, it can improve computational efficiency, protect user privacy, and reduce communication costs.

[0116] In this embodiment, the process of solving the optimal power loss model to obtain the distribution network reconfiguration technology is as follows: the Gurobi solver is used to solve the problem with an error of 0.01%.

[0117] The following example illustrates the multi-stage topology reconfiguration method for distribution networks based on graph computation proposed in this application. Figure 2 shows a schematic diagram of the Institute of Electrical and Electronics Engineers (IEEE)-33 node structure of a distribution network, containing 33 nodes and 37 branches. Five of these branches are tie switches, and the rest are sectionalizing switches. The distribution network is divided into multiple regional grids using spatial or structural indicators. The base voltage is 12.66 kV, with upper and lower voltage limits set at 1.06 pu and 0.94 pu, respectively. The micro-turbine generation cost coefficients are shown in Table 1, where a, b, and c are the micro-turbine cost coefficients inherent to the IEEE nodes. Furthermore, based on the principle of graph computation, graph theory is introduced to transform the IEEE-33 node structure of the distribution network into a schematic diagram for graph computation, as shown in Figure 3.

[0118] Table 1. Cost coefficients for micro-turbine power generation

[0119] In the multi-stage topology reconfiguration method proposed in this application, the opening and closing of segmented switches and tie switches are performed in three stages. A completely new network topology and optimal power loss are obtained in each stage. Simulation tests were conducted on IEEE-33 nodes, setting up four different test scenarios:

[0120] Case 1: Topology reconstruction was not considered;

[0121] Case 2: Consider only the first stage of multi-stage topology reconstruction (based on the optimal flow pattern method);

[0122] Case 3: Only consider the first and third stages of multi-stage topology reconstruction;

[0123] Case 4: The multi-stage topology reconstruction method proposed in this application.

[0124] Table 2 shows the power losses under different scenarios. As can be seen from Table 2, in the IEEE-33 node, the power losses obtained in cases 2-4 are the same. This is because the first step of the reconfiguration method in the IEEE-33 node already achieves the optimal topology and optimal power loss. Without considering the distribution network reconfiguration method, the power loss is 62.556% higher than the method proposed in this application.

[0125] Table 2 Power Loss in Different Scenarios

[0126] Meanwhile, the topology reconfiguration method of this application is performed under the condition of satisfying voltage constraints. As can be seen from Figure 5, at each node and at each time, the distribution network reconfiguration method proposed in this application satisfies the voltage constraints of 1.06 pu and 0.94 pu. The active power obtained by this application is shown in Figure 6.

[0127] As shown in Figure 4, the graph computation method based on the synchronous alternating direction multiplier method in this application can guarantee convergence within 40 iterations, demonstrating good convergence and ensuring the effectiveness of this application. This application constructs a distribution network reconfiguration technology that can accurately perform state transitions of segmented switches and tie switches in the distribution network while ensuring optimal power loss. Furthermore, by utilizing the principles of graph computation, it performs distributed reconfiguration of the distribution network, improving computational efficiency and ensuring the safety of the distribution network voltage.

[0128] The graph-based multi-stage topology reconfiguration technique for distribution networks presented in this application can reduce power losses. Furthermore, this method expands the application of distribution network reconfiguration techniques in reducing power losses, and can be applied to areas such as reducing power losses and improving power supply reliability in power systems.

[0129] On the other hand, as shown in Figure 8, this application also provides a multi-stage topology reconfiguration system for distribution networks based on graph computing, the system comprising:

[0130] The partitioning module is configured to divide the distribution network to be reconfigured into multiple power grid regions using the space or structure of the distribution network.

[0131] The topology setting module is configured to use a multi-stage reconfiguration unit based on an improved optimal flow pattern method and a branch exchange method to obtain different topologies and corresponding power loss values ​​for each topology for each power grid region.

[0132] The multi-stage reconstruction unit based on the improved optimal flow pattern method and branch switching method includes:

[0133] Loop topology area: This is set to close all switches in the initial topology of the distribution network to form a loop topology.

[0134] Graph computation iteration region: Set the number of iterations to be equal to the number of branches in the ring topology. In each iteration, the switches in the ring topology are disconnected in turn, the power loss is calculated in parallel and distributed, and the switch with the minimum power loss is selected as the connecting switch. After the iteration ends, all the connecting switches are disconnected to obtain a radial network topology.

[0135] The graph computation state transition region is set to perform segmented switching and tie-switching state transitions on the obtained radial network topology using a switch pre-classification method, obtaining the topology and power loss value under each state transition, and obtaining the optimal topology and power loss value through parallel distributed computation.

[0136] The graph calculation and judgment module is configured to determine whether each topology and its corresponding power loss value meet the preset conditions of the distribution network reconfiguration model. The preset conditions include: an objective function with the goal of minimizing power loss and inherent constraints. If the preset conditions are not met, the corresponding topology will no longer be considered. If the preset conditions are met, the corresponding topology of each power grid region will be saved.

[0137] The graph calculation iteration region is set as follows:

[0138] Step 2-1: Input the ring topology corresponding to the current power grid area, find all switches K, and initialize the current switch K;

[0139] Step 2-2: Disconnect the k-th switch in the ring topology, calculate the power loss value, and determine whether each node in the ring topology has a path to the substation and whether it satisfies the inherent constraints. If each node has a path to the substation and satisfies the inherent constraints, proceed to step 2-3. If there are nodes that do not have a path to the substation or do not satisfy the inherent constraints, set the objective function value to a large positive value (10). 9 );

[0140] Steps 2-3: After parallel distributed computation, save the current power loss value and disconnect the (k+1)th switch;

[0141] Step 2-4: Determine if k > K is satisfied. If k > K is satisfied, find the switch with the minimum power loss value as the tie switch. When the number of iterations reaches the number of branches in the ring topology, set all the tie switches to open to form a radial network topology. If k > K is not satisfied, go to step 2-2.

[0142] The graph calculation state transition area is set as follows:

[0143] S3-1. Input the ring topology, convert all segment switches to tie switches, and use a multi-stage reconstruction method based on the improved optimal flow pattern method and branch exchange method to obtain the latest topology. Record the segment switches other than types I-III in the current topology as set Ω1, and the segment switches of type II as set Ω2. Type I is where the shortest path from the current node to the substation node in the ring topology consists of n1 or less segment switches; Type II is where the shortest path from the current node to the substation node in the ring topology consists of n2 or less segment switches; Type III is segment switches not in the switch closed loop.

[0144] S3-2. For the segmented switches excluding set Ω1, open one segmented switch at a time in sequence and close all other switches, and execute the process of obtaining the radial network topology, while keeping the segmented switch open until the radial network topology is formed. Save this radial network topology and the corresponding power loss value.

[0145] S3-3. Perform on-off action on the segmented switches within set Ω2. The on-off action is to close a connecting switch for each type II segmented switch that is opened, in order to maintain connectivity. For example, when the switches at nodes 5 and 6 are opened in Figure 2, connecting switch 6-7 needs to be closed to ensure connectivity.

[0146] S3-4. Execute the available switching actions, solve the optimal power loss function in parallel distributed computing, record the topology and power loss value after this switching action, determine whether this switching action reduces power loss, and if it reduces power loss, save the topology and corresponding power loss value obtained after this switching action.

[0147] S3-5. Perform step S3-4 on all segmented switches in set Ω2;

[0148] S3-6. Save the different topologies obtained after executing different on-off actions and the power loss value corresponding to each topology, and close the open / closed segment switches.

[0149] S3-7. Repeat steps S3-2 to S3-6 for the segmented switches in the ring topology.

[0150] The other technical details and methods of this system are the same and will not be repeated in this specification.

[0151] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, portable compact disc read-only memory (CD-ROM), optical storage, etc.) containing computer-usable program code.

Claims

1. A multi-stage topology reconfiguration method for a power distribution network based on graph computation, comprising: dividing the power distribution network to be reconfigured into a plurality of network regions using the space or structure of the power distribution network; obtaining different topologies and power loss values corresponding to each topology for each network region using a multi-stage reconfiguration method based on an improved optimal flow pattern method and a branch exchange method; determining whether each topology and the power loss value corresponding to each topology meet preset conditions of a power distribution network reconfiguration model, the preset conditions including an objective function with a minimum power loss as a target and inherent constraints; in response to the power loss value corresponding to the topology meeting the preset conditions, saving the corresponding topology of each network region; inputting each divided network region into a graph computation library and performing graph computation iterative training using a synchronous alternating direction multiplier method until the switches in all regions of the power distribution network complete the reconfiguration process, and outputting the reconfiguration result of the power distribution network; the multi-stage reconfiguration method based on the improved optimal flow pattern method and the branch exchange method comprises: closing all switches of the initial topology of the power distribution network to form a ring loop topology structure; setting the number of iterations equal to the number of branches in the ring loop topology structure, and in each iteration, sequentially opening the switches in the ring loop topology structure, performing parallel distributed calculation of the power loss value, selecting the switch with the minimum power loss value as a tie switch, and after the iteration ends, opening all the tie switches obtained to obtain a radial network topology; using a switch pre-classification method to perform state conversion of the sectional switches and tie switches of the obtained radial network topology to obtain the topology structure and power loss value under each state conversion, and obtaining the optimal topology structure and power loss value through parallel distributed calculation.

2. The graph-computing-based power distribution network multi-stage topology reconfiguration method of claim 1, wherein, in each iteration, sequentially opening the switches in the ring loop topology structure, performing parallel distributed calculation of the power loss value, selecting the switch with the minimum power loss value as a tie switch, and after the iteration ends, opening all the tie switches obtained to obtain a radial network topology, comprising: inputting the ring loop topology structure corresponding to the current network region, finding all switches K, and initializing the current switch k; opening the kth switch in the ring loop topology structure, calculating the power loss value, and determining whether each node in the ring loop topology structure has a path to the substation and whether the inherent constraints are met; in response to each node having a path to the substation and meeting the inherent constraints, saving the current power loss value and opening the k+1th switch; in response to there being a node without a path to the substation or not meeting the inherent constraints, setting the value of the objective function to a large positive number; determining whether k > K; in response to k > K, finding the switch with the minimum power loss value as a tie switch; when the number of iterations reaches the number of branches in the ring loop topology structure, setting all the tie switches obtained to be open to form a radial network topology.

3. The graph-computing-based power distribution network multi-stage topology reconfiguration method of claim 2, wherein, The obtained radial network topology adopts a switch pre-classification method to perform state conversion of sectional switches and tie switches, to obtain a topology structure and a power loss value under each state conversion, and to obtain an optimal topology structure and power loss value through parallel distributed computing, including: Input a ring loop topology structure, change all sectional switches to tie switches, and obtain the latest topology structure by using a multi-stage reconstruction method based on an improved optimal flow mode method and a branch exchange method, record sectional switches other than types I-III in the current topology structure as a set Ω1, and sectional switches of type II as a set Ω2; wherein, the type I is that a shortest path from a current node to a substation node in the ring loop topology structure is composed of n1 or less sectional switches; the type II is that a shortest path from a current node to a substation node in the ring loop topology structure is composed of n2 or less sectional switches; and the type III is a sectional switch not in a switch closed loop; For sectional switches other than the set Ω1, sequentially disconnect one sectional switch each time, and close all other switches, execute the flow of obtaining the radial network topology, and keep the sectional switch always disconnected until a radial network topology is formed, and save the radial network topology and the corresponding power loss value; Perform an open-close action on sectional switches in the set Ω2; wherein, the open-close action is to close a tie switch each time a sectional switch of type II is disconnected; Perform the available open-close action, solve a power loss optimal function, record a topology structure and a power loss value after the open-close action, judge whether the open-close action reduces the power loss, and in response to the power loss being reduced, save the topology structure and the corresponding power loss value obtained after the open-close action; Save different topology structures obtained after performing different open-close actions and the corresponding power loss values of each topology structure, and close the disconnected sectional switch.

4. The graph-computing-based power distribution network multi-stage topology reconfiguration method of claim 1, wherein, The objective function targeting the minimization of power loss is expressed as: where: p ij , q ij , r ij are the active and reactive power from node i to node j, and the branch resistance between node i and node j, v j , N are the voltage and the number of load nodes at node j, respectively.

5. The graph-computing-based power distribution network multi-stage topology reconfiguration method of claim 4, wherein, The inherent constraints include: Current constraint, represented as: Wherein, formula (2a) represents the active power balance of the node, p demand,i v i G represents the active power and voltage at node i, respectively. ij B ij q are the real and imaginary parts of the admittance matrix in the i-th row and j-th column, respectively. Ωb is the set of all branches. Formula (2b) represents the reactive power balance of the node. demand,i The reactive power of node i is given by equation (2c), which represents the maximum transmission capacity limit between node i and node j. ij It is the maximum apparent power; The power distribution network current constraint, voltage constraint, and microturbine constraint are expressed as: where (3a) is the distribution network voltage constraint, v i is the voltage of node i, v i,max , v i,min are the upper and lower voltage limits of node i, respectively, (3b) is the branch current constraint from node i to node j, I ij is the branch current from node i to node j, I ij,min , I ij,max are the upper and lower branch current limits, respectively, (3c) is the microturbine active power output constraint, p g(i) is the microturbine active power of node i, p g(i),max , p g(i),min are the upper and lower microturbine active power limits of node i, respectively, (3d) is the microturbine reactive power output constraint, q g(i) is the microturbine reactive power of node i, q g(i),max , q g(i),min are the upper and lower microturbine reactive power limits of node i, respectively.

6. The graph-computing-based power distribution network multi-stage topology reconfiguration method of claim 5, wherein, In the process of power distribution network reconstruction, the radiation and connectivity constraints need to be met, which are expressed as: where: n b is the number of branches, dis(i, i ge ) is the distance from node i to substation node i ge .

7. The graph-computing-based power distribution network multi-stage topology reconfiguration method of claim 6, wherein, The topology structure of each divided power grid region satisfying the preset condition is input to a graph calculation library, graph calculation iterative training is performed by using a synchronous alternating direction multiplier method, until the reconstruction process of all switches in the power grid region of the power distribution network is completed, and a result of multi-stage topology reconstruction of the power distribution network is output, including: Nodes and edges of the topology structure of each divided power grid region satisfying the preset condition are described in the form of a matrix by using the principle of graph theory; wherein, the nodes are connected through the edges, and communication information is exchanged through the edges; A compact form of power loss for distribution network reconfiguration is obtained, expressed as follows: where: f is an objective function expression, x n is the decision variable for the nth grid zone; Define helper functions: where: z n is the state vector of node n; Using the alternating direction method of multipliers, equations (4a), (4b) are transformed into: Transform equation (6) into Lagrange function: In the formula, w=l / ρ, l and ρ are respectively a Lagrange multiplier and a penalty factor. Optimized data of the graph database are subjected to distributed computing, and a synchronous alternating direction multiplier method is introduced into power distribution network reconstruction, including: Transforming equation (7) into: In the formulae: exchange power for the nth power grid region, The exchange power of the nth power grid region in the kth iteration; According to the definition of the function h(z n The final iterative form of the synchronous alternating direction multiplier method is expressed as: In the formulae: The average exchange power of the kth iteration; In response to L non-zero eigenvalues in a Laplace matrix, an average value of power exchange of the nodes is obtained after L iterations; Definition of the residual at iteration k+1 When all power grid regions of the power distribution network converge, the power distribution network converges, and a result of multi-stage topology reconstruction of the power distribution network is output.

8. A power distribution network multi-stage topology reconstruction system based on graph computation, comprising: a partition module configured to divide a power distribution network to be reconstructed into a plurality of grid areas using the space or structure of the power distribution network; a topology structure setting module configured to obtain different topology structures and power loss values corresponding to each topology structure for each grid area by using a multi-stage reconstruction unit based on an improved optimal flow pattern method and a branch exchange method; a graph computation judgment module configured to judge whether each topology structure and the power loss value corresponding to each topology structure satisfy preset conditions of a power distribution network reconstruction model, the preset conditions including an objective function with a minimum power loss as a target and inherent constraints; and in response to the power loss value corresponding to a topology structure satisfying the preset conditions, saving the corresponding topology structure of each grid area; a multi-stage reconstruction module configured to input the topology structure of each divided grid area satisfying the preset conditions to a graph computation library, and perform graph computation iterative training by using a synchronous alternating direction multiplier method until the switches in all grid areas of the power distribution network complete the reconstruction process, and output the result of the multi-stage topology reconstruction of the power distribution network; wherein the multi-stage reconstruction unit based on the improved optimal flow pattern method and the branch exchange method comprises: a loop topology structure area configured to close all switches of an initial topology of the power distribution network to form a loop loop topology structure; a graph computation iteration area configured to set the number of iterations equal to the number of branches in the loop loop topology structure, and in each iteration, sequentially open the switches in the loop loop topology structure, perform parallel distributed computation of power loss values, select the switch with the minimum power loss value as a tie switch, and after the iteration ends, open all tie switches obtained to obtain a radial network topology; a graph computation state conversion area configured to perform state conversion of sectional switches and tie switches on the obtained radial network topology by using a switch pre-classification method to obtain the topology structure and power loss value under each state conversion, and obtain the optimal topology structure and power loss value by parallel distributed computation.

9. The graph-computing-based power distribution network multi-stage topology reconfiguration system of claim 8, wherein, The graph computation iteration area is configured to: input the loop loop topology structure corresponding to the current grid area, find all switches K, and initialize the current switch k; open the kth switch in the loop loop topology structure, calculate the power loss value, and judge whether each node in the loop loop topology structure has a path to the substation and satisfies the inherent constraints; in response to each node having a path to the substation and satisfying the inherent constraints, save the current power loss value, and open the k+1th switch; in response to there being a node without a path to the substation or not satisfying the inherent constraints, set the value of the objective function to a large positive number; judge whether k>K is satisfied; in response to k>K, find the switch with the minimum power loss value as the tie switch; and when the number of iterations reaches the number of branches in the loop loop topology structure, set all tie switches obtained to be open to form a radial network topology.

10. The graph-computing-based power distribution network multi-stage topology reconfiguration system of claim 9, wherein, The graph computation state conversion area is configured to: An input ring topology is adopted, all sectional switches are changed into tie switches, and a multi-stage reconstruction method based on an improved optimal flow mode method and a branch exchange method is used to obtain a latest topology, sectional switches other than types I-III in the current topology are recorded as set Ω1, and sectional switches of type II are recorded as set Ω2; wherein, the type I is that a shortest path from a current node to a substation node in the ring topology is composed of n1 or less sectional switches; the type II is that a shortest path from a current node to a substation node in the ring topology is composed of n2 or less sectional switches; and the type III is a sectional switch not in a closed switch loop; For sectional switches other than set Ω1, one sectional switch is sequentially opened each time, and all other switches are closed, the flow of obtaining a radial network topology is executed, and the sectional switch is kept open until a radial network topology is formed, the radial network topology and a corresponding power loss value are saved; An open-close action is performed on sectional switches in set Ω2; wherein, the open-close action is that one type II sectional switch is opened, and one tie switch is closed each time; The available open-close action is executed, a power loss optimal function is solved, a topology after the open-close action and a power loss value are recorded, and whether the open-close action reduces the power loss is judged; in response to the power loss being reduced, the topology and the corresponding power loss value obtained after the open-close action are saved; different topologies obtained after different open-close actions and corresponding power loss values of each topology are saved, and the opened sectional switch is closed.

Images