Network edge adding optimization method applied to urban power communication network

Abstract

The invention discloses a network edge-adding optimization method applied to an urban power communication network, relates to the technical field of power communication networks, and solves the problems that network edge-adding optimization of an existing power communication network cannot give consideration to multiple targets, lacks multi-scene robustness and gives insufficient consideration to fragile areas. The method comprises the following steps: constructing an urban power communication network model; determining a key area, and defining an annular zone; constructing a multi-fault scene damage set, and setting a weight; generating a candidate edge set in a multi-fault scene; evaluating indexes of each candidate edge in each fault scene, summarizing the indexes according to weights, adjusting the weights of the candidate edges in the key area, and calculating scores of the candidate edges; and selecting the most balanced edge adding strategy according to the Pareto frontier and the maxmin, and finally outputting a newly added link and an evaluation index value. The defect that a traditional method is optimal only under a single hypothesis and loses efficacy in actual combats is overcome.

Description

Technical Field

[0001] This invention relates to the field of power communication network technology, and specifically to a network edge optimization method applied to urban power communication networks. Background Technology

[0002] With the acceleration of urbanization, urban electricity demand is constantly growing, and the role of power communication networks in urban power systems is becoming increasingly important. As a crucial component of smart grids, urban power communication networks undertake critical tasks such as real-time monitoring, dispatching, load management, and fault diagnosis of urban power equipment. With the rapid development of smart grids, urban power communication networks are no longer simply information transmission channels, but rather core platforms supporting the efficient operation and stable power supply of the power system. However, with the increasing load on urban power communication networks, the increasing complexity of network structures, and the threats of natural disasters and external malicious attacks, the vulnerability of urban power communication networks is gradually becoming apparent. Especially in terms of reliability at critical nodes and in critical areas, the robustness of the power communication network has become a significant factor affecting the stability of urban power systems.

[0003] In fact, the vulnerability of urban power and communication networks has been exposed many times in history.

[0004] The accidents were all caused by malfunctions or deliberate sabotage of the information and communication systems in the power grid.

[0005] Therefore, network optimization of the power communication network is a necessary measure to ensure the normal operation of the smart grid. This experiment aims to construct a complex network model including tower nodes and real nodes, and to use innovative algorithms to more comprehensively and accurately assess and improve the resilience of the urban power communication network. Summary of the Invention

[0006] This invention addresses the problems of existing power communication network edge optimization failing to consider multiple objectives, lacking robustness across multiple scenarios, and insufficiently addressing vulnerable areas by providing a network edge optimization method applicable to urban power communication networks.

[0007] A network edge optimization method applied to urban power communication networks is implemented through the following steps:

[0008] Step 1: Construct a model of the urban power communication network; simultaneously, conduct a vulnerability analysis of the urban power network to identify key areas;

[0009] Step 2: Define the annular zone based on the key areas identified in Step 1;

[0010] Step 3: Construct a set of damage scenarios with multiple failure scenarios and assign weights to each scenario;

[0011] Step 4: In multi-fault scenarios, a candidate edge set is generated by comprehensively adopting cross-component bridging, cut-point neighbor closure, community bridging, and loop closure and around-region bridging strategies.

[0012] Step 5: Evaluate the indicators of each candidate edge under each fault scenario and summarize them according to their weights. At the same time, adjust the weights of candidate edges located in key areas and calculate the scores of the candidate edges.

[0013] Step 6: Select the most balanced edge-adding strategy based on the Pareto front and maxmin, and select candidate edges until the upper limit is reached or there is no more positive return. Finally, output the newly added links and evaluation index values.

[0014] The beneficial effects of this invention are:

[0015] This invention models local circular damage areas and incorporates them into edge-addition decisions. It prioritizes "loop weaving" and "circular bridging" in the ring zone outside the damage area, enabling the network to maintain main network connectivity even when critical areas fail, significantly reducing the risk of the topology being split into multiple islands. At the same time, it introduces scenario weight vectors to perform weighted evaluations of various types of failures such as "targeted attacks, random failures, and vulnerable area damage," which can adaptively balance the expected optimal performance with the worst-case scenario guarantee, avoiding the shortcomings of traditional methods that are optimal under a single assumption but fail in actual combat.

[0016] This invention integrates structural candidates such as "cross-component bridging, cut-point neighbor closure, and community bridging" with geographical distance constraints. This approach not only fills critical topological vulnerabilities but also effectively controls the length of new links and construction costs. Compared to single-index methods like degree / betweenness or pure geographical nearest neighbor methods, it is more balanced and feasible. Furthermore, it applies positional penalties to candidate edges within damaged areas and prioritizes loop candidates, ensuring that new links are concentrated in areas most likely to generate bypass redundancy and reconnection. This avoids allocating budget to in-loop edges that would fail directly in critical scenarios, thereby achieving higher resilience benefits with the same investment. Attached Figure Description

[0017] Figure 1 This is a flowchart of the network edge optimization method applied to urban power communication networks as described in this invention;

[0018] Figure 2 This is a schematic diagram illustrating the concept of the damaged area in the method described in this invention;

[0019] Figure 3 This is a topology diagram of the power communication network in a certain city.

[0020] Figure 4 Figure showing the results of edge optimization for the power communication network in a certain city.

[0021] Figure 5The graphs show the network performance results before and after adding edges, and after the network is damaged by adding edges. Detailed Implementation

[0022] Specific Implementation Method 1: Combination Figures 1 to 5 This embodiment describes a network edge optimization method applied to urban power communication networks. The method is implemented through the following steps:

[0023] Step 1: In the urban power network structure, spatial modeling of the power communication network of a certain city's urban area is carried out; at the same time, vulnerability analysis of the urban power network is conducted to identify key areas;

[0024] Urban power communication sites are abstracted as network nodes, and optical fiber communication links are abstracted as network edges. Since the mutual transmission between services can be abstracted as a weighted undirected graph, a power communication network model G exists as follows:

[0025] Based on the direct mapping of station and tower nodes including latitude and longitude coordinate information, the urban power communication network model is defined as follows: , For a collection of sites, that is , Indicates the first There are 10 sites, and the number of sites is 100. E is the set of edges corresponding to communication links in the power communication network, i.e. , Representing a power communication network model The site With the site The edges between them. In the set Zhongruo =1 indicates that at the site With the site If there is an edge connecting them, =0, which means that at the site With the site There are no edge connections.

[0026] In this embodiment, based on the spatial distribution of urban power communication sites and towers, and combined with historical fault cases and potential risk points, areas susceptible to concentrated attacks or natural disasters are selected as key areas. These areas are modeled using the coordinates of a circle's center. The damage radius r parameter is determined and used as the basis for subsequent setting and optimization evaluation of the annular zone.

[0027] Step 2: Define the annular zone based on the identified key areas;

[0028] Employing a key area-based power communication network model To carry out the damage, based on the determined coordinates of the center of the circle. Set up an annulus area for the critical area with a damage radius of r. As shown Figure 2 in the circular area indicated by the red dashed line, which is the critical area. Define the critical area as D={ | , where and are the longitude and latitude coordinates of site ; that is, all sites, edges, and pole and tower nodes within this area will be damaged, and the damaged sites and edges will be removed from the model. As shown Figure 2 the red dashed line link in is the failed link, and the set of failed links is E D ; the red nodes are the failed sites, and the set of failed sites is V D .

[0029] In this embodiment, the damage method adopted is: obtain the location information of each site in the power communication network model , traverse the location of each site, and calculate the Euclidean distance between the site and the center of the critical area . If , it is considered that the site is within the critical area. Traverse all sites within the critical area, and delete the sites and their associated edges. Traverse each edge, and calculate the Euclidean distance between the edge and the center of the critical area . If , it is considered that the edge crosses the critical area.

[0030] Since the nodes / links within the critical area will "fail" with the scenario, adding edges within the area has little effect. Therefore, alternative paths are established in the area adjacent to the area (annulus) outside the area, which can most effectively maintain connectivity and reduce detours and enhance resilience when damage occurs. The annulus area is a priority edge-adding buffer zone that is "close to the critical area but not within the area", used to weave detour redundancy and cross-component short bridges, so that the network can still maintain high connectivity and small efficiency loss when this vulnerable area is attacked. Focus on defining the annulus area:

[0031] R={(x,y)∣ r<d((x,y),c)≤(1 + γ)r}; where (x,y) is the location of a certain node in the network, c is the center location of the critical area, d((x,y),c) represents the geometric distance between the node location and the center of the circle, and γ represents the thickness coefficient of the annulus area. The value of γ is determined by experimental comparison of multiple γ values. The existence of the critical area is not only used for evaluation, but also directly affects the edge-adding strategy.

[0032] Step Three, damage sets and weights for multi-fault scenarios;

[0033] To cover three typical failure scenarios of the power communication network (1. targeted attacks, 2. random background failures, 3. localized disasters / construction damage (critical area damage)), this implementation selects five representative scenarios to form a set S.

[0034] Construct scene set S ;

[0035] Targeted attack scenarios: Delete the k nodes with the highest betweenness centrality (scenario s1); Delete the k edges with the highest betweenness centrality (scenario s2).

[0036] Random background failure scenario: with a ratio of p node Randomly delete nodes (scenario s3); in proportion p edge Randomly delete edges (Scenario s4).

[0037] Critical area damage scenario: Delete V D Its associated edge E D (Scene s5).

[0038] Assign scene weight vector w =(1,1,1,1,β), where the value of the key region weight β is determined experimentally: with β as the independent variable, the center and radius of the circle are fixed and the other parameters remain unchanged. Multiple random reproduction experiments are carried out on multiple weight values ​​of β∈{1,2,3,4,5}, and the network model is optimized multiple times. The overall optimization scenario index M(G) after adding edges and after damage is recorded, M(G)∈{E(G), C(G), R(G)}, where E(G) is the network efficiency, C(G) is the connectivity, and R(G) is the robustness. For each β, the mean and standard error of each index after damage are summarized, and the β with the largest mean of each index is used as the weight value.

[0039] In this embodiment, the performance evaluation only calculates three scenario indicators: network efficiency E(G), connectivity C(G), and robustness R(G).

[0040] The formula for network efficiency E(G) (based on the shortest hop count) is as follows:

[0041] ;

[0042] in, For node v i With node v j The shortest path length between, where n is the total number of nodes in the network;

[0043] The formula for connectivity C(G) (the percentage of the largest connected component) is as follows:

[0044] ;

[0045] in, V is the maximum number of connected subgraphs in the network model G, and |V| is the total number of nodes in the current network.

[0046] The formula for robustness R(G) (default optimization objective) is as follows:

[0047] ;

[0048] Where S(p) is the percentage of the largest connected subgraph size after removing a proportion of nodes that were previously removed (removed in descending order of betweenness centrality); This represents the maximum removal percentage.

[0049] Step 4: Candidate edge generation;

[0050] Under multi-scenario constraints, candidate edge generation is no longer a single rule-driven process. Instead, it combines the damage characteristics and structural diagnosis results of different failure scenarios to generate a subset of candidate edges for each scenario and integrates them at the global level. Specifically, this includes:

[0051] 1) Critical area damage scenario: Inside and outside the damage circle and its annular zone, candidate edges are mainly considered for annular closure and bridging around the area. By constructing alternative paths on the outer edge of the damage area, the connectivity defects caused by the regional failure are compensated, thus forming a global candidate edge subset C1.

[0052] 2) Targeted attack scenario: When a high betweenness node or edge is attacked, candidate edges are preferentially closed by cut point neighbor and cross-component bridging to mitigate the risk of global breakage caused by the removal of critical nodes / links, thereby forming a global candidate edge subset C2.

[0053] 3) Random failure scenario: Under the condition of random failure of nodes or links, the candidate edges focus on community bridging and local redundancy reinforcement. By increasing the shortcuts across communities and the local redundant paths, the average reachability and robustness of the network as a whole are improved, thereby forming a global candidate edge subset C3.

[0054] The candidate edge subsets generated in each scenario are integrated to form a global candidate edge set C.

[0055] Step 5: Candidate edge scoring and critical region penalty;

[0056] Step 51: Scene gain refers to the improvement in the target metric after adding candidate edges e to graph G within a set of fault scenarios S (e.g., deleting high betweenness vertices, randomly deleting edges, damaged regions). Scene metric gain is applied to any candidate edge... In each scene The gain set of the above calculation scenario metrics { , } ,in, , These represent the network efficiency, connectivity, and robustness gain values ​​after adding candidate edge e to network model G.

[0057] ;

[0058] in, This is the initial set of indicator values ​​for the network; }∈ , It adds candidate edges to the network model G. The set of three scenario-specific metrics for the network model G: network efficiency, connectivity, and robustness. This indicates that the newly added candidate edge can alleviate the scenario. Damage below;

[0059] Step 52: Set of scene metric gains for each scene Calculate the expected gain (weighted) for each scene s i Corresponding to a scene weight vector We can obtain the overall performance improvement of the network after adding candidate edge e, that is: the expected gain value. It can be expressed as follows:

[0060] ;

[0061] in, ,in, , , Adding candidate edges respectively The improvement in network efficiency, connectivity, and robustness in various scenarios.

[0062] Step 53: If candidate edges If the geometric position falls within the critical region D, then its target index... Adjustments are needed; adjust the formula as follows: ,otherwise, =1;

[0063] Candidate side rating for:

[0064] ;

[0065] Step 6: Select the optimal edge-adding strategy based on the Pareto front and maxmin;

[0066] Robust edge addition optimization and evaluation focusing on key regions (greedy edge selection and stopping criteria) selects the value of the new edge K on the current network to optimize the overall performance of the network model under various failure scenarios.

[0067] Step 61: Only retain Candidates with at least one positive dimension are identified. Compare each component in the candidate set C: if there exists... make ≥ And if at least one dimension is strictly greater than, then it is called Dominate By removing the dominated candidate edges, we obtain the frontier H.

[0068] Step 6.2: To eliminate dimensional differences, perform interval normalization on each component within the frontier H. ∈ Define the final score of adding edge e. The scores are for normalized network efficiency, connectivity, and robustness, respectively. Within the frontier H, a max-min selection method is used: comparing the improvement of each candidate edge across the three metrics, first finding its minimum improvement value, and then selecting the candidate edge that maximizes this minimum value. In other words, the final selected candidate edge should perform evenly across the three metrics, without any particularly prominent weaknesses. The optimal edge is added to the network model G, updating the network structure; the optimal edge is then recorded in the newly added edge sequence.

[0069] Step 63: Continue until the number of added candidate edges reaches the upper limit h or there are no more positive gain edges in the candidate set C.

[0070] Output: New edge sequence, indicators before and after adding edges, and indicators after damage to the key area; also output the edge-added graph and a comparison graph of the three indicators.

[0071] Specific Implementation Method Two: Combination Figures 2 to 5 This embodiment describes the network edge optimization method applied to urban power communication networks as described in Specific Embodiment 1: To illustrate the process of the network edge optimization method based on damaged areas, Figure 2 This example illustrates regional damage, with black nodes representing sites and dashed lines indicating damaged areas (circles). The resilience analysis of a power fiber optic cable network in an urban area of ​​Jilin Province is used as an example.

[0072] 1. Construct a power communication network model;

[0073] In a power communication network in a city in Jilin Province, substations, communication stations, and communication sites in various power plants are abstracted as stations, and the displayed geographical coordinates of each communication station correspond to their positions in the model space; the connections between communication stations are abstracted as undirected edges.

[0074] Construct a model of a power communication network in a part of the urban area of ​​a city in Jilin Province. ,like Figure 3 As shown, Figure 3This is a schematic diagram of a portion of the power communication network in a certain city's urban area. The network includes 282 sites and 329 optical cables.

[0075] According to step two in Specific Implementation Method 1, the circular spatial damage form and method are determined, and an annular zone is defined to accommodate bypass redundancy. This example uses the same critical region as above: damage center c = (125.361, 43.839), radius r = 0.2; and an annular zone R with thickness γ is defined outside the circle. In this implementation, the size of the annular zone is determined based on 10 sets of experimental results. The connectivity C remains constant at 1.00 after adding edges, and only fluctuates slightly after damage (approximately 0.986~1.0000), indicating that the network skeleton already possesses high integrity; the change in efficiency E(G) is also relatively limited. In contrast, robustness R(G) shows more significant differentiation between groups, indicating that the improvement in network resilience mainly comes from the enhancement of local redundancy rather than a further increase in overall connectivity. Based on the principle of "marginal benefit" and mechanism matching, the weight parameters of "skeleton stability" and "redundancy / substitutability" in the resilience objective are set to γ=0.3. The redundancy-first configuration guides the algorithm to generate more loop closures and bridging edges, thereby substantially reducing the bypass cost after damage. When significant network fragmentation is observed (e.g., C<0.95), γ can be dynamically increased to 0.6–0.7 to prioritize skeleton reconnection.

[0076] During damage assessment, the spherical distance (or approximate Euclidean distance) from the center of the circle is first calculated based on the latitude and longitude of the station. When d(vi,c)≤r, vi is determined to be within the damage zone, and the node and its associated edges are deleted from the graph, thus simulating the failure of all nodes and links within the circle. Since adding edges within the damage zone provides limited improvement in robustness, the strategy shifts to prioritizing the construction of alternative paths outside the ring zone to maintain the backbone connectivity of the urban area and reduce detour costs when damage occurs.

[0077] 3. Construct and weight a multi-scenario damage set to highlight the protection capability for fixed critical areas. The scenario set S contains five categories: targeted attacks (s1, s2), random failures (s3, s4), and critical area damage (s5). To reflect the emphasis on fixed critical areas, the scenario weight vector is set as w = (1, 1, 1, 1, β), where the scenario weight for vulnerable areas is β = 3. To determine the scenario weight β for critical areas, five repeated experiments were conducted on β ∈ {1, 2, 3, 4, 5}. Using post-damage robustness Rdmg as the main evaluation index, the results show that when β increases from 1 to 3, the mean value of Rdmg increases from 0.09366 to 0.09539 (+1.85%), while further increasing β from 3 to 5, Rdmg only increases slightly from 0.09539 to 0.09558 (+0.19%), indicating that the benefit is clearly saturated. Meanwhile, the connectivity Cdmg after destruction jumps from 0.99288 to 0.99359 when β≥3, while the efficiency and connectivity resilience remain basically stable after adding edges (for example, the network efficiency Eafter after destruction is about 0.1453–0.1457, with very little change), indicating that increasing β does not bring additional costs.

[0078] Considering the trade-off between "benefits and costs," β=3 is the minimum weight to reach the benefit inflection point: it significantly improves resilience and connectivity after damage while avoiding extremely low marginal benefits and potential overfitting risks caused by further increasing the weight. Therefore, in this implementation, β=3 is used as the default scenario weight. For each scenario si∈S, the damaged graph G(si) is obtained for robustness evaluation of candidate edges.

[0079] 4. Generate a candidate edge set C under geographical constraints, taking into account both structural weak points and redundancy around key areas. Candidate sources include: cross-component bridging, community bridging, cut vertex neighbor closure, loop closure, and loop bridging, supplemented by a small amount of random exploration to avoid local optima. The candidate edge set is deduplicated, and |C| ≤ Cmax.

[0080] 5. Perform multi-scenario robustness scoring on candidate edges and add a critical region location penalty to guide more robust edge connections. For any candidate edge e={u,v}, calculate the scenario metric gain for each scenario si. Then, the scene weight w is used to perform expected aggregation to obtain a weighted expected gain. Simultaneously, if the candidate endpoint falls within the damage circle, it is multiplied by a score adjustment value of 0.3 within the critical region to suppress in-circle repair and encourage out-of-circle detours. For multi-objective scenarios... The Pareto front is obtained by performing non-dominated screening, and the front solution is normalized and then the maximin equilibrium criterion (maximizing the smallest dimension of the three-dimensional gain) is applied.

[0081] 6. Robust edge addition optimization, focusing on key areas, is conducted on the current urban network. This experiment uses adding 5 edges as an example. The algorithm iterates as follows: In each round, the cross-scenario indicator gain is calculated and weighted on the complete candidate set. Candidates that "do not improve" the selected indicator set are eliminated. From the remaining candidates, one edge is selected and added to G according to "Pareto front + maximin". Then, the local candidates are dynamically expanded around the nearest neighbor nodes at both ends of the newly added edge, and the next round begins. The algorithm stops when the budget of 5 edges is reached, or when the candidates are exhausted / the constraints are not met, or when there are no candidates on the front that bring positive improvement. The final output is the set of newly added edges.

[0082] like Figure 4 As shown, Figure 4 The image shows the results of edge optimization on the network; the red areas represent critical regions, and the orange areas represent loop regions. This optimization added 5 edges (numbered sequentially as (30,454), (30,563), (400,595), (297,592), and (125,452)). The gain results for each edge show a typical diminishing marginal effect: the first two edges contribute the most significantly, with the first edge having the highest gain (dE=0.01073, dC=0.09846, dR=0.01843), effectively strengthening key bridging relationships; the subsequent three edges mainly serve to refine and strengthen the network and optimize redundancy. Since the network connectivity has reached 1.0000 in steady state, adding edges no longer increases connectivity but rather enhances the network's efficiency and robustness under both steady-state and multi-scenario perturbation conditions by shortening the average shortest path and providing additional detour paths.

[0083] like Figure 5 As shown, Figure 5The presentation includes three sets of metrics and visualizations: before edge addition, after edge addition, and after damaged region failure. A bar chart compares the three states: before edge addition, after edge addition, and after damaged region failure. First, the connectivity is 1.0000 both before and after edge addition, indicating the network was fully connected from the start. This explains why "connectivity remained unchanged but there was still a gain" (the gain comes from path shortening rather than merging isolated points into the main network). Network efficiency improved from 0.1351 to 0.1517, a 12.29% increase; robustness also improved from 0.1351 to 0.1517. Under stress testing with damaged region failure, connectivity decreased to 0.9961, and the number of connected subgraphs increased from 1 to 2, meaning a very small number of nodes were cut out of the principal component, but the main network remained generally connected. At this point, efficiency only slightly decreased from 0.1517 to 0.1504 (a decrease of 0.86%), while robustness decreased from 0.1517 to 0.1498 (a decrease of 1.25%). Compared to before the edge was added, even after the damage, efficiency was still about 11.3% higher than the baseline, and robustness remained at a similar level. Combining the distribution of ring=21 in the logs and the penalty for candidates within the region, it can be determined that the newly added edges primarily serve loop bridging and loop closure, providing redundant channels across communities / critical regions for the mainnet. Therefore, under the premise that the steady-state C=1 remains unchanged, it can still significantly improve accessibility and performance retention under disturbances.

[0084] This edge-addition strategy achieves the goals of greater efficiency in steady state and greater robustness under disturbances. Specifically, it significantly shortens the average shortest path and improves resilience without changing the full connectivity level. Performance only slightly declines under damage impact, validating the effectiveness of robust guided edge addition (prioritizing shorter edges, avoiding loops / circling loops) around the damaged region. Further, if it is desired to maintain connectivity at 1 after damage, one or two emergency bridging edges could be added to candidates outside the loop and across potential cut points / bridge edges. However, based on the current chart, the existing five edges already provide a cost-effective improvement.

[0085] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0086] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A network edge optimization method applied to urban power communication networks, characterized by: This method is implemented by the following steps: Step 1: Construct an urban power communication network model; meanwhile, analyze the vulnerability of the urban power network to determine the key areas; Step 2: Define an annulus area according to the key areas determined in Step 1; Step 3: Construct a damage set for multiple fault scenarios and set weights for each scenario; Step 4: Under multiple fault scenarios, comprehensively adopt cross-component bridging, cut-point neighbor closure, community bridging, and annulus closure and around-region bridging strategies to generate a candidate edge set; Step 5: Evaluate the metrics for each candidate edge under each fault scenario and summarize them according to the weights. At the same time, adjust the weights of the candidate edges located in the key areas and calculate the scores of the candidate edges; Step 6: Select the most balanced edge addition strategy according to the Pareto front and maxmin. Select candidate edges until the upper limit is reached or there is no positive gain anymore, and finally output the newly added links and the evaluation metric values.

2. The network edge optimization method applied to urban power communication networks according to claim 1, characterized in that: In step one, the urban power communication network model is as follows: ,in, For a collection of sites, that is , Indicates the first There are 10 sites, and the number of sites is 100. E represents the set of edges corresponding to communication links in the power communication network, i.e. , Representing a power communication network model Middle Station With the site The edge between; in the set Zhongruo =1 indicates that at the site With the site If there is an edge connecting them, =0, which means that at the site With the site There are no edge connections.

3. The network edge optimization method applied to urban power communication networks according to claim 1, characterized in that: In step two, the key region is defined as: D={ | ,in, and For the site latitude and longitude coordinates; Here are the coordinates of the center of the critical area, and r is the damage radius; Set the annulus area as a priority edge addition buffer zone close to the key area but not within the area; Define the annulus area as: R = {(x, y) | r < d((x, y), c) ≤ (1 + γ)r}; where, (x, y) is the position of a certain node in the network model G, c is the center position of the key area, d((x, y), c) represents the geometric distance between the node position and the center of the circle, and γ is the thickness coefficient of the annulus area.

4. The network edge optimization method applied to urban power communication networks according to claim 1, characterized in that: The damage method used is: obtaining a power communication network model. The location information of each station is processed, and the location of each station is traversed to calculate the Euclidean distance between the station and the center of the key area. v, if If a site is found to be within a critical region, then the site is considered to be located within the critical region. Iterate through all sites within the critical region, delete the site and its associated edges, and for each edge, calculate the Euclidean distance between the edge and the center of the critical region. ,if If so, then it is considered that the edge crosses the critical area.

5. The network edge optimization method applied to urban power communication networks according to claim 1, characterized in that: In Step 3, set the weight β of the key area. Perform multiple optimizations and damages on the network model respectively at multiple weight values of β ∈ {1, 2, 3, 4, 5}, and record the overall optimization metrics M(G) after adding edges and after damage. M(G) ∈ {E(G), C(G), R(G)}, where E(G) is the network efficiency, C(G) is the connectivity, and R(G) is the robustness. Summarize the mean and standard error of each metric after damage for each β, and take the β with the largest mean of each metric as the weight value.

6. The network edge optimization method applied to urban power communication networks according to claim 5, characterized in that: In Step 3, the formula for the network efficiency E(G) is as follows: ； in, For node v i With node v j The shortest path length between them, where n is the total number of nodes in the network model G; The formula for the connectivity C(G) is as follows: ； Among them, | Let |V| be the maximum number of connected subgraphs in the network model G, and |V| be the total number of nodes in the current network. The formula for the robustness R(G) is as follows: ； Where S(p) is the percentage of the largest connected subgraph size after removing the first p percentage of nodes; This represents the maximum removal percentage.

7. The network edge optimization method applied to urban power communication networks according to claim 1, characterized in that: The specific process of Step 5 is as follows: Step 51: Add candidate edges e to the network model G, in each fault scenario s i The set of gains for calculating scene metrics on ∈S { , } ; ； In the formula, , These are the network efficiency, connectivity, and robustness gain values ​​of the network model G after adding candidate edge e, respectively. For the initial set of index values ​​of the network, {E( ),C( ),R( )}∈M( ), M( ) for scene s i Based on the network model G, candidate edges are added. The set of network efficiency, connectivity, and robustness metrics for the network model G. Step 5.2: Set of scene metric gains for each scene Calculate the expected gain value; for each scene s i Corresponding to a scene weight vector The formula for the expected gain is as follows: ； in, ;in, , , Adding candidate edges respectively The improvement values ​​of network efficiency, connectivity, and robustness in the scenario metrics of the post-network model G; Step 53: If candidate edges If the geometric position falls within the critical region D, then the expected gain value... Adjustments are made, and the adjustment formula is as follows: ;otherwise, =1; Candidate edge score:

8. The network edge optimization method applied to urban power communication networks according to claim 1, characterized in that: In Step 6, compare the improvement amplitudes of each candidate edge in the scenario metrics of network efficiency, connectivity, and robustness. Finally, select the candidate edge with the most balanced improvement of network metrics as the optimal edge, add the optimal edge to the network model G, and update the network structure; record the optimal edge in the newly added edge sequence; When the number of candidate edges reaches the upper limit h or there are no more positively gainful edges in the candidate set, output the newly added links and the evaluation metric values.

Images