Distribution network area division method and terminal based on edge computing

Abstract

The present invention provides a method and terminal for dividing distribution network regions based on edge computing, comprising the following steps: obtaining preset indicators in the distribution network, constructing a target data model of the distribution network according to the preset indicators, performing nonlinear programming on the target data model to obtain the optimal number of clusters of the target data model; obtaining information on terminals to be allocated in the distribution network, generating a multi-perspective data set according to the information on terminals to be allocated, merging the multi-perspective data sets to obtain a consensus data set; performing multi-perspective clustering on the optimal number of clusters and the consensus data set to obtain a clustering result. Dividing the distribution network into regions based on the edge computing method to achieve reasonable division of regions not only alleviates the communication pressure of the main station, but also speeds up the response speed of fault removal. At the same time, the present invention performs regional division based on multiple preset indicators to comprehensively consider various factors of distribution network regional division and edge computing center deployment.

Description

Technical Field

[0001] The present invention relates to the field of distribution network area division, and in particular to a distribution network area division method and terminal based on edge computing. Background Art

[0002] The 10kV distribution network directly connects to end-user electricity users and is crucial for ensuring their electricity consumption experience. Its fundamental mission is to provide users with safe, stable, and high-quality electricity. However, compared to transmission networks, distribution networks have numerous nodes and branches, a complex structure, and are more prone to failures. With the integration of diverse distributed power sources into distribution networks and their expansion and upgrade, the topology of distribution networks is becoming increasingly complex. Regional protection, a method for efficient and rapid distribution network protection, is currently attracting considerable attention.

[0003] In the regional centralized architecture of a distribution network regional protection solution, a single distribution master station controls multiple distribution substations. A distribution substation serves as the regional control center, responsible for processing real-time information transmitted by smart terminals within the region when a fault occurs. It then transmits the fault processing results and disconnection instructions to the smart terminals in each distribution room, thereby eliminating the fault. The connection between distribution substations and terminals is very similar to the connection between edge computing centers and terminals in edge computing, with the distribution master station acting as a cloud computing center. Therefore, edge computing's advantages of decentralized computing power and cloud-edge collaboration are perfectly suited for distribution networks, ensuring rapid distribution network protection.

[0004] In the traditional distribution network fault handling mode, only the distribution network master station handles the fault. The computing task of the distribution network master station is too heavy, and it is difficult to ensure the fault handling speed. At the same time, the communication distance between each terminal and the distribution network master station is too long, the communication quality is difficult to ensure, the communication speed is slow, and the system delay is large. In the existing distribution network area division method, on the one hand, most studies only consider a single indicator for simple division when performing area division. For example, the existing technology [1] (Distribution network edge division method and application research based on stable connection [J]. Electric Power Information and Communication Technology, 2020, 18(01):26-32.DOI:10.16543 / j.2095-641x.electric.power.ict.2020.01.004) only divides the distribution network edge area according to the even number of feeders; on the other hand, in recent years, the continuous access of multiple distributed power sources to the distribution network has led to the change of the distribution network topology structure. The existing area division technology solution cannot be adjusted when the distribution network topology structure changes. For example, the existing technology [2] (edge ​​computing unit optimization configuration method for rapid processing of distribution network faults [J]. Electric Power Construction, 2022, 43(03):31-41) only considers multiple factors such as communication and topology under the current topology structure when dividing the distribution network area. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a distribution network area division method and terminal based on edge computing, so as to realize comprehensive and reasonable regional division of the distribution network, and to dynamically divide the areas according to changes in the distribution network topology structure, thereby improving the flexibility of the distribution network area division.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A method for dividing distribution network regions based on edge computing, comprising the steps of:

[0008] Obtaining preset indicators in the distribution network, constructing a target data model of the distribution network according to the preset indicators, performing nonlinear programming on the target data model, and obtaining an optimal number of clusters for the target data model;

[0009] Acquire information of terminals to be allocated in the distribution network, generate a multi-perspective data set based on the information of the terminals to be allocated, and merge the multi-perspective data sets to obtain a consensus data set;

[0010] Multi-perspective clustering is performed on the optimal number of clusters and the consensus data set to obtain a clustering result.

[0011] In order to solve the above technical problems, another technical solution adopted by the present invention is:

[0012] A distribution network area division terminal based on edge computing includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, each step of the distribution network area division method based on edge computing is implemented.

[0013] The beneficial effects of the present invention are as follows: in the regional centralized architecture of the distribution network, the division and selection of the distribution network substation and its coordinated area, that is, the edge center and edge area in edge computing are crucial. In the division of the protection area, in order to make the protection process fast and efficient, it is necessary to ensure fast communication and high-speed and accurate calculation, and the protection area cannot be too large; at the same time, considering the economic issue, the protection area cannot be too small. Therefore, in the selection of the regional edge computing center (i.e., the distribution substation), it is necessary to comprehensively consider the communication effect and some important nodes for selection. The present invention divides the distribution network into regions based on the edge computing method to achieve reasonable division of regions, which not only alleviates the communication pressure of the main station, but also speeds up the response speed of fault removal, and provides a guarantee for the safety and stability of the distribution network. At the same time, the present invention can divide the region based on a plurality of preset indicators, so as to comprehensively consider the factors of various distribution network regional divisions and edge computing center deployment. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A flowchart of a method for dividing distribution network areas based on edge computing provided by an embodiment of the present invention;

[0015] Figure 2 A flowchart of a method for dividing distribution network areas based on edge computing provided in the first embodiment of the present invention;

[0016] Figure 3 A specific flow chart of a method for dividing distribution network regions based on edge computing provided in Example 1 of the present invention;

[0017] Figure 4 A new terminal area division method based on edge computing provided in the second embodiment of the present invention;

[0018] Figure 5 A schematic diagram of the structure of a distribution network area division terminal based on edge computing provided by an embodiment of the present invention;

[0019] Description of labels:

[0020] 1. A distribution network area division terminal based on edge computing; 2. Memory; 3. Processor. DETAILED DESCRIPTION

[0021] To illustrate the technical content, achieved objectives and effects of the present invention in detail, the following description is given in conjunction with the embodiments and accompanying drawings.

[0022] Please refer to Figure 1 , an embodiment of the present invention provides a method for dividing distribution network regions based on edge computing, comprising the steps of:

[0023] Obtaining preset indicators in the distribution network, constructing a target data model of the distribution network according to the preset indicators, performing nonlinear programming on the target data model, and obtaining an optimal number of clusters for the target data model;

[0024] Acquire information of terminals to be allocated in the distribution network, generate a multi-perspective data set based on the information of the terminals to be allocated, and merge the multi-perspective data sets to obtain a consensus data set;

[0025] Multi-perspective clustering is performed on the optimal number of clusters and the consensus data set to obtain a clustering result.

[0026] From the above description, it can be seen that the beneficial effects of the present invention are: in the regional centralized architecture of the distribution network, the division and selection of the distribution network substation and its coordinated area, that is, the edge center and edge area in edge computing are crucial. In the division of protection areas, in order to make the protection process fast and efficient, it is necessary to ensure fast communication and high-speed and accurate calculation, and the protection area cannot be too large; at the same time, considering economic issues, the protection area cannot be too small. Therefore, in the selection of regional edge computing centers (i.e., distribution substations), it is necessary to comprehensively consider the communication effect and some important nodes for selection. The present invention divides the distribution network into regions based on the edge computing method to achieve reasonable division of regions, which not only alleviates the communication pressure of the main station, but also speeds up the response speed of fault removal, and provides a guarantee for the safety and stability of the distribution network. At the same time, the present invention can divide the region based on a plurality of preset indicators, and comprehensively consider the factors of various distribution network regional divisions and edge computing center deployment.

[0027] Furthermore, the preset indicators include communication indicators, economic indicators and efficiency indicators, the communication indicators include latency, bandwidth and packet loss rate, the economic indicators include the deployment cost of the edge computing server, and the efficiency indicators include the computing resource utilization rate of the edge computing server;

[0028] The step of obtaining preset indicators in the distribution network, constructing a target data model of the distribution network according to the preset indicators, and performing nonlinear programming on the target data model to obtain the optimal number of clusters of the target data model is specifically:

[0029] Construct a target data model of the distribution network based on the preset indicators:

[0030] minF(k)=ω1T′+ω2P′ loss +ω3C′;

[0031] Among them, ω1, ω2, and ω3 are the weight values ​​of delay, packet loss rate, and deployment cost respectively; T′ is the normalized communication delay; P loss ′ is the normalized packet loss rate; C′ is the normalized deployment cost;

[0032] Constrain the target data model:

[0033]

[0034] Among them, B0 is the bandwidth required for the service of the terminal to be allocated; N is the total number of terminals to be allocated; k is the number of divided areas; φ1 is the bandwidth redundancy coefficient; φ2 is the bandwidth redundancy coefficient; B given is the maximum bandwidth limit; R i is the quantified value of the average computing resources required by the i-th terminal to be allocated; R is the quantified value of the total computing resources of each computing center; φ3 is the computing power resource redundancy rate; T given is the maximum delay limit; is the maximum bit error rate limit; C givem is the maximum cost limit;

[0035] A nonlinear programming solution is performed on the target data model according to the constraints to obtain an optimal number of clusters of the target data model.

[0036] As can be seen from the above description, the preset indicators include communication indicators, which are used to avoid communication problems such as slow communication speed and large system delays caused by the long communication distance between each terminal and the distribution master station when the distribution network terminals are divided into regions, making it difficult to ensure communication quality; the preset indicators also include economic indicators, which are used to reduce the deployment cost of the system server; in addition, the preset indicators also include efficiency indicators, which are used to improve the response speed of the distribution network in handling faults and avoid the problem of excessive computing tasks. In this way, various factors of distribution network regional division and edge computing center deployment are comprehensively considered.

[0037] Furthermore, the terminal information to be allocated includes topological structure and location information; the multi-view data set includes a first data set and a second data set;

[0038] The acquiring of information of terminals to be allocated in the distribution network and the generating of a data set according to the information of terminals to be allocated are specifically as follows:

[0039] generating a first data set according to the topological structure of the terminals to be allocated;

[0040] generating a second data set according to the location information of the terminal to be assigned;

[0041] The first data set and the second data set are merged to obtain a consensus data set.

[0042] As can be seen from the above description, the first dataset is generated based on the terminal topology. Regionalization based on the first dataset increases the likelihood that electrically connected terminals will be assigned to the same area. The second dataset is generated based on the terminal's location information. Regionalization based on the second dataset increases the likelihood that terminals with similar geographical locations or belonging to the same administrative region will be assigned to the same area, reducing communication distances and facilitating management.

[0043] Furthermore, the topology structure includes the electrical connection relationship between the terminals to be allocated and the feeder fault frequency; the location information includes longitude, latitude and administrative area;

[0044] The generating of the first data set according to the topological structure of the terminal to be allocated is specifically as follows:

[0045] The weight value of the topology structure is calculated according to the feeder fault frequency:

[0046] w=ωp1+(1-ω)p2;

[0047] Wherein, ω is the weight index, p1 and p2 are the normalized results corresponding to the number of distributed power sources connected to the terminal to be allocated and the fault frequency respectively;

[0048] Generate an adjacency matrix of the topological structure according to the weight values:

[0049] W i =(W ij ,…,0,…,W ik )

[0050] Wherein, the adjacency matrix W is an n×n n-dimensional matrix, n is the total number of the terminals to be allocated; W i is the i-th row of the adjacency matrix W; W ij is the edge weight between the i-th terminal to be assigned and the j-th terminal to be assigned; W ik is the edge weight between the i-th terminal to be assigned and the k-th terminal to be assigned;

[0051] generating a Laplacian matrix of the topological structure according to the adjacency matrix, calculating eigenvalues ​​of the Laplacian matrix, selecting eigenvectors corresponding to the eigenvalues ​​that meet preset conditions, and generating a first data set after dimensionality reduction according to the eigenvectors;

[0052] The generating of the second data set according to the location information of the terminal to be allocated is specifically as follows:

[0053]

[0054] Among them, the second data set A p is an n×3 matrix, where n is the total number of terminals to be allocated; The second data set A p The i-th row of .

[0055] As can be seen from the above description, the first dataset is a weighted adjacency matrix. This weighting is used to assign weights to the topological structure of the terminal to be assigned, taking into account the number of distributed power sources connected to the terminal and the frequency of feeder failures. The second dataset contains the location information of each terminal to be assigned, specifically its specific geographic location and administrative division expressed in longitude and latitude. Organizing and summarizing these datasets into different categories facilitates the subsequent division of distribution network regions and data management.

[0056] Furthermore, the first data set and the second data set are merged to obtain a consensus data set:

[0057] A=(ω1A c ,ω2A p )

[0058] The consensus data set A is an n×6 matrix, where n is the total number of terminals to be allocated; A c For the first data set, A p is the second data set, ω1 is the first data set A c The weight value of ω2 is the second data set A p The weight value of .

[0059] From the above description, we can see that multi-perspective clustering includes two principles, namely the complementary principle and the consensus principle. The complementary principle stipulates that in order to describe data objects more comprehensively and accurately, multiple perspectives should be used, which is reflected in the equation as different perspectives correspond to different data sets; the consensus principle stipulates that the consistency of multiple different perspectives can be maintained to the greatest extent possible, which is reflected in the fact that different perspectives correspond to the same consensus data set.

[0060] Furthermore, the clustering result includes a partition result and a cluster center corresponding to each region;

[0061] The multi-perspective clustering is performed on the optimal number of clusters and the consensus data set, and the clustering results obtained are specifically:

[0062] The optimal number of clusters and the consensus data set are input into a preset k-means algorithm to obtain the partitioning results and the cluster center corresponding to each area.

[0063] From the above description, we can see that combining the consensus matrix with the k-means clustering algorithm optimizes the selection of the optimal number of clusters and cluster centers, making the clustering algorithm prediction results more accurate.

[0064] Furthermore, after performing multi-perspective clustering on the optimal number of clusters and the consensus data set to obtain the clustering results, the method further includes:

[0065] The geographical distance between each terminal to be assigned in the area and the cluster center corresponding to the area is calculated, and the terminal to be assigned with the smallest geographical distance is marked as the edge computing center.

[0066] From the above description, it can be seen that the terminal with the closest geographical distance in each area is selected as the edge computing center according to the cluster center. That is, the cluster center obtained by data calculation is sunk to the actual terminal and used as the edge computing center. This can ensure that all terminals in the area can meet the communication indicator requirements, facilitate the management and control of terminals in the area, and speed up the response speed of fault removal.

[0067] Furthermore, after performing multi-perspective clustering on the optimal number of clusters and the consensus data set to obtain the clustering results, the method further includes:

[0068] If there is a new terminal in the distribution network, the Euclidean distance between all edge computing centers that still have computing power surplus and the new terminal is calculated, the first edge computing center with the smallest Euclidean distance is selected, and it is verified whether the new terminal meets the preset indicators. If so, the new terminal is added to the area to which the first edge computing center belongs.

[0069] From the above description, it can be seen that after the design of the distribution network area division is completed, the distribution network system can still dynamically adjust the partitioning results according to the changes in the distribution network topology. For example, when a new terminal is added, it can automatically divide the area for it according to the preset indicators without manual modification, thereby improving the degree of automation of the area division system.

[0070] Furthermore, it also includes:

[0071] If the new terminal does not meet the preset indicators, after eliminating the first edge computing center, the second edge computing center with the smallest Euclidean distance is selected, and it is verified whether the new terminal meets the preset indicators until the new terminal area is successfully divided.

[0072] It can be seen from the above description that when dividing the new terminal area, the area division is performed according to the preset indicators, so the preset indicators provide a reference standard for the area division of the new terminal; in the case of a variable distribution network topology structure, the distribution network area division method of the present invention can still achieve accurate area division, ensuring the sustainability and flexibility of the distribution network area protection method.

[0073] Another embodiment of the present invention provides a distribution network area division terminal based on edge computing, including a memory, a processor, and a computer program stored on the memory and running on the processor. When the processor executes the computer program, it implements the various steps of the distribution network area division method based on edge computing.

[0074] As can be seen from the above description, the beneficial effects of the present invention are: based on the edge computing method, the distribution network is divided into regions, and the regions are rationally divided, which not only relieves the communication pressure of the master station, but also speeds up the response speed of fault removal, providing a guarantee for the safety and stability of the distribution network. At the same time, the present invention can perform regional division based on multiple preset indicators, and comprehensively consider various factors of distribution network regional division and edge computing center deployment.

[0075] The embodiments of the present invention provide a distribution network area division method and terminal based on edge computing, which can be applied to the distribution network system to solve the problems of low processing speed of traditional distribution network fault handling methods and high data processing pressure of distribution network master stations. The following is an illustration of the method through specific embodiments:

[0076] Please refer to Figures 1 to 3 , embodiment 1 of the present invention is:

[0077] A method for dividing distribution network regions based on edge computing, comprising the steps of:

[0078] S1. Obtain preset indicators in the distribution network, construct a target data model of the distribution network according to the preset indicators, perform nonlinear programming on the target data model, and obtain the optimal number of clusters of the target data model.

[0079] Specifically, the preset indicators include communication indicators, economic indicators, and efficiency indicators. The communication indicators include latency, bandwidth, and packet loss rate. The economic indicators include the deployment cost of the edge computing server. The efficiency indicators include the computing resource utilization rate of the edge computing server.

[0080] S11. Construct a target data model of the distribution network according to the preset indicators:

[0081] minF(k)=ω1T′+ω2P′ loss +ω3C′;

[0082] Among them, ω1, ω2, and ω3 are the weight values ​​of delay, packet loss rate, and deployment cost respectively; T′ is the normalized communication delay; P loss ′ is the normalized packet loss rate; C′ is the normalized deployment cost.

[0083] In this embodiment, the target data model is a normalized target data model, and the normalization process is:

[0084]

[0085]

[0086]

[0087] Where T is the communication delay, T min is the minimum value of communication delay, T max is the maximum value of communication delay; P is the packet loss rate, P min is the minimum value of packet loss rate, P max is the maximum packet loss rate; C is the deployment cost, C min is the minimum deployment cost, C max is the maximum deployment cost.

[0088] Specifically,

[0089]

[0090] Among them, v p is the data transmission rate on the link; k is the number of divided areas; N is the total number of terminals to be allocated; d ij is the distance between the i-th terminal and the j-th terminal; m is the number of routing hops from the terminal to be assigned to the computing center; ρ is the routing service efficiency; is the average packet length; a and b are the transition probabilities between two states of the Gilbert model, a is the probability of transitioning from no packet loss to packet loss, and b is the probability of transitioning from packet loss to no packet loss.

[0091] S12. Constrain the target data model:

[0092]

[0093] Among them, B0 is the bandwidth required for the service of the terminal to be allocated; N is the total number of terminals to be allocated; k is the number of divided areas; φ1 is the bandwidth redundancy coefficient; φ2 is the bandwidth redundancy coefficient; B given is the maximum bandwidth limit; R i is the quantified value of the average computing resources required by the i-th terminal to be allocated; R is the quantified value of the total computing resources of each computing center; φ3 is the computing power resource redundancy rate; T given is the maximum delay limit; is the maximum bit error rate limit; C given is the maximum cost limit.

[0094] S13. Perform nonlinear programming on the target data model according to the constraints to obtain the optimal number of clusters K of the target data model.

[0095] S2. Obtain information of terminals to be allocated in the distribution network, generate a multi-perspective dataset based on the information of terminals to be allocated, and merge the multi-perspective datasets to obtain a consensus dataset.

[0096] Specifically, the terminal information to be allocated includes topological structure and location information; the multi-view data set includes a first data set and a second data set;

[0097] In this embodiment, the topological structure of the terminals to be allocated includes the topological structure of terminals such as the switchgear terminal DTU (Distribution Terminal Unit), the distribution transformer terminal TTU (Distribution Transformer Supervisory Terminal Unit), the feeder terminal FTU (Feeder Terminal Unit), the distributed power source DG (Distributed Generator) and the tie switch; the failure frequency of the feeder refers to the failure frequency of the connecting line between any two connected terminals; the computing power of the terminals to be allocated includes parameters such as the average computing task data of each terminal; the location information of the terminals to be allocated includes the geographical location of each terminal and the administrative region to which it belongs.

[0098] In an embodiment, the multi-view dataset and the consensus dataset are both in matrix form.

[0099] S21. Generate a first data set according to the topological structure of the terminals to be allocated, specifically:

[0100] The topology structure includes the electrical connection relationship between the terminals to be allocated and the feeder fault frequency; the location information includes longitude, latitude and administrative area;

[0101] The weight value of the topology structure is calculated according to the feeder fault frequency:

[0102] w=ωp1+(1-ω)p2;

[0103] Wherein, ω is the weight index, p1 and p2 are the normalized results corresponding to the number of distributed power sources connected to the terminal to be allocated and the fault frequency respectively;

[0104] It should be noted that in the first dataset A cIn the data set, the weight value of the connection line between the terminal and the distributed power source is higher than the weight value of the connection line between ordinary terminals, and the weight value of the connection line with a high terminal failure frequency is even higher. Therefore, by normalizing the number of distributed power sources connected to the terminal and the failure frequency and calculating the weight value, the connection relationship and failure frequency of the terminal can be directly obtained from the first data set.

[0105] S212: Generate an adjacency matrix of the topological structure according to the weight values:

[0106] W i =(W ij ,…,0,…,W ik )

[0107] Wherein, the adjacency matrix W is an n×n n-dimensional matrix, n is the total number of the terminals to be allocated; W i is the i-th row of the adjacency matrix W; W ij is the edge weight between the i-th terminal to be assigned and the j-th terminal to be assigned; W ik is the edge weight value between the i-th terminal to be assigned and the k-th terminal to be assigned.

[0108] In this embodiment, assuming that there are n terminals to be assigned, the adjacency matrix W is an n×n n-dimensional matrix; the row vector W i That is, the data of the i-th terminal to be assigned, where it is connected to the j-th terminal to be assigned and the k-th terminal to be assigned, and the edge weight values ​​are W ij and W ik , and the data of the i-th terminal to be assigned that is not connected to the other terminals to be assigned is taken as an example, so the row vector Expressed as

[0109] W i =(0, ..., W ij ,…,0,…,W ik ,…,0);

[0110] That is, the row vector The elements in represent the edge weight values ​​connecting the i-th terminal to be assigned and other terminals to be assigned.

[0111] S213: Generate a Laplacian matrix of the topological structure according to the adjacency matrix, calculate the eigenvalues ​​of the Laplacian matrix, select eigenvectors corresponding to the eigenvalues ​​that meet preset conditions, and generate a first data set A after dimensionality reduction according to the eigenvectors. c ;

[0112] Specifically, the eigenvectors corresponding to the first three smallest non-zero eigenvalues ​​are selected, and these three eigenvectors constitute the first dataset A after dimensionality reduction. cAmong them, the first data set A c is an n×3 matrix, where n is the total number of terminals to be allocated.

[0113] S22: Generate a second data set A based on the location information of the terminal to be allocated. p , specifically:

[0114]

[0115] Among them, the second data set A p is an n×3 matrix, where n is the total number of terminals to be allocated; The second data set A p The i-th row of .

[0116] In this embodiment, it is assumed that there are n terminals to be assigned. p In the example, we first number the different administrative regions, and the row vector Represents the data of the i-th terminal.

[0117] S23. Merge the first dataset and the second dataset to obtain a consensus dataset, specifically:

[0118] A=(ω i A c ,ω2A p )

[0119] The consensus data set A is an n×6 matrix, where n is the total number of terminals to be allocated; A c For the first data set, A p is the second data set, ω1 is the first data set A c The weight value of ω2 is the second data set A p The weight value of .

[0120] S3. Perform multi-perspective clustering on the optimal number of clusters and the consensus data set to obtain a clustering result.

[0121] Specifically, the clustering result includes the partitioning result and the cluster center corresponding to each area;

[0122] S31 , inputting the optimal number of clusters and the consensus data set into a preset k-means algorithm to obtain a partitioning result and a cluster center corresponding to each region.

[0123] The S3 also includes:

[0124] S32: Calculate the geographical distance between each terminal to be assigned in the area and the cluster center corresponding to the area, and mark the terminal to be assigned with the smallest geographical distance as the edge computing center.

[0125] S4. If there is a new terminal in the distribution network, calculate the Euclidean distance between all edge computing centers that still have computing power surplus and the new terminal, select the first edge computing center with the smallest Euclidean distance, and verify whether the new terminal meets the preset indicators. If so, add the new terminal to the area to which the first edge computing center belongs.

[0126] If the new terminal does not meet the preset indicators, after eliminating the first edge computing center, the second edge computing center with the smallest Euclidean distance is selected, and it is verified whether the new terminal meets the preset indicators until the new terminal area is successfully divided.

[0127] It should be noted that the preset indicator is a communication indicator.

[0128] Please refer to Figure 4 , the second embodiment of the present invention is:

[0129] A specific implementation of S4 in a distribution network area division method based on edge computing.

[0130] In an optional implementation, the new terminal includes a distributed power source.

[0131] When a new distributed power source is added to the distribution network, the current partitioning result needs to be adjusted so that the new distributed power source is placed in one of the areas and meets the preset indicators. Specifically, after step S3, the following steps are further performed:

[0132] S41. If a newly added distributed power source exists in the distribution network, obtain the geographical locations, existing loads, and regions to which terminals connected to the newly added distributed power source belong of all edge computing centers;

[0133] The data format of the edge computing center is:

[0134] x i =(longitude, latitude, load number, region number, region number);

[0135] The data format of the newly added distributed power supply is:

[0136] y = (longitude, latitude, 0, area number of connected terminal 1, area number of connected terminal 2);

[0137] When a distributed power source is connected to the distribution network, there are two possible connection modes: the first is to connect directly to the edge terminal, and the second is to connect between two terminals. Therefore, a newly added terminal may be connected to one or two terminals. In order to keep the data length consistent, two data positions are reserved in the data format:i In , the area number is repeated twice; in y, if the newly added distributed power terminal is only connected to one terminal, the area number of the connected terminal is repeated twice.

[0138] S42. Select an edge computing center with surplus computing power from among all edge computing centers based on their geographical locations, existing loads, and the regions to which the terminals connected to the newly added distributed power sources belong.

[0139] S43. Calculate the Euclidean distance between all edge computing centers that still have computing power surplus and the newly added distributed power sources.

[0140] The Euclidean distance formula is specifically:

[0141]

[0142] Among them, i is the spatial dimension, x i is the edge computing center coordinate, y i The coordinates of the newly added distributed power supply.

[0143] This will obtain the Euclidean distance between the coordinates of the newly added distributed power source and each edge computing center that still has computing power margin.

[0144] S44. Select the first edge computing center with the smallest Euclidean distance and verify whether the newly added distributed power source meets the communication index. If so, add the newly added distributed power source to the area to which the first edge computing center belongs; if not, select the second edge computing center with the second smallest Euclidean distance and verify again whether it meets the communication index, until the newly added distributed power source is added to the edge computing center and can meet the communication index, it means that the area division of the newly added distributed power source is successful.

[0145] Please refer to Figure 5 , the third embodiment of the present invention is:

[0146] A distribution network area division terminal 1 based on edge computing includes a memory 2, a processor 3, and a computer program stored in the memory 2 and running on the processor 3. When the processor 3 executes the computer program, the various steps of the distribution network area division method based on edge computing described in Example 1 and Example 2 are implemented.

[0147] In summary, the present invention provides a method and terminal for regional division of a distribution network based on edge computing. In a centralized regional distribution network architecture, the division and selection of distribution substations and their coordinated regions, namely, edge centers and edge regions in edge computing, are crucial. Regarding the division of protection areas, to ensure rapid and efficient regional protection, it is necessary to ensure fast communication and high-speed, accurate computation. The protection area cannot be too large; at the same time, to consider economic efficiency, the protection area cannot be too small. Therefore, the present invention simultaneously constructs a multi-objective data model based on preset communication, economic, and efficiency indicators. Based on the combined effects of the multi-objective data model, it addresses the problem of most studies that simply consider a single metric when performing regional division. Furthermore, when clustering and partitioning terminals, a weighted terminal connection matrix is ​​generated based on the terminal topology and fault frequency. This ensures that terminals with electrical connections are more likely to be assigned to the same area. Simultaneously, a location affiliation information matrix is ​​generated based on the terminal's location information. This allows terminals with similar geographical locations or belonging to the same administrative region to be grouped into the same area, further reducing communication distances and facilitating management. Moreover, the present invention is based on the cluster centers and partitioning results generated based on the topological structure, and can realize automatic zoning in response to changes in the topological structure, thereby ensuring the sustainability and flexibility of the distribution network regional protection method. Therefore, in the selection of regional edge computing centers (i.e., distribution substations), the present invention comprehensively considers the communication effects and some important nodes for selection, which not only alleviates the communication pressure of the main station, but also speeds up the response speed of fault removal, thus providing a guarantee for the safety and stability of the distribution network. At the same time, regional division can be carried out based on a number of preset indicators, realizing a comprehensive and comprehensive consideration of factors such as communication, cost, topology, and administration in the regional division of various distribution networks and the deployment of edge computing centers.

[0148] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent transformations made using the contents of the present invention's description and drawings, or directly or indirectly applied in related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for dividing distribution network regions based on edge computing, characterized in that: Including steps: Obtaining preset indicators in the distribution network, constructing a target data model of the distribution network according to the preset indicators, performing nonlinear programming on the target data model, and obtaining an optimal number of clusters for the target data model; Acquire information of terminals to be allocated in the distribution network, generate a multi-perspective data set based on the information of the terminals to be allocated, and merge the multi-perspective data sets to obtain a consensus data set; Performing multi-perspective clustering on the optimal number of clusters and the consensus data set to obtain a clustering result; The terminal information to be allocated includes topological structure and location information; the multi-view data set includes a first data set and a second data set; The acquiring of information of terminals to be allocated in the distribution network and the generating of a data set according to the information of terminals to be allocated are specifically as follows: generating a first data set according to the topological structure of the terminals to be allocated; generating a second data set according to the location information of the terminal to be assigned; Merging the first data set and the second data set to obtain a consensus data set; The topology structure includes the electrical connection relationship between the terminals to be allocated and the feeder fault frequency; the location information includes longitude, latitude and administrative area; The generating of the first data set according to the topological structure of the terminal to be allocated is specifically as follows: The weight value of the topology structure is calculated according to the feeder fault frequency: ； in, is the weight index, and are respectively the normalized results corresponding to the number of distributed power sources connected to the terminal to be allocated and the fault frequency; Generate an adjacency matrix of the topological structure according to the weight values: The adjacency matrix W is an n×n dimensional matrix, where n is the total number of terminals to be allocated; is the i-th row of the adjacency matrix W; is the edge weight between the i-th terminal to be assigned and the j-th terminal to be assigned; W ik is the edge weight between the i-th terminal to be assigned and the k-th terminal to be assigned; generating a Laplacian matrix of the topological structure according to the adjacency matrix, calculating eigenvalues ​​of the Laplacian matrix, selecting eigenvectors corresponding to the eigenvalues ​​that meet preset conditions, and generating a first data set after dimensionality reduction according to the eigenvectors; The generating of the second data set according to the location information of the terminal to be allocated is specifically as follows: =(longitude, latitude, administrative region number); Among them, the second data set is an n×3 matrix, where n is the total number of terminals to be allocated; For the second data set A p The i-th row of .

2. A method for dividing distribution network regions based on edge computing according to claim 1, characterized in that: The preset indicators include communication indicators, economic indicators and efficiency indicators. The communication indicators include latency, bandwidth and packet loss rate. The economic indicators include the deployment cost of the edge computing server. The efficiency indicators include the computing resource utilization rate of the edge computing server. The step of obtaining preset indicators in the distribution network, constructing a target data model of the distribution network according to the preset indicators, and performing nonlinear programming on the target data model to obtain the optimal number of clusters of the target data model is specifically: Construct a target data model of the distribution network based on the preset indicators: ； Among them, ω1, ω2, and ω3 are the weight values ​​of latency, packet loss rate, and deployment cost, respectively; is the normalized communication delay; is the normalized packet loss rate; is the normalized deployment cost; Constrain the target data model: in, The bandwidth required for the service of the terminal to be allocated; is the total number of terminals to be allocated; is the number of divided regions; is the bandwidth redundancy coefficient; is the bandwidth redundancy coefficient; is the maximum bandwidth limit; is the quantified value of the average computing resources required by the i-th terminal to be assigned; A quantitative value of the total computing resources of each computing center; is the computing power resource redundancy rate; is the maximum delay limit; is the maximum bit error rate limit; is the maximum cost limit; A nonlinear programming solution is performed on the target data model according to the constraints to obtain an optimal number of clusters of the target data model.

3. A method for dividing distribution network regions based on edge computing according to claim 1, characterized in that: The first data set and the second data set are merged to obtain a consensus data set: The consensus data set A is an n×6 matrix, where n is the total number of terminals to be allocated; A c For the first data set, A p is the second data set, ω1 is the first data set A c The weight value of ω2 is the second data set The weight value of .

4. The method for dividing distribution network regions based on edge computing according to claim 1, characterized in that: The clustering result includes the partitioning result and the cluster center corresponding to each area; The multi-perspective clustering is performed on the optimal number of clusters and the consensus data set, and the clustering results obtained are specifically: The optimal number of clusters and the consensus data set are input into a preset k-means algorithm to obtain the partitioning results and the cluster center corresponding to each area.

5. A method for dividing distribution network regions based on edge computing according to claim 4, characterized in that: After performing multi-perspective clustering on the optimal number of clusters and the consensus data set to obtain a clustering result, the method further includes: The geographical distance between each terminal to be assigned in the area and the cluster center corresponding to the area is calculated, and the terminal to be assigned with the smallest geographical distance is marked as the edge computing center.

6. A method for dividing distribution network regions based on edge computing according to claim 5, characterized in that: After performing multi-perspective clustering on the optimal number of clusters and the consensus data set to obtain a clustering result, the method further includes: If there is a new terminal in the distribution network, the Euclidean distance between all edge computing centers that still have computing power surplus and the new terminal is calculated, the first edge computing center with the smallest Euclidean distance is selected, and it is verified whether the new terminal meets the preset indicators. If so, the new terminal is added to the area to which the first edge computing center belongs.

7. A method for dividing distribution network regions based on edge computing according to claim 6, characterized in that: Also includes: If the new terminal does not meet the preset indicators, after eliminating the first edge computing center, the second edge computing center with the smallest Euclidean distance is selected, and it is verified whether the new terminal meets the preset indicators until the new terminal area is successfully divided.

8. A distribution network area division terminal based on edge computing, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the processor implements the various steps of the method for dividing distribution network areas based on edge computing as described in any one of claims 1 to 7.

Images