US-12621213-B2 - Method and system for multi-stage topology reconfiguration of distribution network based on graph computing

US12621213B2US 12621213 B2US12621213 B2US 12621213B2US-12621213-B2

Abstract

Provided are a method for multi-stage topology reconfiguration of a distribution network based on graph computing and a system for multi-stage topology reconfiguration of a distribution network based on graph computing. The method includes providing a distribution network reconfiguration method based on the optimal flow pattern method and the branch exchange method; constructing a multi-stage reconfiguration model for the distribution network with an optimal power loss model; providing a multi-stage reconfiguration calculation method for the distribution network based on graph computing; and solving the optimal power loss model based on graph computing to obtain the distribution network reconfiguration technique.

Inventors

Zijian Hu
Yandi WANG
Qilong QIAN
Ziqiang Xu
Ye Ji
Mingming Cai
Hong Zhu
Xiaofeng Wang
Zhan LV
Honghua XU
Weiliang Wang
Zhiwei Zhang
Jianhua Zhou
Xiaodong Cao

Assignees

STATE GRID NANJING SUPPLY POWER COMPANY

Dates

Publication Date: 20260505
Application Date: 20250609
Priority Date: 20240723

Claims (6)

1 . A method for multi-stage topology reconfiguration of a distribution network based on graph computing, comprising: dividing a to-be-reconfigured distribution network into a plurality of grid regions by using a space of the distribution network or a structure of the distribution network; obtaining different topology structures and a power loss value corresponding to each topology structure of the different topology structures by applying a multi-stage reconfiguration method based on an improved optimal flow pattern method and a branch exchange method to each grid region of the plurality of grid regions; determining whether each topology structure and the power loss value corresponding to each topology structure satisfy a preset condition of a distribution network reconfiguration model, wherein the preset condition comprises: a target function aimed at minimizing power loss and inherent constraints; and in response to a power loss value corresponding to a topology structure satisfying the preset condition, storing the topology structure for each grid region; and inputting each divided grid region into a graph computing library, performing iterative training for graph computing by using a synchronous alternating direction method of multipliers until a reconfiguration process for switches in all the plurality of grid regions in the distribution network is completed, and outputting a result of the multi-stage topology reconfiguration of the distribution network; wherein the multi-stage reconfiguration method based on the improved optimal flow pattern method and the branch exchange method comprises: closing all switches in an initial topology of the distribution network to form a ring-shaped loop topology structure; setting a number of iterations to a number of branches in the ring-shaped loop topology structure, in each of the iterations, sequentially opening the switches in the ring-shaped loop topology structure, calculating power loss values in a parallel distributed manner, and selecting a switch with a minimum power loss value as a tie switch; and after the iterations are completed, opening all obtained tie switches to acquire a radial network topology; and performing state transitions on sectionalizing switches and tie switches of the acquired radial network topology by using a switch pre-classification method to obtain a topology structure and a power loss value for each of the state transitions, and acquiring an optimal topology structure and a power loss value through parallel distributed computing; wherein performing the state transitions on the sectionalizing switches and the tie switches of the acquired radial network topology by using the switch pre-classification method to obtain the topology structure and the power loss value for each of the state transitions, and acquiring the optimal topology structure and the power loss value through the parallel distributed computing comprises: inputting the ring-shaped loop topology structure, converting all sectionalizing switches to tie switches, acquiring a latest topology structure by applying the multi-stage reconfiguration method based on the improved optimal flow pattern method and the branch exchange method, recording sectionalizing switches other than sectionalizing switches of types I to III in the current topology structure as a set Ω 1 , and recording sectionalizing switches of type II as a set Ω 2 ; wherein type I refers to a sectionalizing switch to which a shortest path from a current node to a substation node in the ring-shaped loop topology structure consists of no more than n 1 sectionalizing switches belongs; type II refers to a sectionalizing switch to which a shortest path from the current node to the substation node in the ring-shaped loop topology structure consists of no more than n 2 sectionalizing switches belongs; and type III refers to a sectionalizing switch that is not in a closed switch loop; for sectionalizing switches outside the set Ω 1 , sequentially opening one of the sectionalizing switches at a time, closing all other switches, executing a process for acquiring the radial network topology, keeping the one sectionalizing switch open until the radial network topology is formed, and storing the radial network topology and a corresponding power loss value; performing an open-close action on the sectionalizing switches of type II in the set Ω 2 ; wherein the open-close action comprises opening one of the sectionalizing switches of type II and simultaneously closing one tie switch; executing an available open-close action, solving a power loss optimization function, recording a topology structure and a power loss value after the available open-close action, and determining whether the available open-close action reduces power loss; and in response to a reduction in the power loss, storing the topology structure and the corresponding power loss value obtained after the available open-close action; and storing different topology structures obtained after executing different open-close actions and a power loss value corresponding to each topology structure of the different topology structures, and closing opened sectionalizing switches.
2 . The method for multi-stage topology reconfiguration of the distribution network based on graph computing according to claim 1 , wherein in each of the iterations, sequentially opening the switches in the ring-shaped loop topology structure, calculating the power loss values in the parallel distributed manner, and selecting the switch with the minimum power loss value as the tie switch; and after the iterations are completed, opening all the obtained tie switches to acquire the radial network topology comprises: inputting a ring-shaped loop topology structure corresponding to a current grid region, identifying all K switches, and initializing a current switch k; opening a k-th switch in the ring-shaped loop topology structure, calculating a power loss value, and determining whether each node in the ring-shaped loop topology structure has a path to a substation and satisfies the inherent constraints; in response to each node having the path to the substation and satisfying the inherent constraints, storing the current power loss value and opening a (k+1)-th switch; and in response to a node lacking the path to the substation or failing to satisfy the inherent constraints, setting a value of the target function to a large positive number; and determining whether k is greater than K; in response to k>K, finding a minimum power loss value and setting a corresponding branch switch to an open state to form the radial network topology.
3 . The method for multi-stage topology reconfiguration of the distribution network based on graph computing according to claim 1 , wherein the target function aimed at minimizing power loss is expressed as: - C = min ⁢ ∑ i = 1 , j = 1 N ⁢ ( p ij + q ij v j ) 2 ⁢ r ij ⁢ ∀ i , ∀ j ∈ N , ( 1 ) wherein p ij , q ij , and r ij denote active power between a node i and a node j, reactive power between the node i and the node j, and branch resistance between the node i and the node j, respectively, and v j and N denote a voltage of the node j and a number of load nodes, respectively.
4 . The method for multi-stage topology reconfiguration of the distribution network based on graph computing according to claim 3 , wherein the inherent constraints comprise: power flow constraints, expressed as: { p demand , i + ∑ i , j ∈ Ω b ⁢ v i ⁢ v j ( G ij ⁢ cos ⁢ θ ij + B ij ⁢ sin ⁢ θ ij ) = 0 ⁢ ( 2 ⁢ a ) q demand , i + ∑ i , j ∈ Ω b ⁢ v i ⁢ v j ( G ij ⁢ sin ⁢ θ ij - B ij ⁢ cos ⁢ θ ij ) = 0 ⁢ ( 2 ⁢ b ) ( v i ⁢ v j ) 2 [ ( G ij ⁢ cos ⁢ θ ij + B ij ⁢ sin ⁢ θ ij ) 2 + ( G ij ⁢ sin ⁢ θ ij - B ij ⁢ cos ⁢ θ ij ) 2 ] ≤ S ij _ 2 , ( 2 ⁢ c ) wherein formula (2a) represents an active power balance of a node, which is a prerequisite for maintaining frequency stability in the distribution network, p demand,i and v i denote active power of the node i and a voltage of the node i, respectively, G ij and B ij denote a real part and an imaginary part of an admittance matrix at an i th row and a j th column, respectively, and fb denotes a set of all branches; formula (2b) represents a reactive power balance of the node, which is a prerequisite for maintaining voltage stability in the distribution network; and g demand,i denotes reactive power of the node i; formula (2c) represents a maximum transmission capacity limit between the node i and the node j, and S ij denotes a maximum apparent power; and distribution network current constraints, voltage constraints, and micro-turbine constraints, expressed as: { v i , min ≤ v i ≤ v i , max ⁢ ( 3 ⁢ a ) I ij , min ≤ I ij ≤ I ij , max ( 3 ⁢ b ) p g ⁡ ( i ) , min ≤ p g ⁡ ( i ) ≤ p g ⁡ ( i ) , max ⁢ ( 3 ⁢ c ) q g ⁡ ( i ) , min ≤ q g ⁡ ( i ) ≤ q g ⁡ ( i ) , max , ( 3 ⁢ d ) wherein formula (3a) represents a voltage constraint of the distribution network, ensuring a safe operation of the distribution network and electrical equipment, v i denotes a voltage of the node i, v i,max and v i,min denote an upper voltage limit of the node i and a lower voltage limit of the node i, respectively, formula (3b) represents a branch current constraint from the node i to the node j, ensuring safety of lines, I ij denotes a branch current from the node i to the node j, I ij,min and I ij,max denote an upper current limit of a branch and a lower current limit of the branch, respectively, formula (3c) represents an active power output constraint of a micro-turbine, p g(i) denotes active power of the micro-turbine at the node i, p g(i),max and p g(i),min denote an upper active power limit of the micro-turbine at the node i and a lower active power limit of the micro-turbine at the node i, respectively, formula (3d) represents a reactive power output constraint of the micro-turbine, q g(i) denotes reactive power of the micro-turbine at the node i, and q g(i),max and q g(i),min denote an upper reactive power limit of the micro-turbine at the node i and a lower reactive power limit of the micro-turbine at the node i, respectively.
5 . The method for multi-stage topology reconfiguration of the distribution network based on graph computing according to claim 4 , wherein radiality and connectivity constraints satisfied during a process of the multi-stage topology reconfiguration of the distribution network are expressed as follows: { N - n b = 1 ⁢ ( 4 ⁢ a ) dis ⁢ ( i , i ge ) < ∞ ⁡ ( i ≠ i ge ) , ( 4 ⁢ b ) wherein n b denotes a number of branches, and dis(i, i ge ) denotes a distance from the node i to a substation node i ge .
6 . The method for multi-stage topology reconfiguration of the distribution network based on graph computing according to claim 5 , wherein inputting the topology structure satisfying the preset condition for each divided grid region into the graph computing library, performing the iterative training for graph computing by using the synchronous alternating direction method of multipliers until the reconfiguration process for the switches in all the grid regions of the distribution network is completed, and outputting the result of the multi-stage topology reconfiguration of the distribution network comprises: describing, by using a principle of graph theory, nodes and edges of the topology structure satisfying the preset condition for each divided grid region in a matrix form; wherein the nodes are connected through the edges, and communication information is exchanged through the edges; wherein the distribution network consists of nodes and branches, the nodes are connected by the branches to transmit information, and the nodes and the branches of the distribution network form basic concepts of graph computing; by using basic principles of the graph computing, a structure of the distribution network is transformed into graph relationships, the nodes are set as vertices, and optimization data are stored in a graph database for parallel computing to improve computational efficiency; and each vertex is configured to perform optimal parallel optimization calculations for distribution network power losses based on information transmitted from adjacent vertices and update boundary information based on an optimization result; the acquired topology structure is transformed into the graph relationships, each node in the topology structure is set as a vertex, and optimization data are stored in the graph database to improve computational efficiency; and each vertex is configured to perform optimal parallel optimization calculations for distribution network power losses based on information transmitted from adjacent vertices and update boundary information based on an optimization result; acquiring a compact form of power loss for the multi-stage topology reconfiguration of the distribution network, expressed as: { min f ⁡ ( x n ) s . t . ( 2 ) ~ ( 4 ) , ( 5 ) wherein ƒ is an expression of the target function, and x n denotes a decision variable for an n-th grid region; performing distributed computing on optimized data in the graph database and introducing the synchronous alternating direction method of multipliers into the distribution network reconfiguration; expressing, according to a definition of a function h(z n ), a final iterative form of the synchronous alternating direction method of multipliers as follows: { x n k + 1 = arg min x n { f ⁡ ( x n ) + τ 2 ⁢  x n ex - x n ex , k + x _ ex , k + w k  2 2 } w k + 1 = w k + x _ ex , k + 1 , wherein τ and w denote a Lagrange penalty factor and a Lagrange multiplier respectively, x n ex denotes exchanged power of an n-th grid region, x n ex , k denotes exchanged power of the n-th grid region at a k-th iteration, and x _ ex , k denotes average exchanged power at the k-th iteration; a principle of discrete consistency theory, in response to a Laplacian matrix having L non-zero eigenvalues, obtaining an average value of power exchange of a node after L iterations; and defining a residual that ϛ k + 1 =  x n k + 1 - x n k  2 2 ≤ ϛ min at a (k+1)-th iteration, and outputting the result of the multi-stage topology reconfiguration of the distribution network until all the plurality of grid regions in the distribution network converge and the distribution network converges.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS This is a continuation of International Patent Application No. PCT/CN2025/090377, filed on Apr. 22, 2025, which claims priority to Chinese Patent Application No. 202410985204.9 filed with the China National Intellectual Property Administration (CNIPA) on Jul. 23, 2024, the disclosures of which are incorporated herein by reference in their entireties. TECHNICAL FIELD This application relates to the technical field of energy storage in distribution networks, for example, to a method for multi-stage topology reconfiguration of a distribution network based on graph computing and a system for multi-stage topology reconfiguration of a distribution network based on graph computing. BACKGROUND The distribution network is one of the fundamental components of the power grid. It involves all aspects of user electricity consumption, featuring a vast and complex structure. The operational quality of the distribution network directly affects the quality of electricity supply to users. The primary function of the distribution network is to transmit 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 increases, the system scale expands, and computations become increasingly complex. Traditional centralized computing has poor scalability and high communication costs, making it difficult to meet the computational demands of distribution networks. In the field of distribution network research, since the impedance of branches is significantly greater than resistance, reducing network losses has always been a critical issue to address. Distribution network reconfiguration, by altering the network topology, that is, by changing the open and closed states of sectionalizing switches and tie switches between nodes, transfers load nodes and reduces power losses. Current research on distribution network reconfiguration primarily focuses on three methods, that is, a mathematical optimization method, a heuristic method, and an intelligent optimization algorithm. Although mathematical optimization can quickly find an optimal solution, it is not directly applicable to non-convex nonlinear problems and is overly dependent on initial values. The heuristic method is simple and capable of high-quality searches for optimal solutions, but as the number of nodes increases, the workload grows and computational efficiency decreases. The intelligent optimization algorithm, while effective for small-scale systems, struggles to converge to optimal solutions in large-scale node systems. Additionally, with rising communication costs and growing user demands for privacy, efficiently and securely reducing network losses is a significant challenge for distribution network reconfiguration. For example, Chinese Patent Application No. 202310097347.1 discloses a distributed photovoltaic capacity allocation method considering distribution network reconfiguration, which uses a Cuckoo Search Algorithm to achieve a more reasonable capacity configuration of the distribution network and employs the optimal flow pattern method for reconfiguration. Chinese Patent Application No. 202111106696.2 discloses a method for fault handling and network reconfiguration suitable for a distribution network with strong uncertainty, which achieves optimized post-fault network reconfiguration through the branch exchange method. However, these methods have the following shortcomings: (1) As the number of nodes increases, the workload becomes excessive, and computational efficiency decreases. (2) These methods cannot accurately and effectively perform distribution network reconfiguration and are influenced by the original topology structure. Specifically, the optimal flow pattern method has a slow solving speed and struggles to search for the global optimum due to mutual current influence. The branch exchange method leads to a high workload and low computational efficiency as the number of nodes grows, and the power flow calculations are affected by non-independent topologies. (3) During distribution network reconfiguration, these methods cannot securely protect user information while effectively reducing network losses. SUMMARY This application provides a method for multi-stage topology reconfiguration of a distribution network based on graph computing and a system for multi-stage topology reconfiguration of a distribution network based on graph computing. In a first aspect, this application provides a method for multi-stage topology reconfiguration of a distribution network based on graph computing. The method includes the steps below. A to-be-reconfigured distribution network is divided into multiple grid regions by using a space or structure of the distribution network. Different topology structures and a power loss value corresponding to each topology structure are obtained by applying a multi-stage reconfiguration