A network base station planning method and system based on correlation-based intelligent resource allocation

Abstract

This invention relates to the field of network base station planning technology, and provides a network base station planning method and system based on correlation-based intelligent resource allocation, comprising: Step S1: constructing an initial population for the network base station planning problem and establishing a one-to-many mapping relationship; Step S2: extracting the changing trend, calculating the correlation index, and dynamically dividing the current population solution set into an exact optimization group, a conditional optimization group, and a probabilistic optimization group; Step S3: executing differentiated dynamic resource allocation and evolutionary strategies based on the solution set partitioning results; Step S4: iteratively executing, dynamically updating the correlation index and solution set partitioning until the termination condition is met, and outputting the optimal two-level decision scheme. This invention can perform network base station planning more effectively.

Description

Technical Field

[0001] This invention relates to the field of network base station planning technology, and more specifically, to a network base station planning method and system based on correlation-based intelligent resource allocation. Background Technology

[0002] Network base station planning is a crucial step in telecommunications infrastructure construction, and the quality of its decisions directly impacts network coverage, service quality, and return on investment. In actual commercial operations, base station planning typically involves two decision-makers with differing interests: the tower company responsible for constructing base station towers and other infrastructure, and the various operators leasing these facilities to provide telecommunications services. The tower company minimizes its net expenditure (construction costs minus rental income) by planning base station locations, while the operators maximize their net revenue by selecting different equipment configurations (such as coverage and capacity combinations) on existing base stations. This process naturally constitutes a two-level optimization problem, with the tower company acting as the upper-level leader and the operators as the lower-level followers.

[0003] The core challenge in solving this type of two-layer planning problem lies in the enormous computational complexity brought about by its inherent nested structure. For each candidate base station construction scheme of the upper-layer leader (the tower company), it is necessary to accurately solve the optimal configuration response problem of multiple lower-layer operators. This process leads to high-frequency and high-cost evaluation of the lower-layer objective function, causing the overall computational overhead to increase exponentially with the problem size (number of candidate base stations, number of operators, types of configuration schemes).

[0004] To address the aforementioned challenges, existing technologies have explored various approaches. For instance, traditional methods like the Karush-Kuhn-Tucker (KKT) conditional transformation method attempt to replace the lower-level problem with its optimality conditions, transforming a two-level problem into a single-level constraint problem. However, this requires the lower-level problem to possess favorable mathematical properties (such as convexity and differentiability), making it difficult to apply to complex, non-convex, and discrete network base station planning scenarios in practice. Another approach is nested evolutionary algorithms, which directly utilize evolutionary algorithms to handle upper and lower-level optimizations separately. While offering strong generality, each generation of evolution requires running a complete lower-level evolutionary optimization process for each upper-level individual, resulting in an extremely heavy computational burden and poor scalability. In recent years, surrogate model-assisted evolutionary algorithms have been proposed, replacing some lower-level precise evaluations by constructing computationally less computationally expensive approximate models. However, training these surrogate models requires a large amount of sample data, and the models lack generalization ability, requiring retraining when the problem scenario changes, thus limiting their application in practical dynamic programming environments.

[0005] In summary, existing network base station planning methods generally suffer from the following technical shortcomings when dealing with two-layer decision-making problems involving tower companies and multiple operators: First, they fail to effectively identify the intrinsic relationship between the upper and lower layer objective functions, applying the same precision to all upper-layer solutions, resulting in a significant waste of computational resources on solutions with no potential or negative correlation. Second, they lack a dynamic adaptive resource allocation mechanism, failing to intelligently adjust the optimization investment in the lower-layer problem based on different states such as cooperation, irrelevance, or conflict between the upper and lower layer objectives during the current search process. Third, existing methods are computationally inefficient, making it difficult to obtain high-quality two-layer planning solutions within an acceptable timeframe, especially when the number of candidate base stations is large and operator requirements are complex, often leading to computational bottlenecks.

[0006] Therefore, there is an urgent need for a network base station planning method that can intelligently perceive the correlation between upper and lower layer targets, dynamically allocate computing resources, and significantly reduce the solution overhead of the lower layer, so as to improve the solution efficiency and quality of the two-layer decision scheme. Summary of the Invention

[0007] The present invention provides a network base station planning method and system based on correlation-based intelligent resource allocation, which can overcome some or all the defects of the prior art.

[0008] According to a network base station planning method based on correlation-based intelligent resource allocation of the present invention, the method includes the following steps: Step S1: Construct the initial population for the network base station planning problem, where the upper-level decision variable is the base station construction location vector X, and the lower-level decision variable is the configuration scheme y of each telecom operator on the existing base stations; establish a one-to-many mapping relationship, that is, one upper-level base station construction scheme corresponds to multiple lower-level operator configuration scheme samples, in order to retain the distribution information of the lower-level solutions; Step S2: During the evolution process, extract the upper-level objective function value of the tower company. F The lower-level objective function values ​​of each operator f The changing trend is used to calculate the correlation index, which measures the strength of the correlation between upper and lower level objectives. Based on the correlation index, the current population solution set is dynamically divided into the exact optimization group, the conditional optimization group, and the probabilistic optimization group. Step S3: Based on the partitioning results of the solution set, execute differentiated dynamic resource allocation and evolutionary strategies to update the individual groups, where: For the precise optimization group, a full-scale search strategy is adopted, using correlation to guide the synchronous evolution of base station construction variables and operator configuration variables. For the condition optimization group, a threshold criterion for changes in base station construction variables is introduced. Full-scale lower-level operator configuration optimization is only initiated when the change in the base station construction plan exceeds the set threshold. For probabilistic optimization groups, probability values ​​are set to determine the optimization of the lower-level problem; A crossover operator based on distribution information is adopted, which uses the lower-level configuration distribution information accumulated in a pair of multiple populations to guide the upper-level base station construction individuals to perform crossover and mutation; at the same time, the lower-level operator configuration scheme also performs corresponding crossover and mutation. Step S4: Repeat steps S2 and S3, dynamically updating the correlation index and solution set partitioning until the termination condition is met, and output the optimal two-level decision scheme.

[0009] Preferably, in step S1, the upper-level decision variable is the base station construction location vector. Indicates whether or not in position j Build base stations; upper objective function F The function to minimize the net expenditure of the tower company is:

[0010] First item The first item is the total construction cost, and the second item is the total rental income that the tower company collects from all operators. It refers to the number of operators. It is the number of candidate base station locations. This refers to the number of optional configuration options for the base station. In position j The construction cost of building a base station Is the operator at the base station? j Select configuration scheme k , It is the rent that the tower company collects from the operators. It is related to the base station j The relevant distance or attenuation factor; The lower-level decision variable is the configuration scheme of each telecom operator on the existing base stations. y Each operator i In each existing base station j Choose a configuration scheme k， Right now ,and ; Lower-level objective function f The function for maximizing the net revenue of each operator:

[0011] Operators The returns are: Coverage Returns Plus 3x capacity benefits Subtract the rent paid ; It was a decision made at the top; if the tower company hadn't built the base stations, If the value is zero, then the term is zero. Establish a one-to-many mapping relationship: each individual group Includes an upper-layer base station construction plan and Sample configuration schemes for different lower-level operators , ,…, This is to preserve the distribution information of the lower-level solutions.

[0012] Preferably, in step S1, the initial population is generated as follows: Set the population size to ; For each individual group : Randomly generate an upper-level solution Each dimension of the variable takes the value 1 with a preset probability, and otherwise takes the value 0; according to A configuration scheme is randomly selected for each operator and each existing base station. k ,get K A number of different lower-level solution samples; Forming individual groups ; Initial population and initialize the set It is used to store the precisely optimized feasible solution.

[0013] Preferably, the specific method for dynamically partitioning the solution set in step S2 is as follows: When the decrease in the net expenditure of the tower company is accompanied by the increase in the net revenue of the operator, it is determined to be positively correlated and classified into the precise optimization group. When the changes of both are not obvious or exhibit nonlinear fluctuations, they are classified into the aforementioned conditional optimization group; When a decrease in the tower company's net expenditure leads to a decrease in the operator's net revenue, it is determined to be a negative correlation and classified into the aforementioned probability optimization group.

[0014] Preferably, the correlation index in step S2 is It is used to measure whether the relationship between upper-level and lower-level targets in the current search area is cooperative, irrelevant, or conflicting.

[0015] As a preferred option, step S3 further refines each individual group based on similarity indicators. The update mechanism is implemented, including the similarity index. The calculation formula is:

[0016] in:

[0017] Indicates from N The number of combinations of selecting 2 elements from n distinct elements; for the selected nth element... i Any two samples in the group data and If variable F Trends and variables f Consistent, then The value is 1; otherwise, it is 0. The similarity index is used to measure the degree of similarity between lower-level solution samples within an individual group. The value is between 0 and 1. The closer the value is to 1, the more significant the similarity; otherwise, the weaker the similarity. Based on two thresholds , ,and Classify and process them accordingly: like If the similarity is high, execute type 1 update: precisely solve for the current upper-level solution. The corresponding lower-level problem yields the optimal lower-level solution. The exact solution will be ( Store the data in set A, clear the current individual group, and randomly generate a new individual group. like The similarity is determined to be average, so execute type 2 update: calculate the maximum upper-level target value of the current individual group. ,like If the solution is correct, the lower-level problem is solved precisely and the exact solution is stored in A. Then, the A is cleared and the entire group of individuals is randomly generated again. Otherwise, the upper-level solution is kept unchanged, and a new lower-level sample is randomly generated again. like If the similarity is low, execute type 3 update: solve the lower-level problem with 50% probability and store the exact solution in A and completely regenerate the individual group; keep the upper-level solution unchanged with 50% probability and regenerate the lower-level sample randomly.

[0018] Preferably, the crossover operator based on distribution information described in step S3 adopts the following method when generating the offspring population: Utilize the current population Generate offspring population from set A : (1) Based on the upper-level objective function F Evaluate each exact solution in set A, and select the optimal group of individuals, denoted as . ; (2) For i=1 to Generate offspring individuals in a cyclical manner: Randomly select from A For each individual, calculate the mean of its upper-level solution vector. M ; from Two different individual groups were randomly selected. and ; Generate offspring:

[0019] in, It is the first i The new individual after the individual is updated It is the first i Individual updates x Quantity, It is the first i The individual j indivual y The updated values ​​of the components It is the best individual in set A. It is a random selection from set A The mean value of an individual. and yes For any two individual groups, the scaling factor , It is the mean individual x Quantity, Individual i of x Quantity, It is a random group of two individuals x Quantity, It is the mean individual y Quantity, Individual i The j indivual y Quantity, It is the first of two random individual groups j indivual y Quantity, It is the number of individual groups. yes The number of dimensional components.

[0020] This invention provides a network base station planning system based on correlation-based intelligent resource allocation, which adopts the aforementioned network base station planning method based on correlation-based intelligent resource allocation.

[0021] The beneficial effects of this invention are as follows: This invention abandons the traditional strategy of indiscriminately solving all lower-level problems, and introduces a correlation index and a dynamic solution set partitioning mechanism. By measuring the cooperative, irrelevant, or conflicting relationships between the objective functions of the upper-level (tower company) and lower-level (operator) in real time, the population solution set is divided into precise optimization group, conditional optimization group, and probabilistic optimization group, and differentiated resource allocation is implemented accordingly. For positively or negatively correlated solution sets, a full-scale search or a crossover operator based on distribution information is used to avoid frequent high-precision solutions to lower-level problems; full lower-level optimization is only initiated for solutions in the conditional optimization group that meet specific threshold criteria. This mechanism significantly reduces the number of evaluations of the lower-level objective function, and reduces the overall computational overhead of the network base station planning problem while ensuring solution accuracy. It is particularly suitable for large-scale planning scenarios with a large number of candidate base stations and complex operator configuration schemes.

[0022] This invention, through a one-to-many population structure, preserves the distribution information of lower-level operator configuration schemes, providing rich data support for correlation analysis. Based on this, for precisely optimized groups exhibiting positive correlations, a correlation-guided synchronous evolution strategy between upper and lower levels is adopted, ensuring that the reduction in the tower company's net expenditure and the increase in the operator's net revenue are realized simultaneously, effectively promoting synergy between upper and lower levels. For probabilistic optimized groups with conflicts, a crossover operator based on distribution information intelligently focuses on the Pareto front region, obtaining a better compromise solution between balancing the tower company's construction costs and the operator's revenue.

[0023] The relevance index and solution set partitioning of this invention are dynamically updated during the evolutionary process, automatically adjusting the optimization strategy according to the changing trend of the target in the current search region. When the search process shifts from global exploration to local refinement, the partitioning results change accordingly, and the corresponding resource allocation strategy also switches adaptively, without requiring the user to preset any rules or thresholds. Furthermore, this invention introduces an intra-group similarity index and its update mechanism, which can intelligently determine the diversity status of lower-level samples and automatically decide whether to clear and regenerate individual groups, thereby maintaining the population's exploratory ability and avoiding premature convergence. This adaptability enables the method of this invention to maintain stable and efficient solution performance when facing base station networks of different sizes, different numbers of operators, and dynamically changing rental and cost parameters, demonstrating good generalization ability and engineering practicality. Attached Figure Description

[0024] Figure 1 This is a flowchart of a network base station planning method based on correlation-based intelligent resource allocation in an embodiment. Detailed Implementation

[0025] To further understand the content of this invention, a detailed description of the invention will be provided in conjunction with the accompanying drawings and embodiments. It should be understood that the embodiments are merely illustrative and not limiting of the invention.

[0026] Example like Figure 1 As shown, this embodiment provides a network base station planning method based on correlation-based intelligent resource allocation, which includes the following steps: Step S1: Construct the initial population for the network base station planning problem, where the upper-level decision variable is the base station construction location vector X, and the lower-level decision variable is the configuration scheme y of each telecom operator on the existing base stations; establish a one-to-many mapping relationship, that is, one upper-level base station construction scheme corresponds to multiple lower-level operator configuration scheme samples, in order to retain the distribution information of the lower-level solutions; Step S2: During the evolution process, extract the upper-level objective function value of the tower company. F The lower-level objective function values ​​of each operator f The changing trend is used to calculate the correlation index, which measures the strength of the correlation between upper and lower level objectives. Based on the correlation index, the current population solution set is dynamically divided into the exact optimization group, the conditional optimization group, and the probabilistic optimization group. Step S3: Based on the partitioning results of the solution set, execute differentiated dynamic resource allocation and evolutionary strategies to update the individual groups, where: For the precise optimization group, a full-scale search strategy is adopted, using correlation to guide the synchronous evolution of base station construction variables and operator configuration variables. For the condition optimization group, a threshold criterion for changes in base station construction variables is introduced. Full-scale lower-level operator configuration optimization is only initiated when the change in the base station construction plan exceeds the set threshold. For probabilistic optimization groups, probability values ​​are set to determine the optimization of the lower-level problem; A crossover operator based on distribution information is adopted, which uses the lower-level configuration distribution information accumulated in a pair of multiple populations to guide the upper-level base station construction individuals to perform crossover and mutation; at the same time, the lower-level operator configuration scheme also performs corresponding crossover and mutation. Step S4: Repeat steps S2 and S3, dynamically updating the correlation index and solution set partitioning until the termination condition is met, and output the optimal two-level decision scheme.

[0027] In step S1, the upper-level decision variable is the base station construction location vector. Indicates whether or not in position j Build base stations; upper objective function F The function to minimize the net expenditure of the tower company is:

[0028] First item The first item is the total construction cost, and the second item is the total rental income that the tower company collects from all operators. It refers to the number of operators. It is the number of candidate base station locations. This refers to the number of optional configuration options for the base station. In position j The construction cost of building a base station Is the operator at the base station? j Select configuration scheme k , It is the rent that the tower company collects from the operators. It is related to the base station j The relevant distance or attenuation factor; in fact, the tower company wants to keep construction costs as low as possible while maximizing rental income, so the overall goal is to minimize net expenditure.

[0029] Constraints:

[0030] Ensure that the proportion of base stations built is not less than , This is the minimum base station density requirement to meet basic coverage and emergency communication needs.

[0031] The lower-level decision variable is the configuration scheme of each telecom operator on the existing base stations. y Each operator i In each existing base station j Choose a configuration scheme k， Right now ,and ; Lower-level objective function f The function for maximizing the net revenue of each operator:

[0032] Operators The returns are: Coverage Returns Plus 3x capacity benefits Subtract the rent paid ; It was a decision made at the top; if the tower company hadn't built the base stations, If the value is zero, then the item is zero; the operator attempts to maximize its net profit.

[0033] Constraints:

[0034] For each operator and each existing base station Only one configuration option can be selected. (That is, multiple configurations cannot be used at the same time).

[0035] Establish a one-to-many mapping relationship: each individual group Includes an upper-layer base station construction plan and Sample configuration schemes for different lower-level operators , ,…, This is to preserve the distribution information of the lower-level solutions.

[0036] In step S1, the initial population is generated as follows: Set the population size to ; For each individual group : Randomly generate an upper-level solution Each dimension of the variable takes the value 1 with a preset probability, and otherwise takes the value 0; according to A configuration scheme is randomly selected for each operator and each existing base station. k ,get K A number of different lower-level solution samples; Forming individual groups ; Initial population and initialize the set It is used to store the precisely optimized feasible solution.

[0037] The specific method for dynamically partitioning the solution set in step S2 is as follows: When the decrease in the net expenditure of the tower company is accompanied by the increase in the net revenue of the operator, it is determined to be positively correlated and classified into the precise optimization group. When the changes of both are not obvious or exhibit nonlinear fluctuations, they are classified into the aforementioned conditional optimization group; When a decrease in the tower company's net expenditure leads to a decrease in the operator's net revenue, it is determined to be a negative correlation and classified into the aforementioned probability optimization group.

[0038] The correlation index mentioned in step S2 is It is used to measure whether the relationship between upper-level and lower-level targets in the current search area is cooperative, irrelevant, or conflicting.

[0039] In step S3, each individual group is further analyzed based on similarity indicators. The update mechanism is implemented, including the similarity index. The calculation formula is:

[0040] in:

[0041] Indicates from N The number of combinations of selecting 2 elements from n distinct elements; for the selected nth element... i Any two samples in the group data and If variable F Trends and variables f Consistent, then The value is 1; otherwise, it is 0. The similarity index is used to measure the degree of similarity between lower-level solution samples within an individual group. The value is between 0 and 1. The closer the value is to 1, the more significant the similarity; otherwise, the weaker the similarity. Based on two thresholds , ,and Classify and process them accordingly: like If the similarity is high, execute type 1 update: precisely solve for the current upper-level solution. The corresponding lower-level problem yields the optimal lower-level solution. The exact solution will be ( Store the data in set A, clear the current individual group, and randomly generate a new individual group. like The similarity is determined to be average, so execute type 2 update: calculate the maximum upper-level target value of the current individual group. ,like If the solution is correct, the lower-level problem is solved precisely and the exact solution is stored in A. Then, the A is cleared and the entire group of individuals is randomly generated again. Otherwise, the upper-level solution is kept unchanged, and a new lower-level sample is randomly generated again. like If the similarity is low, execute type 3 update: solve the lower-level problem with 50% probability and store the exact solution in A and completely regenerate the individual group; keep the upper-level solution unchanged with 50% probability and regenerate the lower-level sample randomly.

[0042] The crossover operator based on distribution information described in step S3 specifically employs the following method when generating the offspring population: Utilize the current population Generate offspring population from set A : (1) Based on the upper-level objective function F Evaluate each exact solution in set A, and select the optimal group of individuals, denoted as . ; (2) For i=1 to Generate offspring individuals in a cyclical manner: Randomly select from A For each individual, calculate the mean of its upper-level solution vector. M ; from Two different individual groups were randomly selected. and ; Generate offspring:

[0043] in, It is the first i The new individual after the individual is updated It is the first i Individual updates x Quantity, It is the first i The individual j indivual y The updated values ​​of the components It is the best individual in set A. It is a random selection from set A The mean value of an individual. and yes For any two individual groups, the scaling factor , It is the mean individual x Quantity, Individual i of x Quantity, It is a random group of two individuals x Quantity, It is the mean individual y Quantity, Individual i The j indivual y Quantity, It is the first of two random individual groups j indivual y Quantity, It is the number of individual groups. yes The number of dimensional components.

[0044] In step S4, the termination condition is reaching the maximum number of iterations or satisfying the convergence accuracy; the final solution output is: after the loop ends, evaluate all exact solutions in set A, and select the solution with the smallest upper-level objective function F as the optimal base station construction and resource allocation scheme.

[0045] This embodiment provides a network base station planning system based on correlation-based intelligent resource allocation, which adopts the above-mentioned network base station planning method based on correlation-based intelligent resource allocation.

[0046] This embodiment introduces a one-to-many population structure to effectively explore the correlation between the goals of leaders and followers. The decision variables are then divided into one of three categories—exact optimization, conditional optimization, or probabilistic optimization—depending on the measured strength of the correlation. This classification allows for a differential evolution strategy, applying different levels of precision to each group. This avoids seeking exhaustive, high-precision solutions for each follower's problem, thus saving significant computational resources.

[0047] The present invention and its embodiments have been described above illustratively. This description is not restrictive, and the figures shown are only one embodiment of the present invention; the actual structure is not limited thereto. Therefore, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the present invention, such designs should fall within the protection scope of the present invention.

Claims

1. A network base station planning method based on correlation-based intelligent resource allocation, characterized in that, Includes the following steps: Step S1: Construct the initial population for the network base station planning problem, where the upper-level decision variable is the base station construction location vector X, and the lower-level decision variable is the configuration scheme y of each telecom operator on the existing base stations; establish a one-to-many mapping relationship, that is, one upper-level base station construction scheme corresponds to multiple lower-level operator configuration scheme samples, in order to retain the distribution information of the lower-level solutions; Step S2: During the evolution process, extract the upper-level objective function value of the tower company. F The lower-level objective function values ​​of each operator f The changing trend is used to calculate the correlation index, which measures the strength of the correlation between upper and lower level objectives. Based on the correlation index, the current population solution set is dynamically divided into the exact optimization group, the conditional optimization group, and the probabilistic optimization group. Step S3: Based on the partitioning results of the solution set, execute differentiated dynamic resource allocation and evolutionary strategies to update the individual groups, where: For the precise optimization group, a full-scale search strategy is adopted, using correlation to guide the synchronous evolution of base station construction variables and operator configuration variables. For the condition optimization group, a threshold criterion for changes in base station construction variables is introduced. Full-scale lower-level operator configuration optimization is only initiated when the change in the base station construction plan exceeds the set threshold. For probabilistic optimization groups, probability values ​​are set to determine the optimization of the lower-level problem; A crossover operator based on distribution information is adopted, which uses the lower-level configuration distribution information accumulated in a pair of multiple populations to guide the upper-level base station construction individuals to perform crossover and mutation; at the same time, the lower-level operator configuration scheme also performs corresponding crossover and mutation. Step S4: Repeat steps S2 and S3, dynamically updating the correlation index and solution set partitioning until the termination condition is met, and output the optimal two-level decision scheme.

2. The network base station planning method based on correlation-based intelligent resource allocation according to claim 1, characterized in that, In step S1, the upper-level decision variable is the base station construction location vector. , indicating whether in position j Build base stations; upper-level objective function F The function to minimize the net expenditure of the tower company is: ； First item The first item is the total construction cost, and the second item is the total rental income that the tower company collects from all operators. It refers to the number of operators. It is the number of candidate base station locations. This refers to the number of optional configuration options for the base station. In position j The construction cost of building a base station Is the operator at the base station? j Select configuration scheme k , It is the rent that the tower company collects from the operators. It is related to the base station j The relevant distance or attenuation factor; The lower-level decision variable is the configuration scheme of each telecom operator on the existing base stations. y Each operator i In each existing base station j Choose a configuration scheme k， Right now ,and ; Lower-level objective function f The function for maximizing the net revenue of each operator: ； Operators The returns are: Coverage Returns Plus 3x capacity benefits Subtract the rent paid ; It was a decision made at the top; if the tower company hadn't built the base stations, If the value is zero, then the term is zero. Establish a one-to-many mapping relationship: each individual group Includes an upper-layer base station construction plan and Sample configuration schemes for different lower-level operators , ,…, This is to preserve the distribution information of the lower-level solutions.

3. The network base station planning method based on correlation-based intelligent resource allocation according to claim 2, characterized in that, In step S1, the initial population is generated as follows: Set the population size to ; For each individual group : Randomly generate an upper-level solution Each dimension of the variable takes the value 1 with a preset probability, and otherwise takes the value 0; according to A configuration scheme is randomly selected for each operator and each existing base station. k ,get K A number of different lower-level solution samples; Forming individual groups ; Initial population and initialize the set It is used to store the precisely optimized feasible solution.

4. The network base station planning method based on correlation-based intelligent resource allocation according to claim 3, characterized in that, The specific method for dynamically partitioning the solution set in step S2 is as follows: When the decrease in the net expenditure of the tower company is accompanied by the increase in the net revenue of the operator, it is determined to be positively correlated and classified into the precise optimization group. When the changes of both are not obvious or exhibit nonlinear fluctuations, they are classified into the aforementioned conditional optimization group; When a decrease in the tower company's net expenditure leads to a decrease in the operator's net revenue, it is determined to be a negative correlation and classified into the aforementioned probability optimization group.

5. A network base station planning method based on correlation-based intelligent resource allocation according to claim 4, characterized in that, The correlation index mentioned in step S2 is It is used to measure whether the relationship between upper-level and lower-level targets in the current search area is cooperative, irrelevant, or conflicting.

6. A network base station planning method based on correlation-based intelligent resource allocation according to claim 5, characterized in that, In step S3, each individual group is further analyzed based on similarity indicators. The update mechanism is implemented, including the similarity index. The calculation formula is: ； ； Indicates from N The number of combinations of selecting 2 elements from n distinct elements; for the selected nth element... i Any two samples in the group data and If variable F Trends and variables f Consistent, then The value is 1; otherwise, it is 0. The similarity index is used to measure the degree of similarity between lower-level solution samples within an individual group. The value is between 0 and 1. The closer the value is to 1, the more significant the similarity. Conversely, it indicates that the similarity is weaker; Based on two thresholds , ,and Classify and process them accordingly: like If the similarity is high, execute type 1 update: precisely solve for the current upper-level solution. The corresponding lower-level problem yields the optimal lower-level solution. The exact solution will be ( Store the data in set A, clear the current individual group, and randomly generate a new individual group. like The similarity is determined to be average, so execute type 2 update: calculate the maximum upper-level target value of the current individual group. ,like If the solution is correct, the lower-level problem is solved precisely and the exact solution is stored in A. Then, the A is cleared and the entire group of individuals is randomly generated again. Otherwise, the upper-level solution is kept unchanged, and a new lower-level sample is randomly generated again. like If the similarity is low, execute type 3 update: solve the lower-level problem with 50% probability and store the exact solution in A and completely regenerate the individual group; keep the upper-level solution unchanged with 50% probability and regenerate the lower-level sample randomly.

7. A network base station planning method based on correlation-based intelligent resource allocation according to claim 6, characterized in that, The crossover operator based on distribution information described in step S3 specifically employs the following method when generating the offspring population: Utilize the current population Generate offspring population from set A : (1) Based on the upper-level objective function F Evaluate each exact solution in set A, and select the optimal group of individuals, denoted as . ; (2) For i=1 to Generate offspring individuals in a cyclical manner: Randomly select from A For each individual, calculate the mean of its upper-level solution vector. M ; from Two different individual groups were randomly selected. and ; Generate offspring: ； in, It is the first i The new individual after the individual is updated It is the first i Individual updates x Quantity, It is the first i The individual j indivual y The updated values ​​of the components It is the best individual in set A. It is a random selection from set A The mean value of an individual. and yes For any two individual groups, the scaling factor , It is the mean individual x Quantity, Individual i of x Quantity, It is a random group of two individuals x Quantity, It is the mean individual y Quantity, Individual i The j indivual y Quantity, It is the first of two random individual groups j indivual y Quantity, It is the number of individual groups. yes The number of dimensional components.

8. A network base station planning system based on correlation-based intelligent resource allocation, characterized in that, It employs a network base station planning method based on correlation-based intelligent resource allocation as described in any one of claims 1-7.

Images