Ad-hoc network method and system for power distribution Internet of Things

By building a distribution-weighted directed graph model and improving ant colony algorithm, dynamically adjusting the network topology, the problem of insufficient flexibility and adaptability in the distribution IoT self-organizing network is solved, load balancing and resource optimization are achieved, and the network topology is improved. The adaptability and flexibility of the network.

CN119052811AInactive Publication Date: 2024-11-29GUANGDONG UNIV OF SCI & TECH

Patent Information

Application Number
CN202411083521.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-08
Publication Date
2024-11-29
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing power distribution IoT self-networking technology lacks flexibility and adaptability, the network topology structure is fixed, difficult to dynamically adjust, unbalanced load distribution, low resource utilization, lack of intelligent optimization mechanism, and cannot meet the needs of complex application scenarios. .

Method used

By obtaining the physical and data state information of the node, a distribution-weighted directed graph model is built, and the optimization target algorithm and the improved ant colony algorithm are used to dynamically adjust the network topology to achieve load balancing and multi-objective optimization.

Benefits of technology

It improves the adaptability and flexibility of the network, realizes load balancing, improves resource utilization, fully reflects the concept of intelligence and adaptability, and can independently optimize and reconstruct the network topology according to real-time state.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119052811A_ABST
    Figure CN119052811A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of ad-hoc networks, and discloses a power distribution Internet of Things ad-hoc network method and system. Comprising the following steps: acquiring node physical state information and node data state information of n nodes in the power distribution network topology; obtaining an adjacent matrix among the n nodes according to the physical state information of the nodes; constructing a power distribution weighted directed graph model based on the node data state information and the adjacent matrix; based on the power distribution weighted directed graph model, calculating the shortest weighted path between the nodes and the path overhead corresponding to the shortest weighted path by adopting an optimization target algorithm; based on the shortest weighted path between the nodes and the path overhead corresponding to the shortest weighted path, the optimal path scheme is dynamically obtained by adopting the improved ant colony algorithm, the power distribution network topology is adjusted by utilizing the optimal path scheme, and the adaptability, the flexibility and the resource utilization rate of the network are greatly improved by dynamically optimizing the network topology and path planning.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of self-organizing networks, and more specifically, to a self-organizing network method and system for a power distribution Internet of Things. Background Art

[0002] The patent application with the publication number CN112261087A discloses a self-organizing networking method and system for the power distribution Internet of Things. The self-organizing networking method based on the sensor takes the gateway as the root node, and takes the receipt of a new networking request or the addition of a new network node as the networking condition, and broadcasts the networking through the networked nodes; topology recognition relies on the self-organizing network as the basis, and establishes an adjacency matrix to realize topology recognition. It has the advantages of wide interconnection, reliable communication, wide adaptability and strong anti-interference.

[0003] However, the existing distribution network still lacks flexibility and adaptability, the network topology is relatively fixed, and it is difficult to dynamically adjust according to the actual load and environmental conditions. The adaptability is limited. Faced with increasingly complex application scenarios, this defect will become more and more prominent; secondly, there are deficiencies in network modeling. There is a deviation between the model and the actual network situation, which affects the accuracy and practicality of the model; in addition, the problem of unbalanced load distribution is common. Some nodes are overloaded, while other nodes have idle resources, resulting in low resource utilization and affecting network performance; in addition, the path planning algorithm is relatively simple, often only focusing on single-objective optimization, lacking comprehensive consideration of multiple objectives, and it is difficult to obtain a comprehensive and reasonable optimal solution; finally, there is still a lack of intelligent optimization mechanism. The network cannot optimize and make decisions autonomously according to the real-time status, and requires manual intervention. The degree of intelligence is low, and it is difficult to meet future development needs.

[0004] In view of this, the present invention proposes a self-organizing network method and system for power distribution Internet of Things to solve the above problems. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solutions: 1. A self-organizing network method for power distribution Internet of Things, characterized by comprising: S1, obtaining node physical state information and node data state information of n nodes in the power distribution network topology;

[0006] S2. Obtain the adjacency matrix between n nodes according to the node physical state information;

[0007] S3, constructing a power distribution weighted directed graph model based on node data status information and adjacency matrix;

[0008] S4. Based on the power distribution weighted directed graph model, an optimization target algorithm is used to calculate the shortest weighted path between nodes and the path cost corresponding to the shortest weighted path;

[0009] S5. Based on the shortest weighted path between nodes and the path cost corresponding to the shortest weighted path, an improved ant colony algorithm is used to dynamically obtain the optimal path solution, and the distribution network topology is adjusted using the optimal path solution.

[0010] Further, the node physical status information includes node location coordinates, node communication distance and node communication bandwidth;

[0011] The node data status information includes the node's data flow, data cache capacity, data transmission delay, data packet loss rate, node energy level and data transmission volume.

[0012] Furthermore, the adjacency matrix is ​​obtained by:

[0013] Initialize an n×n sparse matrix A and construct a spatial index data structure of the node, traverse all nodes i, and calculate the communication capability of node i based on the node physical state information. Among them, Pt is the transmission power, Gt is the transmission antenna gain, Gr is the receiving antenna gain, γ is the wavelength of the wireless signal, Pr is the receiving sensitivity, L is the system loss factor, f(E_i,ρ_i) is the influence function; E_i is the node energy level of node i, g(ρ_i) is the environment function; ρ_i is the characteristic parameter of the environment where node i is located;

[0014] Influence function f(E_i,ρ_i)=E_i×exp(-ρ_i);

[0015] Environment Functions Among them, α, β and is an adjustable parameter;

[0016] Characteristic parameters Among them, F_iJ is the distance from node i to the Jth obstacle, r_i is the perception radius of node i; P_I is the power intensity of the Ith interference source, D_iI is the distance from node i to the Ith interference source; w_1 and w_2 are weight coefficients;

[0017] Based on the communication capability ra_i of node i, its communication coverage area R_i is determined; R_i is a circular area with the node position coordinates of node i as the center and the communication capability ra_i as the radius;

[0018] Use the spatial index data structure to find the neighbor node set N_i in the region R_i; for each node in the neighbor node set N_i of node i, record it as a neighbor node s;

[0019] Based on the node position coordinates between node i and neighbor node s, calculate the geographic distance dist(i,s) between node i and neighbor node s;

[0020] If dist(i, s) ≤ min(ra_i, ra_s), there is a physical connection relationship between node i and neighbor node s, where ra_s is the communication capability of neighbor node s; if there is a physical connection relationship, assign 1 to A[i, s], otherwise keep it as 0. A[i, s] is an element in the sparse matrix A. Repeat until all nodes are traversed; the finally obtained sparse matrix A is the adjacency matrix.

[0021] Furthermore, the construction method of the spatial index data structure includes:

[0022] Define the boundary region Y of the spatial index data structure; create a root node of the empty spatial index data structure, and the boundary region of the root node is the boundary region Y;

[0023] Initially define a small grid side length. According to the area S_j of the boundary region of leaf node j and the number of nodes N_j, dynamically adjust the small grid side length to obtain the ideal small grid side length;

[0024] The way of dynamically adjusting the small grid side length includes:

[0025] Set the minimum value l_min and the maximum value l_max of the small grid side length; based on the number of nodes N_j in the boundary region of leaf node j, calculate the expected number of small grids M_j in the boundary region of leaf node j;

[0026] Among them, is the ceiling function;

[0027] Then, based on the expected number of small grids M_j, the number of nodes N_j, and the area S_j, calculate the expected small grid side length in the boundary region of leaf node j

[0028] If l_j > l_max, set the ideal small grid side length to l_max; if l_j < l_min, set the ideal small grid side length to l_min; otherwise, set the ideal small grid side length to l_j;

[0029] Traverse all nodes. For each node i, start from the root node of the spatial index data structure. If the current node is a leaf node j, based on the obtained ideal small grid side length, divide its boundary region into small grids, traverse all nodes in the current boundary region, and according to the coordinate node position coordinates of each node, assign it to the corresponding small grid, and calculate the node density by dividing the number of nodes in the small grid by the area; preset a density threshold interval [m_min, m_max], where m_min is the lower limit of the node density and m_max is the upper limit of the node density;

[0030] Traverse all small grids, and for small grid k with node density greater than or equal to m_max, divide small grid k into 4 smaller sub-grids, and redistribute the nodes in the original small grid to the 4 sub-grids; for small grids with node density less than m_min, check its adjacent small grids, find adjacent small grids with node density less than m_min, and use these small grids as grids to be merged; merge the grids to be merged into a large grid, delete the original small grids, until the node density of all small grids is within the density threshold interval [m_min, m_max]; perform shape optimization on all large grids to obtain the final spatial index data structure.

[0031] Furthermore, the method of performing shape optimization on all large grids includes:

[0032] Collect the node position coordinates of all nodes in the grid to be merged to form a point set P;

[0033] Calculate the convex hull PU(P) of the point set P. On the convex hull PU(P), for any pair of parallel edges X1=[(ux,uy),(vx,vy)] and X2=[(u'.x,u'.y),(v'.x,v'.y)], where ux is the abscissa of the starting point of the parallel edge X1, uy is the ordinate of the starting point of the parallel edge X1, vx is the abscissa of the end point of the parallel edge X1, and vy is the ordinate of the end point of the parallel edge X1; u'.x is the abscissa of the starting point of the parallel edge X2, u'.y is the ordinate of the starting point of the parallel edge X2, v'.x is the abscissa of the end point of the parallel edge X2, and v'.y is the ordinate of the end point of the parallel edge X2; calculate the distance d between the parallel edges X1 and X2. X1,X2 ,

[0034]

[0035] Traverse all pairs of parallel edges on the convex hull PU(P), calculate the distance between them, and find the pair of parallel edges corresponding to the minimum distance d_min; use this pair of parallel edges as opposite sides to construct a minimum perimeter rectangle, and the other two sides of the minimum perimeter rectangle are perpendicular to the pair of parallel edges; use the minimum perimeter rectangle as the boundary area of ​​the merged large grid, and complete the shape optimization of all large grids.

[0036] Furthermore, the method of constructing the power distribution weighted directed graph model includes:

[0037] Initialize an n×n weighted adjacency matrix W, with all elements initially set to 0. Traverse all pairs of nodes i and f. If the corresponding element A[i,f] of node i and node f in the adjacency matrix is ​​1, calculate the edge weight w_if from node i to node f and assign it to W[i,f]. W[i,f] is an element in the weighted adjacency matrix W.

[0038] Among them, da_if represents the data flow from node i to node f, m_da is the maximum data flow in the entire network, del_if represents the data transmission delay from node i to node f, m_del represents the maximum data transmission delay in the entire network, dr_if represents the data packet loss rate from node i to node f; en_i is the node energy level of node i, en_f is the node energy level of node f; α1, β1, γ1 and δ1 are all balancing weight parameters;

[0039] Repeat until all nodes are traversed to obtain a complete weighted adjacency matrix W; based on the weighted adjacency matrix W, construct a preliminary power distribution weighted directed graph model G = (V, E, W1); where V is a set of nodes, E is a set of edges, representing the physical connection relationship between nodes, and W1 is a set of edge weights, representing the data transmission status between nodes; the preliminary power distribution weighted directed graph model is hierarchically linked to obtain a power distribution weighted directed graph model.

[0040] Furthermore, the method of hierarchically linking the preliminary power distribution weighted directed graph model includes:

[0041] Define the hierarchy of the preliminary power distribution weighted directed graph model, which includes the number of levels M3 and the node type; traverse all nodes, assign corresponding levels to them according to the type of the node, and number the levels to obtain the level number;

[0042] According to the level of the nodes, the nodes belonging to the same level are grouped. For each level q, an empty subgraph G_q = (V_q, E_q) is initialized, and all the nodes of this level are added to V_q. All pairs of nodes in this level are traversed. If there is a physical connection relationship, an edge is added from the subgraph G_q to E_q. V_q is the set of nodes in the subgraph G_q, and E_q is the edge set between the nodes in the subgraph G_q.

[0043] Traverse all edges z in E. If the edges z belong to different levels, add a special edge between the corresponding subgraphs;

[0044] For nodes in the same level, the edge weights are calculated in the same way as when constructing the weighted adjacency matrix W;

[0045] For the special edge (c, b) connecting different levels, define its weight w_cb, where c and b are the two nodes connected by the special edge;

[0046] Where, L_c is the level number of the level to which node c belongs, L_b is the level number of the level to which node b belongs, dS_cb is the physical distance between node c and node b, BW_cb is the bandwidth of the communication link between node c and node b, lay_cb is the communication delay between node c and node b, θ1, μ1 and ε1 are special edge weight parameters; M3 subgraphs and the special edges connecting them constitute the power distribution weighted directed graph model.

[0047] Furthermore, the method of calculating the shortest weighted path between nodes and the path cost corresponding to the shortest weighted path includes:

[0048] Define any node in the power distribution weighted directed graph model as the source node s', and define nodes other than the source node s' as the target node; for the source node s', initialize its distance estimation value di(s') and path cost estimation value co(s') to 0; for other nodes v' except the source node s', initialize v''s distance estimation value di(v') and path cost estimation value co(v') to infinity;

[0049] If there is a direct edge L' between v' and s', update di(v') and co(v') to the weight of edge L';

[0050] Create a priority queue pq, add all nodes v' and their current distance estimates di(v') and path cost estimates co(v') to the priority queue pq;

[0051] Define the comprehensive scoring function, the expression of the comprehensive scoring function is:

[0052] sc(v')=α3×di(v')+β3×co(v')+γ3×pe(v'); where sc(v') is the comprehensive score of node v', pe(v') is the penalty function value of node v', and α3, β3 and γ3 are score weight parameters;

[0053] pe(v') = ∑(w_O × FG_O(v')); where w_O is the weight of the Oth penalty factor, and FG_O(v') is the factor function corresponding to the Oth penalty factor;

[0054] Among them, RL(v') is the data cache capacity of node v', SY(v') is the residual energy of node v', and U1 is the factor balance parameter;

[0055] In the priority queue pq, sort the nodes from small to large according to the comprehensive score, and take out the node p' with the smallest comprehensive score; for each priority neighbor node q' of p', calculate the estimated distance from the source node to q' through p' and the estimated path cost from the source node to q' through p';

[0056] And calculate the comprehensive score LP_q'p' from the source node to q' through p'. If LP_q'p' is less than the current comprehensive score of q', update the distance estimate, path cost estimate and comprehensive score of q', and add q' to the priority queue; repeat until all nodes are processed or the target node is found;

[0057] For each node, initialize an empty list PUO and variable tot; start from q' and trace back, add each node to PUO in turn until reaching the source node, and calculate the sum of the weights of the edges between adjacent nodes in PUO as tot; output the shortest weighted path and the corresponding path cost based on PUO and tot.

[0058] Furthermore, the method for obtaining the optimal path solution includes:

[0059] Initialize m4 ant individuals to form an ant colony, each ant individual represents a possible path solution;

[0060] The solution space is constructed using the shortest weighted path between nodes and the corresponding path cost; the solution space contains all feasible path combination solutions;

[0061] Each ant individual gradually builds a complete path in the solution space according to the transition probability; the calculation formula of the transition probability is:

[0062] Where P(k')_q1,q2 represents the probability that the k'th ant individual transfers from node q1 to node q2, τ'_q1,q2 represents the pheromone concentration on the edge from node q1 to node q2, δ'_q1,q2 is the path cost from node q1 to node q2, α4 and β4 are the pheromone importance factor and heuristic importance factor respectively;

[0063] When all ant individuals complete the construction of the path, the fitness function is defined and the value of the fitness function of each path is calculated, which is recorded as fitness; the fitness function is a weighted combination of the path length, path cost, and load balance;

[0064] Load Balancing Among them, x_q1 is the load on node q1, and xtotal is the total load of all nodes;

[0065] According to the fitness of each path, pheromone is released on the corresponding path to obtain the optimal solution of the current iteration;

[0066] The formula for releasing pheromones is:

[0067] τ'_q1,q2=(1-ρ2)×τ'_q1,q2+Δτ'_q1,q2; where ρ2 is the pheromone volatility coefficient, ranging from (0, 1); Δτ'_q1,q2 is the sum of the newly released pheromones of all ant individuals on the edge from node q1 to node q2;

[0068] The entire ant colony is updated according to the optimal solution of the current iteration, a fitness threshold is preset, ant individuals whose fitness function value is less than the fitness threshold are deleted, and new ant individuals are generated to join the ant colony; this is repeated until the preset maximum number of iterations is reached, and the ant individual finally output is the current optimal path solution.

[0069] A power distribution Internet of Things self-organizing network system, which is used to implement a power distribution Internet of Things self-organizing network method, comprises: a data acquisition module, which is used to obtain node physical state information and node data state information of n nodes in the power distribution network topology;

[0070] The matrix construction module is used to obtain the adjacency matrix between n nodes based on the node physical state information;

[0071] The model hierarchical fitting module constructs a power distribution weighted directed graph model based on node data status information and adjacency matrix;

[0072] The path preliminary calculation module, based on the power distribution weighted directed graph model, uses the optimization target algorithm to calculate the shortest weighted path between nodes and the path cost corresponding to the shortest weighted path;

[0073] The solution output module uses an improved ant colony algorithm to dynamically obtain the optimal path solution based on the shortest weighted path between each node and the path cost corresponding to the shortest weighted path, and uses the optimal path solution to adjust the distribution network topology. Each module is connected by wired and / or wireless means.

[0074] The technical effects and advantages of the self-organizing method and system of the power distribution Internet of Things of the present invention are as follows:

[0075] The present invention can realize the intelligent self-organization of the distribution network. By constructing a weighted directed graph model for distribution, the concept of hierarchy is introduced to truly reflect the actual situation of the network and the difference in node importance, distinguish the importance of different types of nodes, and provide a basis for adopting differentiated optimization strategies at different levels. By using an improved ant colony algorithm, combined with the shortest path and path cost information between nodes, the optimal path solution is searched globally, and the network topology is dynamically adjusted, so that the distribution network can be autonomously optimized and reconstructed according to the real-time status, thereby improving the adaptability and flexibility of the network and effectively coping with complex and changeable loads and environments; secondly, load balancing in the distribution network is realized. During planning, the situation where some nodes are overloaded while other nodes are idle is avoided, and resource utilization efficiency is improved. A multi-objective optimization strategy is adopted, taking into account multiple indicators such as path length, cost and load balancing, so that the acquired path solution is more comprehensive and reasonable, fully reflecting the concepts of intelligence and adaptability, and greatly improving the adaptability, flexibility and resource utilization of the network by dynamically optimizing the network topology and path planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 A schematic diagram of a self-organizing network method for power distribution Internet of Things of the present invention;

[0077] Figure 2 This is a schematic diagram of a power distribution Internet of Things self-organizing network system of the present invention. DETAILED DESCRIPTION

[0078] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0079] Example 1

[0080] See also Figure 1 As shown, the power distribution Internet of Things self-organizing network method described in this embodiment includes:

[0081] S1. Obtain node physical status information and node data status information of n nodes in the power distribution network topology;

[0082] S2. Obtain the adjacency matrix between n nodes according to the node physical state information;

[0083] S3, constructing a power distribution weighted directed graph model based on node data status information and adjacency matrix;

[0084] S4. Based on the power distribution weighted directed graph model, an optimization target algorithm is used to calculate the shortest weighted path between nodes and the path cost corresponding to the shortest weighted path;

[0085] S5. Based on the shortest weighted path between nodes and the path cost corresponding to the shortest weighted path, an improved ant colony algorithm is used to dynamically obtain the optimal path solution, and the distribution network topology is adjusted using the optimal path solution to achieve network self-organization and load balancing.

[0086] Furthermore, the node physical status information includes node location coordinates, node communication distance and node communication bandwidth;

[0087] The node data status information includes the node's data flow, data cache capacity, data transmission delay, data packet loss rate, node energy level and data transmission volume.

[0088] It should be noted that the node location coordinates are obtained through GPS positioning or manual input, reflecting the geographical location of the node in the network; the node communication distance is measured or estimated based on the communication equipment parameters, reflecting the communication coverage of the node; the node communication bandwidth is obtained from the communication equipment specification manual, reflecting the data transmission capability of the node; the data accurately describes the spatial distribution and communication capability of the nodes in the network, laying the foundation for building an accurate network topology model.

[0089] Node data traffic is obtained through network monitoring or traffic statistics, reflecting the real-time data transmission needs of the node; data cache capacity is obtained from the node hardware configuration, reflecting the data cache capability of the node; data transmission delay is measured or estimated based on link status, reflecting the real-time performance of data transmission; data packet loss rate is obtained through data packet statistics, reflecting the reliability of data transmission; node energy level is estimated through node self-detection or energy model, reflecting the remaining power of the node; data transmission volume is obtained through traffic statistics, reflecting the data throughput demand of the node.

[0090] Furthermore, the adjacency matrix is ​​obtained by:

[0091] Initialize an n×n sparse matrix A and build a spatial index data structure of nodes to quickly find nodes in a given area.

[0092] Specifically, a boundary area Y of the spatial index data structure is defined. The boundary area is a rectangular area that includes the node position coordinates of all nodes. The node position coordinates of all nodes are traversed to obtain the minimum horizontal coordinate and vertical coordinate, as well as the maximum horizontal coordinate and vertical coordinate, so as to determine the boundary area.

[0093] Create an empty root node of the spatial index data structure, and the boundary area of ​​the root node is the boundary area Y;

[0094] Initially define a small grid side length, for example, 10 meters. According to the area S_j of the boundary area of ​​leaf node j and the number of nodes N_j, dynamically adjust the small grid side length to obtain the ideal small grid side length so that each small grid contains approximately 5-10 nodes.

[0095] Traverse all nodes, for each node i, start from the root node of the spatial index data structure, if the current node is a leaf node j (the bottom-level node, their boundary area will not be divided again, and all actual node data are stored in the leaf node), based on the ideal small grid side length, divide its boundary area into small grids, traverse all nodes in the current boundary area, and assign each node to the corresponding small grid according to its coordinate node position coordinates. Calculate the number of nodes in the small grid divided by the area to obtain the node density; preset a density threshold interval [m_min, m_max], where m_min is the lower limit of the node density and m_max is the upper limit of the node density.

[0096] Traverse all small grids, and for small grid k with node density greater than or equal to m_max, divide small grid k into 4 smaller sub-grids, and redistribute the nodes in the original small grid to the 4 sub-grids; for small grids with node density less than m_min, check its adjacent small grids, find adjacent small grids with node density less than m_min, and use these small grids as grids to be merged; merge the grids to be merged into a large grid, delete the original small grids, until the node density of all small grids is within the density threshold interval [m_min, m_max]; perform shape optimization on all large grids to obtain the final spatial index data structure.

[0097] The constructed spatial index data structure can efficiently search for nodes in a given area. The search process is also a recursive process, starting from the root node. Based on the positional relationship between the query area and the boundary area of ​​the current node, it is decided whether to continue recursively searching for child nodes.

[0098] The method of performing shape optimization on all large grids includes:

[0099] Collect the node position coordinates of all nodes in the grid to be merged to form a point set P;

[0100] Calculate the convex hull PU(P) of the point set P. For any pair of parallel edges X1 = [(u.x, u.y), (v.x, v.y)] and X2 = [(u'.x, u'.y), (v'.x, v'.y)] on the convex hull PU(P), where u.x is the abscissa of the starting point of the parallel edge X1, u.y is the ordinate of the starting point of the parallel edge X1, v.x is the abscissa of the ending point of the parallel edge X1, v.y is the ordinate of the ending point of the parallel edge X1; u'.x is the abscissa of the starting point of the parallel edge X2, u'.y is the ordinate of the starting point of the parallel edge X2, v'.x is the abscissa of the ending point of the parallel edge X2, v'.y is the ordinate of the ending point of the parallel edge X2; calculate the distance d between the parallel edges X1 and X2 X1,X2 ;

[0101]

[0102] Traverse all pairs of parallel edges on the convex hull PU(P), calculate the distances between them, and find the pair of parallel edges corresponding to the minimum distance d_min; using this pair of parallel edges as the opposite sides, construct a rectangle with the minimum perimeter, and the other two sides of the rectangle with the minimum perimeter are the sides perpendicular to this pair of parallel edges; use the rectangle with the minimum perimeter as the boundary region of the merged large grid.

[0103] The method for dynamically adjusting the side length of the small grid includes:

[0104] Set the minimum value l_min and the maximum value l_max of the side length of the small grid. For example, l_min = 5 meters and l_max = 50 meters.

[0105] Based on the number of nodes N_j in the boundary region of the leaf node j, calculate the expected number of small grids M_j in the boundary region of the leaf node j;

[0106] where, is the ceiling function to ensure that there is at least one node in each small grid.

[0107] Then, based on the expected number of small grids M_j, the number of nodes N_j, and the area S_j, calculate the expected side length of the small grid in the boundary region of the leaf node j Make the sizes of the small grids in the dense area and the sparse area more reasonable, and make the side lengths of the small grids in the dense area decrease appropriately and the side lengths of the small grids in the sparse area increase appropriately, so as to better balance the splitting accuracy and the calculation overhead.

[0108] If l_j > l_max, set the ideal side length of the small grid to l_max; if l_j < l_min, set the ideal side length of the small grid to l_min; otherwise, set the ideal side length of the small grid to l_j.

[0109] Traverse all nodes i (i from 1 to n), and calculate the communication capability of node i based on the node physical status information Among them, Pt is the transmission power, Gt is the transmitting antenna gain, Gr is the receiving antenna gain, γ is the wavelength of the wireless signal, Pr is the receiving sensitivity, L is the system loss factor, f(E_i,ρ_i) is the influence function; E_i is the node energy level of node i, g(ρ_i) is the environmental function; ρ_i is the characteristic parameter of the environment in which node i is located.

[0110] It should be noted that the transmit power indicates the power of the radio signal emitted by the transmitter, which is determined by the hardware of the transmitter and can be obtained from the equipment manual or actual measurement; the transmit antenna gain indicates the ability of the transmit antenna to concentrate and radiate the transmit power in a specific direction. The higher the gain value, the more the antenna concentrates energy in the main radiation direction. It depends on the antenna type and design and can be obtained from the antenna manual.

[0111] The receiving antenna gain is similar to the transmitting antenna gain, which indicates the gain of the receiving antenna to the received signal in its radiation direction, and is also obtained from the antenna manual; the receiving sensitivity indicates the minimum signal power that the receiver can detect and demodulate. The higher the sensitivity, the weaker the signal the receiver can detect. It depends on the hardware design and manufacturing process of the receiver and is obtained from the equipment manual.

[0112] The system loss factor represents the power loss caused by various reasons (such as fading, shadow effect, etc.) during the propagation of wireless signals. Specifically, under a known transmission power, multiple receiving points at different distances are selected; the received signal power at each receiving point is measured respectively; the received power under ideal conditions is calculated using a theoretical model (such as a free space propagation model); the measured received power is compared with the ideal received power to calculate the power difference; the system loss factor is this power difference.

[0113] Influence function f(E_i,ρ_i)=E_i×exp(-ρ_i);

[0114] Environment Functions Among them, α, β and It is an adjustable parameter.

[0115] Characteristic parameters Among them, F_iJ is the distance from node i to the Jth obstacle, r_i is the perception radius of node i, which is used to determine the obstacles within the consideration range; P_I is the power intensity of the Ith interference source, D_iI is the distance from node i to the Ith interference source; w_1 and w_2 are weight coefficients.

[0116] The communication coverage area R_i of node i is determined based on its communication capability ra_i; R_i is a circular area with the node position coordinates of node i as the center and the communication capability ra_i as the radius.

[0117] Use the spatial index data structure to search for the neighbor node set N_i in the region R_i; for each node in the neighbor node set N_i of node i, record it as a neighbor node s.

[0118] Based on the node position coordinates between node i and its neighbor node s, calculate the geographic distance dist(i,s) between node i and its neighbor node s.

[0119] If dist(i,s)≤min(ra_i,ra_s), there is a physical connection between node i and neighbor node s, where ra_s is the communication capability of neighbor node s; if there is a physical connection, A[i,s] is assigned a value of 1, otherwise it remains 0, and A[i,s] is an element in the sparse matrix A. Repeat until all nodes are traversed; the final sparse matrix A is the adjacency matrix.

[0120] Furthermore, the methods of constructing a weighted directed graph model for power distribution include:

[0121] Initialize an n×n weighted adjacency matrix W with all elements initially set to 0. Traverse all pairs of nodes i and f. If the corresponding element A[i,f] in the adjacency matrix between nodes i and f is 1, calculate the edge weight w_if from node i to node f and assign it to W[i,f]. W[i,f] is an element in the weighted adjacency matrix W.

[0122] Among them, da_if represents the data flow from node i to node f, m_da is the maximum data flow in the entire network, del_if represents the data transmission delay from node i to node f, m_del represents the maximum data transmission delay in the entire network, dr_if represents the data packet loss rate from node i to node f, and its value range is [0, 1]; en_i is the node energy level of node i, en_f is the node energy level of node f; α1, β1, γ1 and δ1 are all balancing weight parameters, and α1+β1+γ1+δ1=1, which is used to adjust the relative importance of each parameter to the edge weight.

[0123] Repeat until all nodes are traversed to obtain a complete weighted adjacency matrix W; based on the weighted adjacency matrix W, construct a preliminary power distribution weighted directed graph model G = (V, E, W1); where V is a set of nodes, E is a set of edges, representing the physical connection relationship between nodes, and W1 is a set of edge weights, representing the data transmission status between nodes; the preliminary power distribution weighted directed graph model is hierarchically linked to obtain a power distribution weighted directed graph model.

[0124] Define the hierarchy of the preliminary power distribution weighted directed graph model, which includes the number of levels M3 and node types; both are obtained according to the actual structure and equipment type of the distribution network; usually it can be divided into the following layers: transformer layer, switchgear layer, distribution box layer, smart meter layer, etc. Each layer represents a type of node in the distribution network, and nodes at the same level have similar functions and importance.

[0125] Traverse all nodes, assign corresponding levels to them according to the types of nodes, and number the levels to obtain level numbers; for example, 0 represents the transformer level, 1 represents the switch cabinet level, and so on.

[0126] According to the level of the nodes, the nodes belonging to the same level are grouped. For each level q, an empty subgraph G_q = (V_q, E_q) is initialized, all the nodes of this layer are added to V_q, and all pairs of nodes of this layer are traversed. If there is a physical connection relationship, an edge is added from the subgraph G_q to E_q. V_q is the set of nodes in the subgraph G_q, and E_q is the edge set between the nodes in the subgraph G_q.

[0127] Traverse all edges z in E. If the edge z belongs to different levels, add a special edge between the corresponding subgraphs. Special edges are used to connect subgraphs of different levels, reflecting the connection relationship between levels.

[0128] For nodes in the same level, the edge weights are calculated in the same way as when constructing the weighted adjacency matrix W.

[0129] For the special edge (c, b) connecting different levels, define its weight w_cb, where c and b are the two nodes connected by the special edge.

[0130] Where L_c is the level number of the level to which node c belongs, L_b is the level number of the level to which node b belongs, dS_cb is the physical distance between node c and node b, BW_cb is the bandwidth of the communication link between node c and node b, lay_cb is the communication delay between node c and node b (the time required for data to be transmitted from one node to another), θ1, μ1 and ε1 are special edge weight parameters. The influence of each parameter on the special edge weight can be adjusted according to actual needs. For example, if you pay more attention to the difference in hierarchical importance, you can increase the value of θ1 appropriately; if you pay more attention to the physical distance, you can increase The value of .

[0131] M3 subgraphs and the special edges connecting them constitute a power distribution weighted directed graph model; the power distribution weighted directed graph model can better reflect the hierarchical characteristics of the network and provide a basis for adopting different optimization strategies at different levels.

[0132] Furthermore, the method of calculating the shortest weighted path between nodes and the path cost corresponding to the shortest weighted path includes:

[0133] Define any node in the power distribution weighted directed graph model as the source node s', and define the nodes other than the source node s' as the target nodes; for the source node s', initialize its distance estimation value di(s') and path cost estimation value co(s') to 0, and for other nodes v' except the source node s', initialize v''s distance estimation value di(v') and path cost estimation value co(v') to infinity.

[0134] If there is a direct edge L' between v' and s', di(v') and co(v') are updated to the weight of edge L'.

[0135] Create a priority queue pq to store unprocessed nodes, and add all nodes v' and their current distance estimates di(v') and path cost estimates co(v') to the priority queue pq.

[0136] Define the comprehensive scoring function, the expression of the comprehensive scoring function is:

[0137] sc(v')=α3×di(v')+β3×co(v')+γ3×pe(v'); where sc(v') is the comprehensive score of node v', pe(v') is the penalty function value of node v', and α3, β3 and γ3 are score weight parameters.

[0138] pe(v') = ∑(w_O × FG_O(v')); where w_O is the weight of the Oth penalty factor, and FG_O(v') is the factor function corresponding to the Oth penalty factor, which is used to describe the relevant characteristics or status of node v'. Among them, RL(v') is the data cache capacity of node v', SY(v') is the remaining energy of node v' (the current remaining available energy level of the node), and U1 is the factor balance parameter, which ranges from [0, 1].

[0139] In the priority queue pq, the nodes are sorted from small to large according to the comprehensive score, and the node p' with the smallest comprehensive score is taken out; for each priority neighbor node q' of p' (obtained based on the preset domain radius), the estimated distance from the source node to q' through p' and the estimated path cost from the source node to q' through p' are calculated.

[0140] And calculate the comprehensive score LP_q'p' from the source node to q' through p'. If LP_q'p' is less than the current comprehensive score of q', update the distance estimate, path cost estimate and comprehensive score of q', and add q' to the priority queue; repeat until all nodes are processed or the target node is found.

[0141] For each node, initialize an empty list PUO to store the shortest path from the source node to q', and initialize a variable tot to store the total cost of the path; starting from q', trace back in reverse, add each node to PUO in turn until reaching the source node, and calculate the sum of the weights of the edges between adjacent nodes in PUO as tot; based on PUO and tot, output the shortest weighted path and the corresponding path cost.

[0142] Furthermore, the method for obtaining the optimal path solution includes:

[0143] Initialize m4 ant individuals to form an ant colony, each ant individual represents a possible path solution; the number of ant individuals m4 is set according to the problem scale and the algorithm convergence speed.

[0144] The solution space is constructed using the shortest weighted path between each node and the corresponding path cost; the solution space contains all feasible path combination schemes; specifically, each element represents a path from a source node to a destination node, including the node sequence of the path and the corresponding path cost value.

[0145] The solution space is represented using the following data structure:

[0146] Solution space = {(source node 1, destination node 1): [(node ​​sequence 1, path cost 1), (node ​​sequence 2, path cost 2), ...], (source node 1, destination node 2): [(node ​​sequence 1, path cost 1), ...], ... (source node n, destination node n): [(node ​​sequence 1, path cost 1), ...]}.

[0147] Each ant individual gradually builds a path in the solution space according to the transition probability; the transition probability is inversely proportional to the length of the path and the path cost;

[0148] Specifically, the calculation formula of the transition probability is:

[0149] Among them, P(k')_q1,q2 represents the probability that the k'th ant individual transfers from node q1 to node q2, τ'_q1,q2 represents the pheromone (a virtual chemical pheromone used to simulate the behavior of real ants releasing and sensing pheromones during foraging) concentration on the edge from node q1 to node q2. The higher the pheromone concentration, the greater the probability of being selected. δ'_q1,q2 is the path cost from node q1 to node q2. α4 and β4 are the pheromone importance factor and heuristic importance factor, respectively, which are used to control the relative importance of the two types of information.

[0150] When all ant individuals complete the construction of the path, the fitness function is defined and the value of the fitness function of each path is calculated, which is recorded as fitness; the fitness function is a weighted combination of the path length, path cost, and load balancing.

[0151] Load Balancing Among them, x_q1 is the load on node q1, and xtotal is the total load of all nodes;

[0152] According to the fitness of each path, pheromone is released on the corresponding path. The higher the pheromone concentration, the greater the probability that the path will be selected; the optimal solution of the current iteration is obtained.

[0153] Specifically, the formula for releasing pheromones is:

[0154] τ'_q1,q2=(1-ρ2)×τ'_q1,q2+Δτ'_q1,q2; where ρ2 is the pheromone volatility coefficient, and its value range is (0, 1); Δτ'_q1,q2 is the sum of the newly released pheromones by all ant individuals on the edge from node q1 to node q2.

[0155] Update the entire ant colony according to the optimal solution of the current iteration, preset the fitness threshold, delete the ant individuals whose fitness function value is less than the fitness threshold, and generate new ant individuals to join the ant colony;

[0156] Repeat until the preset maximum number of iterations is reached, and the final output ant individual is the current optimal path plan.

[0157] Based on the obtained optimal path plan, the topology of the distribution network is replanned and adjusted to enable the network to self-organize and achieve balanced load distribution.

[0158] By simulating the foraging behavior of ants, combining the shortest path and path cost between nodes, searching for the optimal path solution globally, and dynamically adjusting the network topology, the self-organization and load balancing of the distribution network are effectively achieved.

[0159] This embodiment can realize the intelligent self-organization of the distribution network. By constructing a weighted directed graph model for distribution, the concept of hierarchy is introduced to truly reflect the actual situation of the network and the difference in node importance, distinguish the importance of different types of nodes, and provide a basis for adopting differentiated optimization strategies at different levels. By using the improved ant colony algorithm, combined with the shortest path and path cost information between each node, the optimal path solution is searched globally, and the network topology is dynamically adjusted, so that the distribution network can be autonomously optimized and reconstructed according to the real-time status, thereby improving the adaptability and flexibility of the network and effectively coping with complex and changeable loads and environments; secondly, load balancing in the distribution network is realized. During planning, the situation where some nodes are overloaded while other nodes are idle is avoided, and resource utilization efficiency is improved. A multi-objective optimization strategy is adopted, taking into account multiple indicators such as path length, cost and load balancing, so that the obtained path solution is more comprehensive and reasonable, fully reflecting the concept of intelligence and adaptability, and greatly improving the adaptability, flexibility and resource utilization of the network by dynamically optimizing the network topology and path planning.

[0160] Example 2

[0161] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, which provides a power distribution Internet of Things self-organizing network system, including:

[0162] A data acquisition module, used to obtain node physical status information and node data status information of n nodes in the power distribution network topology;

[0163] The matrix construction module is used to obtain the adjacency matrix between n nodes based on the node physical state information;

[0164] The model hierarchical fitting module constructs a power distribution weighted directed graph model based on node data status information and adjacency matrix;

[0165] The path preliminary calculation module, based on the power distribution weighted directed graph model, uses the optimization target algorithm to calculate the shortest weighted path between nodes and the path cost corresponding to the shortest weighted path;

[0166] The solution output module uses an improved ant colony algorithm to dynamically obtain the optimal path solution based on the shortest weighted path between each node and the path cost corresponding to the shortest weighted path, and uses the optimal path solution to adjust the distribution network topology. The modules are connected by wired and / or wireless means to achieve data transmission between modules.

[0167] Example 3

[0168] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the above-mentioned distribution Internet of Things self-organizing network method is implemented.

[0169] Since the electronic device introduced in this embodiment is an electronic device used to implement a distribution Internet of Things self-organizing network method in the embodiment of this application, based on the distribution Internet of Things self-organizing network method introduced in the embodiment of this application, the technical personnel of this field can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application is not described in detail here. As long as the technical personnel of this field implement the electronic device used in the distribution Internet of Things self-organizing network method in the embodiment of this application, it belongs to the scope of protection of this application.

[0170] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0171] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A self-organizing network method for power distribution Internet of Things, characterized in that: Including: S1. Obtain the node physical state information and node data state information of n nodes in the distribution network topology; S2. Obtain the adjacency matrix between n nodes according to the node physical state information; S3. Build a distribution weighted directed graph model based on the node data state information and the adjacency matrix; S4. Based on the distribution weighted directed graph model, use the optimization objective algorithm to calculate the shortest weighted path between each node and the path cost corresponding to the shortest weighted path; S5. Based on the shortest weighted path between each node and the path cost corresponding to the shortest weighted path, use the improved ant colony algorithm to dynamically obtain the optimal path scheme, and use the optimal path scheme to adjust the distribution network topology.

2. A power distribution Internet of Things self-organizing network method according to claim 1, characterized in that: The node physical state information includes node position coordinates, node communication distance, and node communication bandwidth; The node data state information includes the data traffic, data cache capacity, data transmission delay, data packet loss rate, node energy level, and data transmission volume of the node.

3. A power distribution Internet of Things self-organizing network method according to claim 2, characterized in that: The obtaining method of the adjacency matrix includes: Initialize an n×n sparse matrix A and construct a spatial index data structure of the node, traverse all nodes i, and calculate the communication capability of node i based on the node physical state information. Among them, Pt is the transmission power, Gt is the transmission antenna gain, Gr is the receiving antenna gain, γ is the wavelength of the wireless signal, Pr is the receiving sensitivity, L is the system loss factor, f(E_i,ρ_i) is the influence function; E_i is the node energy level of node i, g(ρ_i) is the environment function; ρ_i is the characteristic parameter of the environment where node i is located; The influence function f(E_i,ρ_i) = E_i×exp(-ρ_i); Environment Functions Among them, α, β and is an adjustable parameter; Characteristic parameters Among them, F_iJ is the distance from node i to the Jth obstacle, r_i is the perception radius of node i; P_I is the power intensity of the Ith interference source, D_iI is the distance from node i to the Ith interference source; w_1 and w_2 are weight coefficients; Determine its communication coverage area R_i based on the communication ability ra_i of node i; R_i is a circular area with the node position coordinates of node i as the center and the communication ability ra_i as the radius; Use the spatial index data structure to find the set of neighbor nodes N_i within the area R_i; for each node in the set of neighbor nodes N_i of node i, denote it as the neighbor node s; Based on the node position coordinates between node i and neighbor node s, calculate the geographical distance dist(i, s) between node i and neighbor node s; If dist(i, s) ≤ min(ra_i, ra_s), there is a physical connection relationship between node i and neighbor node s, where ra_s is the communication ability of neighbor node s; if there is a physical connection relationship, assign A[i, s] to 1, otherwise keep it as 0, A[i, s] is an element in the sparse matrix A, repeat until all nodes are traversed; the finally obtained sparse matrix A is the adjacency matrix.

4. A power distribution Internet of Things self-organizing network method according to claim 3, characterized in that: The construction method of the spatial index data structure includes: Define the boundary area Y of the spatial index data structure; create an empty root node of the spatial index data structure, and the boundary area of the root node is the boundary area Y; Initially define a small grid side length, and dynamically adjust the small grid side length according to the area S_j of the boundary area of leaf node j and the number of nodes N_j to obtain the ideal small grid side length; The method of dynamically adjusting the small grid side length includes: Set the minimum value l_min and the maximum value l_max of the small grid side length; calculate the expected number of small grids M_j within the boundary area of leaf node j based on the number of nodes N_j in the boundary area of leaf node j; in, To round up; Based on the expected number of small grids M_j, the number of nodes N_j and the area S_j, the expected small grid side length in the boundary area of ​​leaf node j is calculated If l_j > l_max, set the ideal small grid side length to l_max, if l_j < l_min, set the ideal small grid side length to l_min; otherwise, set the ideal small grid side length to l_j; Traverse all nodes, for each node i, start from the root node of the spatial index data structure, if the current node is a leaf node j, divide its boundary area into small grids based on the ideal small grid side length, traverse all nodes in the current boundary area, and assign each node to the corresponding small grid according to its coordinate node position coordinates, calculate the number of nodes in the small grid divided by the area, and get the node density; preset a density threshold interval [m_min, m_max], where m_min is the lower limit of node density and m_max is the upper limit of node density; Traverse all small grids, and for small grid k with node density greater than or equal to m_max, divide small grid k into 4 smaller sub-grids, and redistribute the nodes in the original small grid to the 4 sub-grids; for small grids with node density less than m_min, check its adjacent small grids, find adjacent small grids with node density less than m_min, and use these small grids as grids to be merged; merge the grids to be merged into a large grid, delete the original small grids, until the node density of all small grids is within the density threshold interval [m_min, m_max]; perform shape optimization on all large grids to obtain the final spatial index data structure.

5. A power distribution Internet of Things self-organizing network method according to claim 4, characterized in that: The method of performing shape optimization on all large grids includes: Collect the node position coordinates of all nodes in the grid to be merged to form a point set P; Calculate the convex hull PU(P) of the point set P. On the convex hull PU(P), for any pair of parallel edges X1=[(ux,uy),(vx,vy)] and X2=[(u'.x,u'.y),(v'.x,v'.y)], where ux is the abscissa of the starting point of the parallel edge X1, uy is the ordinate of the starting point of the parallel edge X1, vx is the abscissa of the end point of the parallel edge X1, and vy is the ordinate of the end point of the parallel edge X1; u'.x is the abscissa of the starting point of the parallel edge X2, u'.y is the ordinate of the starting point of the parallel edge X2, v'.x is the abscissa of the end point of the parallel edge X2, and v'.y is the ordinate of the end point of the parallel edge X2; calculate the distance d between the parallel edges X1 and X2. X1,X2 , Traverse all pairs of parallel edges on the convex hull PU(P), calculate the distance between them, and find the pair of parallel edges corresponding to the minimum distance d_min; use this pair of parallel edges as opposite sides to construct a minimum perimeter rectangle, and the other two sides of the minimum perimeter rectangle are perpendicular to the pair of parallel edges; use the minimum perimeter rectangle as the boundary area of ​​the merged large grid, and complete the shape optimization of all large grids.

6. A power distribution Internet of Things self-organizing network method according to claim 5, characterized in that: The method of constructing the power distribution weighted directed graph model includes: Initialize an n×n weighted adjacency matrix W, with all elements initially set to 0. Traverse all pairs of nodes i and f. If the corresponding element A[i,f] of node i and node f in the adjacency matrix is ​​1, calculate the edge weight w_if from node i to node f and assign it to W[i,f]. W[i,f] is an element in the weighted adjacency matrix W. Among them, da_if represents the data flow from node i to node f, m_da is the maximum data flow in the entire network, del_if represents the data transmission delay from node i to node f, m_del represents the maximum data transmission delay in the entire network, dr_if represents the data packet loss rate from node i to node f; en_i is the node energy level of node i, en_f is the node energy level of node f; α1, β1, γ1 and δ1 are all balancing weight parameters; Repeat until all nodes are traversed to obtain a complete weighted adjacency matrix W; based on the weighted adjacency matrix W, construct a preliminary power distribution weighted directed graph model G = (V, E, W1); where V is a set of nodes, E is a set of edges, representing the physical connection relationship between nodes, and W1 is a set of edge weights, representing the data transmission status between nodes; the preliminary power distribution weighted directed graph model is hierarchically linked to obtain a power distribution weighted directed graph model.

7. A power distribution Internet of Things self-organizing network method according to claim 6, characterized in that: The method of hierarchically linking the preliminary power distribution weighted directed graph model includes: Define the hierarchy of the preliminary power distribution weighted directed graph model, which includes the number of levels M3 and the node type; traverse all nodes, assign corresponding levels to them according to the type of the node, and number the levels to obtain the level number; According to the level of the nodes, the nodes belonging to the same level are grouped. For each level q, an empty subgraph G_q = (V_q, E_q) is initialized, and all the nodes of this level are added to V_q. All pairs of nodes in this level are traversed. If there is a physical connection relationship, an edge is added from the subgraph G_q to E_q. V_q is the set of nodes in the subgraph G_q, and E_q is the edge set between the nodes in the subgraph G_q. Traverse all edges z in E. If the edges z belong to different levels, add a special edge between the corresponding subgraphs; For nodes in the same level, the edge weights are calculated in the same way as when constructing the weighted adjacency matrix W; For the special edge (c, b) connecting different levels, define its weight w_cb, where c and b are the two nodes connected by the special edge; Where, L_c is the level number of the level to which node c belongs, L_b is the level number of the level to which node b belongs, dS_cb is the physical distance between node c and node b, BW_cb is the bandwidth of the communication link between node c and node b, lay_cb is the communication delay between node c and node b, θ1, μ1 and ε1 are special edge weight parameters; M3 subgraphs and the special edges connecting them constitute the power distribution weighted directed graph model.

8. A power distribution Internet of Things self-organizing network method according to claim 7, characterized in that: The method of calculating the shortest weighted path between nodes and the path cost corresponding to the shortest weighted path includes: Define any node in the power distribution weighted directed graph model as the source node s', and define nodes other than the source node s' as the target node; for the source node s', initialize its distance estimation value di(s') and path cost estimation value co(s') to 0; for other nodes v' except the source node s', initialize v''s distance estimation value di(v') and path cost estimation value co(v') to infinity; If there is a direct edge L' between v' and s', update di(v') and co(v') to the weight of edge L'; Create a priority queue pq, add all nodes v' and their current distance estimates di(v') and path cost estimates co(v') to the priority queue pq; Define the comprehensive scoring function, the expression of the comprehensive scoring function is: sc(v')=α3×di(v')+β3×co(v')+γ3×pe(v'); where sc(v') is the comprehensive score of node v', pe(v') is the penalty function value of node v', and α3, β3 and γ3 are score weight parameters; pe(v') = ∑(w_O × FG_O(v')); where w_O is the weight of the Oth penalty factor, and FG_O(v') is the factor function corresponding to the Oth penalty factor; Among them, RL(v') is the data cache capacity of node v', SY(v') is the residual energy of node v', and U1 is the factor balance parameter; In the priority queue pq, sort the nodes from small to large according to the comprehensive score, and take out the node p' with the smallest comprehensive score; for each priority neighbor node q' of p', calculate the estimated distance from the source node to q' through p' and the estimated path cost from the source node to q' through p'; And calculate the comprehensive score LP_q'p' from the source node to q' through p'. If LP_q'p' is less than the current comprehensive score of q', update the distance estimate, path cost estimate and comprehensive score of q', and add q' to the priority queue; repeat until all nodes are processed or the target node is found; For each node, initialize an empty list PUO and variable tot; start from q' and trace back, add each node to PUO in turn until reaching the source node, and calculate the sum of the weights of the edges between adjacent nodes in PUO as tot; output the shortest weighted path and the corresponding path cost based on PUO and tot.

9. A power distribution Internet of Things self-organizing network method according to claim 8, characterized in that: The method for obtaining the optimal path solution includes: Initialize m4 ant individuals to form an ant colony, each ant individual represents a possible path solution; The solution space is constructed using the shortest weighted path between nodes and the corresponding path cost; the solution space contains all feasible path combination solutions; Each ant individual gradually builds a complete path in the solution space according to the transition probability; the calculation formula of the transition probability is: Where P(k')_q1,q2 represents the probability that the k'th ant individual transfers from node q1 to node q2, τ'_q1,q2 represents the pheromone concentration on the edge from node q1 to node q2, δ'_q1,q2 is the path cost from node q1 to node q2, α4 and β4 are the pheromone importance factor and heuristic importance factor respectively; When all ant individuals complete the construction of the path, the fitness function is defined and the value of the fitness function of each path is calculated, which is recorded as fitness; the fitness function is a weighted combination of the path length, path cost, and load balance; Load Balancing Among them, x_q1 is the load on node q1, and xtotal is the total load of all nodes; According to the fitness of each path, pheromone is released on the corresponding path to obtain the optimal solution of the current iteration; The formula for releasing pheromones is: τ'_q1,q2=(1-ρ2)×τ'_q1,q2+Δτ'_q1,q2; where ρ2 is the pheromone volatility coefficient, ranging from (0, 1); Δτ'_q1,q2 is the sum of the newly released pheromones of all ant individuals on the edge from node q1 to node q2; The entire ant colony is updated according to the optimal solution of the current iteration, a fitness threshold is preset, ant individuals whose fitness function value is less than the fitness threshold are deleted, and new ant individuals are generated to join the ant colony; this is repeated until the preset maximum number of iterations is reached, and the ant individual finally output is the current optimal path solution.

10. A power distribution Internet of Things self-organizing network system, which is used to implement a power distribution Internet of Things self-organizing network method according to any one of claims 1 to 9, characterized in that: include: A data acquisition module, used to obtain node physical status information and node data status information of n nodes in the power distribution network topology; The matrix construction module is used to obtain the adjacency matrix between n nodes based on the node physical state information; The model hierarchical fitting module constructs a power distribution weighted directed graph model based on node data status information and adjacency matrix; The path preliminary calculation module, based on the power distribution weighted directed graph model, uses the optimization target algorithm to calculate the shortest weighted path between nodes and the path cost corresponding to the shortest weighted path; The solution output module uses an improved ant colony algorithm to dynamically obtain the optimal path solution based on the shortest weighted path between each node and the path cost corresponding to the shortest weighted path, and uses the optimal path solution to adjust the distribution network topology. Each module is connected by wired and / or wireless means.

Citation Information

Patent Citations

  • Power distribution Internet of Things ad hoc network method and system

    CN112261087A

Cited By

  • Supply chain network optimization method and system

    CN119273251A

  • Data analysis system driven by intelligent algorithm

    CN119782952A