Power graph database query acceleration method based on graph partitioning and node routing

Abstract

The application discloses a power graph database query acceleration method based on graph division and node routing, belongs to the computer technical field, and aims to solve the problems that the performance of the existing research cannot adapt to the increasingly expanded power system and the cost caused by the storage cost is not considered; the specific steps are as follows: step 1, power grid topology graph division; step 2, setting routing nodes; step 3, query. The application divides the power grid topology graph into several subgraphs with comparable scales, and specifies routing nodes in the subgraphs, wherein the routing nodes store the shortest paths to each node in the belonging subgraph, the nodes on the non-shortest paths are filtered by querying the routing nodes, the query is accelerated, and the shortest path query in the power GIS graph database can be effectively accelerated.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of computer technology, and particularly relates to a power graph database query acceleration method based on graph partitioning and node routing. BACKGROUND

[0002] With the continuous increase of power grid topology structure, the time complexity of querying the shortest path in the power GIS graph database increases sharply, which restricts the topology analysis performance of the power GIS graph database, and greatly increases the risk of power system paralysis.

[0003] With the continuous development of the power system, the power facilities are increasing, and the corresponding equipment data is also increasing sharply. In order to efficiently process the equipment data, the graph database is widely applied in the power system and shows good performance. In order to better perform topology analysis, the GIS system is also applied in the power system. The system combines database technology, computer technology and the scale, structure, properties and real geographical location information of the power grid, and strengthens the topology analysis capability of the power system, which is conducive to promoting the development of smart grid. However, when facing the topology structure of city-level or province-level power grid, the shortest path query, as the basis for the optimal power supply scheme and the best repair path of the distribution network in the topology analysis of the power GIS graph database, faces great pressure. Therefore, the power GIS graph database puts forward higher requirements on the performance of querying the shortest path, and enhances the processing capability of emergency events.

[0004] At present, the strategy of the shortest path in the field of electric power is realized based on algorithm and improved storage structure. At present, for the best repair path of distribution network, an improved differential algorithm and a mathematical model of GIS best repair path of distribution network based on improved differential algorithm are designed, which effectively improves the calculation speed of the shortest path. At present, based on the analysis of the existing Dijkstra algorithm, different weights are given to the edge values of the city traffic graph according to the actual situation of urban traffic, and the Dijkstra algorithm is improved by taking the rectangular restriction search range as the basis, so as to realize the search of the shortest repair path of the distribution network. At present, by adopting the idea of space for time, the storage form of the topological network is improved, wherein three arrays are used to store the data related to the path on the topological graph, the data related to the nodes and the weight of each edge, and the Dijkstra algorithm is used to calculate the best repair path in the shortest time. At present, in order to improve the query efficiency, the idea of space for time is used to propose a node blocking strategy, which filters most of the non-optimal path nodes by specifying the cache data in the blocking node, and quickly realizes the solution of the shortest path. As described above, the existing research is to realize the best repair path of the distribution network by improving the differential algorithm and improving the Dijkstra algorithm, but the performance of the improved algorithm cannot adapt to the increasingly large power system. Another part of the research adopts the strategy of space for time, but these researches do not consider the cost brought by the storage cost. In order to improve the limitations in the above research schemes, the present application proposes a shortest path query algorithm for electric power GIS graph database based on graph partitioning and routing nodes. SUMMARY

[0005] In view of the above technical deficiencies, the purpose of the present application is to provide an electric power graph database query acceleration method based on graph partitioning and node routing, which first divides the overall power grid topology structure, then specifies two nodes constituting the interactive edge of the subgraph in the topological graph as routing nodes, and stores the shortest path to other nodes in the belonging subgraph in the routing node, and finally converts the overall shortest path query into routing node query to optimize the shortest path query efficiency.

[0006] To solve the above technical problems, the present application adopts the following technical scheme:

[0007] The present application provides an electric power graph database query acceleration method based on graph partitioning and node routing, characterized in that the specific steps are as follows:

[0008] Step 1: Power grid topology graph partitioning:

[0009] 1) : add three attributes to all nodes in the topology graph: (Bid, Did, Z), set variable T, T represents the number of nodes in the topology graph that have completed graph partition; Bid represents the identifier of the subgraph to which the node belongs; Did represents whether the node has completed subgraph allocation, where assignment of 1 indicates completion and assignment of 0 indicates non-completion; and z represents the number of existing nodes in a subgraph;

[0010] 2) : create a queue Q, randomly select a node X with Did value of 0 in the topology graph, perform BFS algorithm traversal from node X, add nodes with Did value of 0 that are traversed to the queue Q, and stop traversal at a node with Did value of 1; if the number of nodes in the queue is less than 10 after the first layer of recursion of the BFS algorithm is completed, it indicates that all nodes in the queue have been partitioned to a subgraph, and then all nodes in the queue are dequeued, BFS algorithm is performed from the node and traversed to the surrounding subgraph, the dequeued node is allocated to the subgraph with the fewest nodes, and G++ and Z++ operations are performed; if the number of nodes in the queue reaches 10 after the first round of recursion, all nodes in the queue are dequeued, and the subgraph to which each node belongs is set, Did is assigned a value of 1, T is incremented, and Z is assigned a value of 10, and then the queue is emptied;

[0011] 3) : perform BFS algorithm traversal again from node X, and create a new queue Q; stop traversal when a node with Did value of 0 is reached, and repeat the process of step (2) from the node;

[0012] 4) : if the value of T is equal to the total number of topology nodes, it indicates that the partitioning phase of the power grid topology graph using the SP algorithm has been completed;

[0013] Step 2: Set the routing node:

[0014] 1) : perform BFS algorithm traversal from a node X, determine whether there is an adjacent node with different Bid value from X in the nodes traversed by the first layer of recursion of the BFS algorithm, if there is a node with different Bid value, then node X is the routing node of the subgraph, add attribute isRoute to the routing node, which indicates whether the node is a routing node, assignment of 1 indicates that the node is a routing node, and assignment of 0 indicates that the node is a non-routing node; then add a new attribute array A to the routing node, A = ([X5, [X, X2, X3, X4, X5],..., [X i , [X, X2, X5, X6, X i]]), the array A stores several arrays, each array with index 0 represents the target node of the shortest path stored in the array, and the index 1 represents the shortest path from the routing node X to other nodes in the subgraph; if there is no node with different Bid values, continue the BFS traversal and repeat the step (1);

[0015] 2) Starting from the routing node, the Dijkstra algorithm is executed to obtain the shortest path from the routing node to other nodes in the subgraph, and the shortest path is stored;

[0016] 3) Repeat the step (1) to find all routing nodes in the topology graph and complete the storage of the shortest path;

[0017] Step 3: Query:

[0018] 1) The source node and the destination node belong to the same subgraph: the BFS algorithm is executed from the source node and traversed to the routing node of the subgraph, and it is checked whether the target node exists in the storage data of the routing node, at the same time, the Dijkstra algorithm is executed from the initial position of the starting node and traversed to the target node;

[0019] 2) The source node and the destination node do not belong to the same subgraph: the BFS algorithm is executed from the source node and traversed to the routing node of the subgraph, and it is checked whether the destination node exists in the storage data of the routing node, and then the shortest path is obtained by executing the BFS algorithm between the subgraphs from all routing nodes in the subgraph where the source node is located to the routing node in the subgraph where the destination node is located.

[0020] The beneficial effects of the present application are that: by dividing the power grid topology graph into several subgraphs of comparable size and designating routing nodes in the subgraphs, the routing nodes store the shortest paths to each node in the subgraph, and by querying the routing nodes to filter the nodes on the non-shortest path, the query is accelerated; the query of the shortest path in the power GIS graph database can be effectively accelerated. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0022] Figure 1 The flowchart of the power graph database query acceleration method based on graph division and node routing provided by the present application;

[0023] Figure 2The schematic diagram of graph division provided by the present application;

[0024] Figure 3 The schematic diagram of routing node provided by the present application;

[0025] Figure 4 The time comparison graph of 1-layer recursion to 40-layer in experimental analysis provided by the present application;

[0026] Figure 5 The time comparison graph of 50-layer recursion to 120-layer in experimental analysis provided by the present application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all the other embodiments obtained by those skilled in the art without creative work are within the protection scope of the present application.

[0028] The power graph database query acceleration method based on graph division and node routing has the following specific steps:

[0029] Step 1: Power grid topology graph division: in order to enhance the topology analysis performance of the power GIS graph database and accelerate the shortest path query, the nodes representing transformers, generators and other facilities in the topology graph need to be divided according to the graph division principle, and the graph division principle is used to efficiently manage the power system. The specific implementation steps are as follows:

[0030] (1): add three attributes (Bid, Did, Z) to all nodes in the topology graph, set variable T, T represents the number of nodes in the topology graph that have completed graph division; Bid represents the identifier of the subgraph to which the node belongs; Did represents whether the node has completed subgraph allocation, where 1 indicates completion and 0 indicates non-completion; Z represents the number of existing nodes in a subgraph;

[0031] (2): Create queue Q. Randomly select a node X with a Did value of 0 in the topology graph. Perform BFS algorithm traversal from node X. Add the nodes with Did values ​​of 0 to queue Q. If a node with a Did value of 1 is encountered, stop traversing at that node. If the number of nodes in the queue is less than 10 after the first level of recursion of the BFS algorithm, it means that all the surrounding nodes of all nodes in the queue have been assigned to a certain subgraph. Then dequeue all nodes in the queue. For each node dequeued, start BFS algorithm from that node and traverse to the surrounding subgraphs. Assign the dequeued node to the subgraph with the fewest nodes and perform G++ and Z++ operations. If the number of nodes in the queue reaches 10 after the first round of recursion, dequeue all nodes in the queue. When dequeuing, set the subgraph representation Bid of each node, assign Did to 1, perform T++ operation, assign Z to 10, and clear the queue after completion.

[0032] (3): Start the BFS algorithm again from node X, create a new queue Q, stop the traversal when the node with the Did value of 0 is reached, and repeat the process of step (2) from that node.

[0033] (4): If the value of T is equal to the total number of topology nodes, this indicates that the grid topology partitioning phase of the SP algorithm has been completed;

[0034] Step 2: Set up routing nodes: In this section, routing nodes are first formed based on the characteristics of blocking nodes and routing tables. Secondly, any two nodes that form an interaction edge between subgraphs are routing nodes. Finally, routing nodes can store the shortest paths to other nodes within their respective subgraphs.

[0035] By adding routing nodes to the partitioned subgraphs, the performance of topology analysis in power GIS map databases can be further improved. Therefore, specifying routing nodes in the partitioned subgraphs needs to satisfy an important principle: given node u∈V i Node v∈V j If there exists an edge (u, v) ∈ E, i ≠ j, then node u is a subgraph G. i The routing node, node v is a subgraph G j The routing node. For example... Figure 3 The diagram shown is a schematic of a routing node. The specific implementation steps are as follows:

[0036] (1) : BFS algorithm traversal from a node X, determine whether there is a node with different Bid value adjacent to X in the first layer of BFS algorithm recursive traversal, if there is a node with different Bid value, the node X is the routing node of the subgraph, add attribute isRoute to the routing node, which indicates whether the node is a routing node, value 1 indicates that the node is a routing node, value 0 indicates that the node is a non-routing node; then add new attribute array A to the routing node = ([X5, [X, X2, X3, X4, X5]],..., [X i , [X, X2, X5, X6, X i ]), array A stores several arrays, the 0th position of each array indicates the target node of the shortest path stored in the array, and the 1st position indicates the shortest path from the routing node X to other nodes in the subgraph; if there is no node with different Bid value, continue BFS traversal and repeat step (1);

[0037] (2) : execute Dijkstra algorithm from the routing node to obtain the shortest path from the routing node to other nodes in the subgraph, and store the shortest path;

[0038] (3) : repeat step (1) to find all routing nodes in the topology graph and complete the storage of the shortest path;

[0039] Step 3: query: for the topology graph that has completed SP algorithm, there are two cases for shortest path query analysis, the first case is that the source node and the destination node belong to the same subgraph, and the second case is that the source node and the destination node do not belong to the same subgraph, this part will discuss the two cases; this part uses basic BFS algorithm and Dijkstra algorithm, and this paper focuses on the power GIS graph database query acceleration on the stages of power grid topology graph division and routing node setting.

[0040] (1) the source node and the destination node belong to the same subgraph: execute BFS algorithm from the source node and traverse to the routing node of the subgraph, view the storage data of the routing node to see if there is a target node, at the same time, start node executes Dijkstra algorithm at the initial position to traverse to the target node;

[0041] (2) the source node and the destination node do not belong to the same subgraph: execute BFS algorithm from the source node and traverse to the routing node of the subgraph, view the storage data of the routing node to see if there is a target node, then start from all routing nodes of the subgraph where the source node is located, traverse to the routing node of the subgraph where the destination node is located by executing BFS algorithm between subgraphs, and obtain the shortest path using the routing node.

[0042] The graph partitioning, blocking nodes, and routing tables are explained below:

[0043] like Figure 2 As shown, the graph is divided.

[0044] A graph is a data structure that represents many-to-many relationships, consisting of a finite, non-empty set of vertices and a set of edges between vertices. It is denoted as [Graph]. Let G represent a graph, where V is the set of vertices in graph G and E is the set of edges in graph G, where E = {(u, v) | (u, v ∈ V) ∧ (u ≠ v)}.

[0045] Given a positive integer k, partition a graph G into k disjoint subgraphs, such that G = {G1(V1, E1), ..., G...} k (V k E k )}, 1≤i≤k, this process is called graph partitioning. In graph partitioning, given a positive integer j, where 1≤j≤k, when node u∈V i ,v∈V j If there exists an edge (u, v) ∈ E and i ≠ j, then the edge (u, v) is the interaction edge between subgraphs.

[0046] The original problem is decomposed into multiple subproblems, and each subproblem is solved one by one. This problem-solving process is applicable to graph partitioning theory. Therefore, graph partitioning theory can be applied to the shortest path problem in power GIS graph databases. For power grid topology graph partitioning, two important principles must be met:

[0047] 1) Load balancing. The vertices in each subgraph after partitioning are roughly the same size, avoiding resource waste due to uneven task allocation. That is, for each set V... i , Where |V i | represents the number of nodes in the subgraph after load balancing is satisfied.

[0048] 2) Minimize the number of interaction edges. The number of interaction edges between subgraphs determines the communication overhead between them. To reduce overall communication overhead, the number of interaction edges should be minimized. For example... Figure 1 The diagram shown is a schematic diagram of the completed graph partitioning.

[0049] The graph partitioning, blocking nodes, and routing tables are explained below:

[0050] Graph partitioning:

[0051] A graph is a data structure that represents many-to-many relationships, consisting of a finite, non-empty set of vertices and a set of edges between vertices. It is denoted as [Graph]. Let G represent a graph, where V is the set of vertices in graph G and E is the set of edges in graph G, where E = {(u, v) | (u, v ∈ V) ∧ (u ≠ v)}.

[0052] Given a positive integer k, partition a graph G into k disjoint subgraphs, such that G = {G1(V1, E1), ..., G...} k (V k E k )}, 1≤i≤k, this process is called graph partitioning. In graph partitioning, given a positive integer j, where 1≤j≤k, when node u∈V i ,v∈V j If there exists an edge (u, v) ∈ E and i ≠ j, then the edge (u, v) is the interaction edge between subgraphs.

[0053] The original problem is decomposed into multiple subproblems, and each subproblem is solved one by one. This problem-solving process is applicable to graph partitioning theory. Therefore, graph partitioning theory can be applied to the shortest path problem in power GIS graph databases. For power grid topology graph partitioning, two important principles must be met:

[0054] 1) Load balancing. The vertices in each subgraph after partitioning are roughly the same size, avoiding resource waste due to uneven task allocation. That is, for each set V... i , Where |V i | represents the number of nodes in the subgraph after load balancing is satisfied.

[0055] 2) Minimize the number of interaction edges. The number of interaction edges between subgraphs determines the communication overhead between them. To reduce overall communication overhead, the number of interaction edges should be minimized. For example... Figure 2 The diagram shown is a schematic diagram of the completed graph partitioning.

[0056] Routing table and blocking nodes:

[0057] The proposed node blocking strategy can designate any node in the topology graph as a blocking node. The blocking node itself can cache the shortest path to a certain node, and the invalid nodes with non-optimal paths can be filtered by querying the data stored in the blocking node.

[0058] In computer networks, a routing table is a small database stored in a router that stores paths pointing to specific network addresses. The primary function of the routing table is to find the optimal transmission path for each piece of data passing through the router and efficiently deliver that data to its destination. Routing tables can be configured by an administrator, or they can be automatically adjusted by the router.

[0059] By combining the blocking node strategy and routing table, the topology analysis performance of the power GIS graph database can be further enhanced.

[0060] Experimental analysis:

[0061] Experimental configuration:

[0062] A simulation experiment was performed using real data provided by a power system graph database of a certain provincial State Grid, which has 1 million nodes and 6 million edges. The operating system and hardware configuration used in the simulation experiment are shown in Table 1.

[0063] Table 1 Experimental software and hardware environment

[0064] CPU Intel(R) Core(TM) i7-11700 CPU @ 2.50GHz System Memory 16 GB Operating System CentOS 7

[0065] Query time analysis:

[0066] First, in the power GIS graph database query acceleration stage, the number of subgraph nodes is about 10, and the routing node can store the shortest path to other nodes within the belonging subgraph. When the number of subgraph nodes increases, the data stored in the routing node will increase the query time. Second, the BFS algorithm is used for traversal between routing nodes in subgraphs, which will produce recursion phenomenon, and the increase of recursion layers will increase the query time. Therefore, in order to verify the effectiveness of the SP algorithm, the experiment will compare the case that the number of subgraph nodes is 10 and 20 respectively and the subgraph uses SP algorithm recursion to 120 layers. It is worth noting that the completed experiment shows that the performance of bidirectional BFS algorithm based on node blocking strategy (PBFS) is better than that of BFS algorithm, so the experiment does not compare the performance of BFS algorithm, but compares with PBFS algorithm. In addition, Dijkstra algorithm has low time complexity in the case of small number of nodes, so the time consumed by Dijkstra algorithm in subgraph to execute shortest path query can be ignored. The experimental results are shown in Table 2, Figure 4 , Figure 5 , wherein Table 2 represents the time consumed from the first recursion layer to the 120th layer, and Figure 4 , Figure 5 represents the comparison between time and layer number.

[0067] Table 2 Time comparison of iteration to 120 layers, time (s)

[0068]

[0069] From Table 2, Figure 4 and Figure 5It is found that the query time is kept within 1 S in the case of sub-graphs with 10 and 20 nodes respectively due to the application of graph partition and routing nodes, and the query time of sub-graphs with 20 nodes is always slightly greater than that of sub-graphs with 10 nodes, because the number of sub-graph nodes increases from 10 to 20, and the shortest path data stored in the routing nodes also increases, and the query of the stored data in the routing nodes consumes extra time. Unlike the PBFS algorithm, the time consumed by the algorithm in the recursive case of 1 to 40 layers and 120 layers is not much different from that of the SP algorithm, but in the recursive case of 40 layers to 110 layers, the algorithm takes more time, with the highest being 41.291 seconds and the lowest being 2.774 seconds. The above analysis shows that the SP algorithm of dividing the power grid topology graph into several sub-graphs of comparable size and querying the shortest path through the routing nodes can effectively realize the effect of query acceleration and show excellent performance.

[0070] The method of the present application divides the power grid topology graph into several sub-graphs of comparable size, and specifies routing nodes in the sub-graphs, wherein the routing nodes store the shortest paths to each node in the sub-graphs, and filters the nodes on the non-shortest path by querying the routing nodes to realize accelerated query. The above experiments can verify that the method of the present application can effectively accelerate the query of the shortest path in the power GIS graph database.

[0071] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and variations.

Claims

1. A method for accelerating power graph database queries based on graph partitioning and node routing, characterized in that, The specific steps are as follows: Step 1: Power grid topology diagram division: 1) Add three attributes to all nodes in the topology graph: (Bid, Did, Z). Set the variable T, where T represents the number of nodes in the topology graph that have completed graph partitioning; Bid represents the identifier of the subgraph to which the node belongs; Did represents whether the node has completed subgraph assignment, where a value of 1 indicates completion and a value of 0 indicates incomplete assignment; Z represents the number of nodes currently in a certain subgraph. 2): Create a queue Q. Randomly select a node X with a Did value of 0 in the topology graph. Perform a BFS algorithm traversal from node X, adding nodes with Did values ​​of 0 to queue Q. If a node with a Did value of 1 is encountered, stop traversing at that node. If the number of nodes in the queue is less than 10 after the first level of recursion of the BFS algorithm, it means that all the surrounding nodes of all nodes in the queue have been assigned to a subgraph. Then, dequeue all nodes from the queue. For each dequeueed node, start the BFS algorithm from that node and traverse to the surrounding subgraphs. Assign the dequeueed node to the subgraph with the fewest nodes and perform G++ and Z++ operations, where G represents a graph. If the number of nodes in the queue reaches 10 after the first round of recursion, dequeue all nodes from the queue. When dequeuing, set the subgraph representation Bid of each node, assign Did to 1, perform T++ operation, and assign Z to 10. After completion, clear the queue. 3): Start the BFS algorithm again from node X, create a new queue Q, stop the traversal when the node with the Did value of 0 is reached, and repeat the process of step (2) from that node. 4): If the value of T equals the total number of topology nodes, this indicates that the grid topology partitioning phase of the SP algorithm has been completed; Step 2: Configure the routing node: 1): Start the BFS algorithm traversal from a node X. Determine if there are any adjacent nodes with different Bid values ​​than X among the nodes traversed in the first level of the BFS algorithm. If there are nodes with different Bid values, then node X is the routing node of its subgraph. Add the attribute isRoute to the routing node. This attribute indicates whether the node is a routing node. A value of 1 indicates that the node is a routing node, and a value of 0 indicates that the node is not a routing node. Then add a new attribute array A = ([X5, [X, X2, X3, X4, X5]], ..., [Xi, [X, X2, X5, X6, Xi]]) to the routing node. Array A stores several arrays. The position with index 0 represents the target node of the shortest path stored in the array, and the position with index 1 represents the shortest path from the routing node X to other nodes in its subgraph. If there are no nodes with different Bid values, continue the BFS traversal and repeat step (1). 2): Starting from the routing node, execute Dijkstra's algorithm to find the shortest path from the routing node to other nodes in its subgraph, and store the shortest path; 3): Repeat step (1) to find all routing nodes in the topology graph and complete the storage of the shortest path; Step 3: Query: 1) The source node and the destination node belong to the same subgraph: Execute the BFS algorithm from the source node and traverse to the routing node of the subgraph. Check the stored data of the routing node to see if the target node exists. At the same time, execute Dijkstra's algorithm from the starting node to traverse to the target node. 2) The source node and the destination node do not belong to the same subgraph: The source node executes the BFS algorithm and traverses to the routing nodes of its own subgraph. It checks that the destination node is not present in the stored data of the routing nodes. Then, starting from all the routing nodes in the subgraph where the source node is located, it traverses to the routing nodes in the subgraph where the destination node is located by executing the BFS algorithm between subgraphs, and uses the routing node to obtain the shortest path.

Images