Power distribution network key load recovery method based on graph calculation

Abstract

A power distribution network key load recovery method based on graph calculation comprises the steps that a power distribution network is modeled into an undirected graph G = (V, E), V represents a node set and comprises all loads and distributed power sources in the power distribution network, and the loads are divided into key loads and non-key loads; e represents an edge set and comprises switches in the power distribution network; determining a shortest path from each distributed power supply to each key load by using a Dijkstra algorithm; constructing feasible recovery trees of the nodes of the distributed power supply, determining each recovery tree by a combination of two or more key loads related to the distributed power supply, and performing load flow calculation to check the recovery feasibility of the key loads in each recovery tree; the optimal recovery tree combination is obtained by optimizing the load recovery model and utilizing a mixed integer linear programming method to solve. The method has the advantages that when the main power supply is interrupted, the recovery of the key load is optimized, the switching operation frequency is reduced, and the implementation efficiency is improved.

Description

Technical Field

[0001] The invention relates to an evaluation method, and in particular to a distribution network key load recovery method based on graph calculation. Background Art

[0002] In the power system, the distribution network, as the terminal link of power transmission and distribution, bears the important task of safely and reliably delivering electric energy from the transmission system to users. However, with the rapid growth of power load, the power supply reliability of the distribution network has become increasingly prominent, especially in the event of a permanent fault. How to quickly and accurately locate and isolate the fault area, and effectively transfer the load of the non-fault power-off section to other normally operating feeders, has become a key issue that the power industry needs to solve.

[0003] As an important part of new energy, distributed generation is increasingly used in distribution networks. In extreme events, when the main power facilities of the distribution network are unavailable and the fault lasts for a certain period of time, distributed generation can play an important role in supplying power to critical loads. However, the capacity of distributed generation is limited, and each distributed generation must first provide its local load. If there is additional capacity, it can be used to restore the critical loads of the power grid. Therefore, during the restoration of critical loads, the capacity limitations of distributed generation and network operation and topology constraints need to be comprehensively considered to ensure the feasibility and effectiveness of restoration.

[0004] Most of the existing research on distribution network power restoration is based on searching for restoration paths using adjacency matrices. However, it has obvious deficiencies in describing the topological structure of distribution networks and their dynamic changes. Specifically, the following aspects are required: 1. Traditional models are difficult to accurately describe the dynamic characteristics of distribution networks, such as the rapid changes in topological structure when a fault occurs and the dynamic access and exit of distributed power sources. 2. For large-scale distribution networks, traditional algorithms have low computational efficiency when dealing with complex topological structures and large amounts of data, and cannot meet real-time requirements. 3. Traditional methods often analyze each component or region separately, lacking in-depth exploration of the overall correlation of the distribution network, resulting in inaccurate and incomplete evaluation results. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a distribution network key load recovery method based on graph calculation, which can optimize the recovery of key loads when the main power supply is interrupted, reduce the number of switch operations, and improve implementation efficiency.

[0006] The technical solution of the present invention is:

[0007] A distribution network key load restoration method based on graph computing, comprising:

[0008] A graph theory model is established to model the distribution network as an undirected graph G = (V, E), where V represents the node set, including all loads and distributed power sources in the distribution network, and the loads are divided into critical loads and non-critical loads; E represents the edge set, including switches in the distribution network;

[0009] Use Dijkstra algorithm to determine the shortest path from each distributed power source to each critical load;

[0010] Constructing feasible restoration trees of distributed generation nodes, each restoration tree is determined by a combination of two or more key loads associated with the distributed generation, and performing power flow calculations to verify the feasibility of restoring the key loads in each restoration tree;

[0011] By optimizing the load restoration model and solving it using the mixed integer linear programming (MILP) method, the optimal restoration tree combination is obtained to maximize the energy provided to the critical loads and minimize the number of switching operations.

[0012] Further, the Dijkstra algorithm includes the following steps:

[0013] i) Initialization: Create an unvisited node set, mark all key load nodes as unvisited nodes, assign a temporary distance value to each node, set the distance from the distributed power node to itself to 0, and the distance to all other nodes to infinity, and select any distributed power node as the starting node;

[0014] ii) Select a node: Select a node from the unvisited node set that has the smallest distance to the start node as the current processing node;

[0015] iii) Update distance: For each unvisited neighbor node of the current processing node, calculate the distance from the starting node through the current processing node to the neighbor node, and update the shortest path estimate of the neighbor node;

[0016] iv) Update node: mark the currently processed node as visited and remove the node from the set of unvisited nodes;

[0017] v) Repeat: Repeat steps ii) to iv) until all nodes are visited.

[0018] Furthermore, constructing a feasible recovery tree for a distributed power generation node includes the following steps:

[0019] i) Path combination: for each distributed generation node, all the shortest paths to different critical loads are combined to form a possible restoration tree;

[0020] ii) Load flow calculation, which is performed for each possible restoration tree to verify whether it meets the operating constraints of the distribution network, including voltage limits, line current limits, and power capacity limits of the distributed sources;

[0021] iii) Feasibility check, if a restoration tree satisfies all the operational constraints in the load flow calculation, it is considered feasible; otherwise, it is removed from the set of feasible restoration trees.

[0022] Furthermore, when optimizing the load recovery model, the objective function is determined and two objective functions are defined. The first one is to maximize the energy provided to the critical load, and the second one is to minimize the number of switching operations, which are balanced by weight factors.

[0023] Furthermore, when optimizing the load recovery model, constraints are determined, including that each distributed power source can only be powered by one recovery tree, each critical load can only be powered by one recovery tree, and each non-critical load can only be restored once in the recovery tree.

[0024] The present invention is based on the distribution network graph model, restores power supply to the distribution network after a fault, establishes a power flow calculation model that constrains the operation of the distribution network and adapts to the topology changes of the distribution network under the graph calculation framework, uses the minimum spanning tree theory to define the power outage range and construct a recovery path set, determines the priority of the recovery target according to the degree of correlation between power supply reliability and recovery targets, and establishes a topology evolution model on this basis to optimize the recovery strategy. The model can comprehensively utilize the repeated primary and secondary transfers in the power outage area to restore power supply, and determine the optimal recovery strategy according to the recovery target. Its beneficial effects are:

[0025] (1) Graph computing technology represents data in the form of graphs, which is naturally consistent with the topological structure of the distribution network and can accurately and flexibly represent the complex topological structure of the distribution network.

[0026] (2) When dealing with fault recovery in large-scale distribution networks, the Dijkstra algorithm is used to quickly search for the optimal fault recovery path; and combined with graph computing technology, the nodes and edges in the graph can be processed in parallel, greatly improving computing efficiency.

[0027] (3) By analyzing the electrical connection relationship between each component and the impact propagation path when a fault occurs, and comprehensively considering the overall characteristics of the distribution network, a more reasonable and effective fault recovery plan can be formulated.

[0028] In summary, the present invention takes into account the limited power supply capacity of distributed power sources and the critical loads constrained by topology. The total number of switching operations is optimized, thereby helping power grid dispatchers to implement restoration plans more effectively in a shorter time. The method is applicable to large distribution networks and can maximize the use of the excess capacity of distributed power sources to restore the critical loads of the power grid with the minimum number of switching operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 The equivalent diagram of the 118 bus distribution network used in the present invention (corresponding to Example 1);

[0030] Figure 2 It is a schematic diagram of a distributed power supply of the present invention. DETAILED DESCRIPTION

[0031] Example 1

[0032] The capacity of distributed power sources is utilized to supply power to critical loads in extreme events when the main power facilities of the distribution network are unavailable and the fault lasts for a certain period of time. The capacity of distributed power sources is limited. Each distributed power source must first provide its local load, and if there is extra capacity, it can be used to restore the critical loads of the power grid. During the restoration of critical loads, non-critical loads will also be restored along the path between the power source and each critical load. The main goal of the present invention is to select the correct state of the topological path switches of the distribution network to provide the optimal path for the critical loads that are given priority for the restoration of distributed power sources. To this end, the restoration problem is formulated as an optimization problem subject to the power and energy constraints of distributed power sources and the network operation and topology constraints.

[0033] Step 1: Build a graph theory model

[0034] The distribution network is modeled as an undirected graph G = (V, E), where V represents the node set, including all loads and distributed power sources in the distribution network, and the loads are divided into critical loads and non-critical loads; E represents the edge set, including the operable lines (i.e. switches) in the distribution network; each node represents a load or a distributed power source, each edge represents a line connecting two nodes, and the weight of the edge is the total power of the load restored on the line.

[0035] Step 2: Determine the shortest path from distributed energy to load

[0036] Use Dijkstra algorithm to determine the shortest path from each distributed power source to each critical load;

[0037] Dijkstra's algorithm is an algorithm for finding the shortest path from a single source point to all other nodes in a weighted graph. The core idea of ​​the algorithm is the greedy strategy, that is, at each step, the vertex of the currently known shortest path is selected as the next visited vertex. It gradually finds the shortest path by maintaining a set of nodes to be visited and a set of visited nodes, as well as an estimated value of the shortest path from the initial node to each node. The algorithm steps are as follows:

[0038] (1) Initialize nodes: Create a set of unvisited nodes, mark all key load nodes as unvisited nodes, assign a temporary distance value to each node, set the distance from the distributed power node to itself to 0, and the distance to all other nodes to infinity; select any distributed power node as the starting node.

[0039] (2) Select the current processing node: Select a node with the shortest distance from the start node from the set of unvisited nodes as the current processing node.

[0040] (3) Update the distance between nodes: For each unvisited neighbor node of the currently processed node, calculate the distance from the current node to the neighbor node and update its shortest path estimate.

[0041] (4) Update the node to a visited node: mark the currently processed node as visited and remove it from the set of unvisited nodes.

[0042] (5) Repeat algorithm steps (2)-(4) until all nodes have been visited.

[0043] For each distributed power source and critical load, the Dijkstra algorithm will find a path with the minimum total load power, that is, the shortest path.

[0044] Step 3: Construct a feasible restoration tree for distributed generation nodes

[0045] The recovery tree is used to restore two or more critical loads using distributed power sources. The recovery tree of each microgrid is determined by the combination of two or more feasible recovery paths associated with the distributed power source. Using the feasible recovery paths obtained in the first step, all feasible recovery trees for each distributed power source are determined. Then, the power flow calculation is performed to verify the feasibility of critical load recovery in each feasible recovery tree. The construction steps are as follows:

[0046] (1) Path combination: For each distributed generation node, all the shortest paths leading to different critical loads are combined to form a possible recovery tree.

[0047] (2) Load flow calculation: Load flow calculation is performed for each possible restoration tree to verify whether it meets the operating constraints of the distribution network, including voltage limits, line current limits, and power capacity limits of distributed sources.

[0048] (3) Feasibility check: If a recovery tree satisfies all operational constraints in the load flow calculation, it is considered feasible; otherwise, it is removed from the set of feasible recovery trees.

[0049] Step 4: Optimize the load recovery model

[0050] After all feasible restoration trees have been identified, the next step is to select the best combination of restoration trees to maximize the energy delivered to the critical loads and minimize the number of switching operations. This can be achieved by solving a mixed integer linear programming (MILP) problem.

[0051] First, determine the objective function and define two objective functions. The first is to maximize the energy provided to the critical load, and the second is to minimize the number of switching operations. These two objective functions are weighted by the weight factor W. cj and W sw A balance is struck to ensure that the number of switching operations is minimized while ensuring energy supply.

[0052] Secondly, determine the constraints. The constraints of the objective function are:

[0053] (1) Each distributed power source can only be powered by one recovery tree.

[0054] (2) Each critical load can be powered by only one restoration tree.

[0055] (3) Each non-critical load can only be restored once in the recovery tree to avoid intersections between recovery trees.

[0056] Energy supply objective function

[0057]

[0058] Switching operation times objective function

[0059]

[0060] The variable is defined as:

[0061] j: represents the jth recovery tree, and Guni is the set of recovery trees.

[0062] c: represents the cth critical load, C j represents the set of critical loads in the jth restoration tree.

[0063] Z j: A binary variable 0 or 1, indicating whether the recovery tree j is selected.

[0064] T Rj : The service time of restoring tree j.

[0065] P cj : Rated power of critical load c in restoration tree j.

[0066] W cj : The weight factor of critical load c in the recovery tree j.

[0067] S j : The number of switch operations required to isolate and restore tree j.

[0068] W sw : Represents the switch operation cost factor.

[0069] Finally, the system of equations is solved and a MILP solver, such as the intlinprog function in MATLAB, is used to solve the above optimization problem and obtain the optimal combination of recovery trees.

[0070] The present invention is applied to the IEEE118 bus distribution network model, and utilizes the capacity of distributed power sources to restore critical loads. The load calculation of the restoration path and restoration tree is performed using the Newton-Raphson algorithm through the MatPower 7.1 toolbox.

[0071] The equivalent diagram of the 118 bus distribution network used in Example 1 of the present invention is as follows: Figure 1 As shown. The network contains 12 critical loads on buses 4, 7, 14, 46, 47, 55, 60, 63, 73, 79, 92 and 103. The weight coefficients, active power, reactive power and other parameters of the critical loads are shown in Table 1. Critical loads 7, 47, and 103 have high priority and a weight factor of 3; critical loads 14, 63, and 92 have average priority and a weight factor of 2; critical loads 4, 46, 55, 60, 73, and 79 have low priority and a weight factor of 1. The weight coefficients of non-critical loads are set to zero. Six distributed power sources are connected to buses 27, 28, 62, 65, 77, and 110. Table 2 gives the capacity of the generating sources and their local loads when the power factor is 0.9. The equivalent circuit used by each distributed power source includes a source and a local load, as shown Figure 2 As shown in Figure 1. After an extreme event, the main power supply was interrupted and six faults occurred in the distribution network. The corresponding fault lines are 5-6, 20-21, 25-26, 32-33, 67-68, and 101-102, and the downtime is estimated to be 16 hours.

[0072] Table 1

[0073] Critical load Active power (kW) Q(kVar) Weight Factor CL4 34.315 21.845 1 CL7 104.47 61.725 3 CL14 141.9 117.5 2 CL46 39.653 20.758 1 CL47 66.195 42.361 3 CL55 62.1 26.86 1 CL60 80.551 49.156 1 CL63 478.8 463.74 2 CL73 52.699 22.482 1 CL79 294.55 162.47 1 CL92 114.57 81.748 2 CL103 408.43 168.46 3

[0074] Table 2

[0075]

[0076] For distributed generation sources M1, M2, M3, M4, M5, and M6, the feasible restoration paths associated with each distributed generation can be combined to form 9, 15, 17, 42, 3, and 6 feasible restoration trees. A total of 92 feasible restoration trees are established for all distributed generation sources.

[0077] Next, the power delivered by distributed generation to the critical load is maximized, regardless of the number of switching operations. The restored critical loads, the number of restorations, and the power delivered by each distributed generation to the critical load are shown in Table 4. Table 5 shows the operated sectionalizing and bundling switches, as well as the total number of switching operations required to isolate the restoration tree.

[0078] Critical loads 14 and 46 are restored by distributed power source M1, and critical loads 4 and 63 are restored by distributed power source M2. Similarly, critical loads 7, 47, 55, and 60 are restored by distributed power source M3, critical loads 63 and 92 are restored by distributed power source M4, critical load 79 is restored by distributed power source M5, and finally, critical load 103 is restored by distributed power source M6. In this case, the power required by all critical loads is provided by any distributed power source.

[0079] As shown in Tables 3 and 4, the distributed power source M3 supplies three key loads 7, 47 and 60, and its maximum number of switching operations is 10 times, including opening 6 section switches and closing 4 grid-connected switches. As shown in Table 5, the total number of switching operations required for all distributed power sources to restore key loads is 40 times.

[0080] Table 3

[0081]

[0082]

[0083] Table 4

[0084] Distributed Power Restoring critical loads Restored electricity (MWh) M1 CL14,CL46 2.48 M2 CL4,CL63 5.73 M3 CL7,CL47,CL55,CL60 4.83 M4 CL73,CL92 2.14 M5 CL79 3.09 M6 CL103 4.5

[0085] Table 5

[0086] Distributed Power Section switch operation Combined switch operation Switching Operation M1 13-14,26-27,45-46 TM1,T1,T5 6 M2 28-29,4-5,2-10,2-3,1-100,63-64 TM2 7 M3 29-55,49-50,34-35,23-24,8-9,6-7 TM3,T8,T4,T2 10 M4 64-65,65-66,72-73,73-74,92-93,91-96 TM4,T11 8 M5 74-75,78-79,79-80 TM5,T12 5 M6 111-112,102-103,110-111 TM6 4

[0087] Based on graph shortest path theory, a resilience-oriented approach is proposed to utilize the capacity of distributed generation for critical load restoration after extreme events. The goal of restoration is to maximize the energy delivered to critical loads and minimize the number of switching operations while considering load priorities.

[0088] The method includes using Dijkstra algorithm to find the shortest feasible restoration path, using the feasible restoration path to form a feasible restoration tree, and using the MILP optimization method to optimally select the restoration tree. The simulation results of the 118-bus distribution network show that the method can utilize the remaining capacity of the distributed power source, observe the operation constraints of the power grid, and restore the total critical load of 22.77MWh. The simulation results prove the effectiveness of the proposed method.

[0089] The above are only specific embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A distribution network key load restoration method based on graph computing, characterized by: include: A graph theory model is established to model the distribution network as an undirected graph G = (V, E), where V represents the node set, including all loads and distributed power sources in the distribution network, and the loads are divided into critical loads and non-critical loads; E represents the edge set, including switches in the distribution network; Use Dijkstra algorithm to determine the shortest path from each distributed power source to each critical load; Constructing feasible restoration trees of distributed generation nodes, each restoration tree is determined by a combination of two or more key loads associated with the distributed generation, and performing power flow calculations to verify the feasibility of restoring the key loads in each restoration tree; By optimizing the load restoration model and solving it using the mixed integer linear programming method, the optimal restoration tree combination is obtained to maximize the energy provided to the critical loads and minimize the number of switching operations.

2. The method for restoring critical loads in a distribution network based on graph computing according to claim 1 is characterized in that: The Dijkstra algorithm consists of the following steps: i) Initialization: Create an unvisited node set, mark all key load nodes as unvisited nodes, assign a temporary distance value to each node, set the distance from the distributed power node to itself to 0, and the distance to all other nodes to infinity, and select any distributed power node as the starting node; ii) Select a node: Select a node from the unvisited node set that has the smallest distance to the start node as the current processing node; iii) Update distance: For each unvisited neighbor node of the current processing node, calculate the distance from the starting node through the current processing node to the neighbor node, and update the shortest path estimate of the neighbor node; iv) Update node: Mark the current processing node as visited and remove the node from the unvisited node set; v) Repeat: Repeat steps ii) to iv) until all nodes are visited.

3. The method for restoring critical loads in a distribution network based on graph computing according to claim 1 is characterized in that: Constructing a feasible restoration tree for a distributed generation node includes the following steps: i) Path combination: for each distributed generation node, all the shortest paths to different critical loads are combined to form a possible restoration tree; ii) Load flow calculation, which is performed for each possible restoration tree to verify whether it meets the operating constraints of the distribution network, including voltage limits, line current limits, and power capacity limits of the distributed sources; iii) Feasibility check, if a restoration tree satisfies all the operational constraints in the load flow calculation, it is considered feasible; otherwise, it is removed from the set of feasible restoration trees.

4. The method for restoring critical loads in a distribution network based on graph computing according to claim 1 is characterized in that: When optimizing the load recovery model, the objective function is determined and two objective functions are defined. The first one is to maximize the energy provided to the critical load, and the second one is to minimize the number of switching operations, which are balanced by weight factors.

5. The method for restoring critical loads in a distribution network based on graph computing according to claim 1 is characterized in that: When optimizing the load restoration model, the constraints are determined, including the constraints that each distributed power source can only be powered by one restoration tree, each critical load can only be powered by one restoration tree, and each non-critical load can only be restored once in the restoration tree.

Images