A base station site planning method and system based on multiple model optimization
Through optimization methods based on multiple models, combined with genetic algorithms and clustering algorithms, the problems of low accuracy and high cost of base station selection in the existing technology are solved, and accurate planning and efficient coverage of base station site selection are achieved.
Patent Information
- Application Number
- CN202411035212.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-31
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-07-31
AI Technical Summary
The existing single-model base station site selection method has shortcomings in terms of accuracy and multi-scenario applications, and it is impossible to effectively consider the base station coverage range and weak coverage point distribution characteristics, resulting in inaccurate site selection planning and increasing construction costs.
The optimization method based on multiple models is adopted, and the planned area is divided into different regions according to the base station coverage type and weak coverage point distribution characteristics, and the base station is used using the 0-1 planning model, the K-means clustering algorithm and the DBSCAN clustering model respectively. Site selection planning is optimized with genetic algorithms to form a comprehensive optimal base station site selection scheme.
It significantly improves the site selection accuracy of base stations, adapts to the base station planning needs of different scenarios, reduces construction costs, improves network coverage and service quality, and achieves accurate coverage of weak coverage points.
Smart Images

Figure CN118590899B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of base station site selection and planning, and specifically to a base station site planning method and system based on multiple model optimization. Background Art
[0002] With the popularization of broadband and the large-scale use of mobile phones, families or individuals have a stronger demand for the Internet, and the requirements for Internet speed are getting higher and higher. Accordingly, more and more base stations are built to meet the needs of high business volume and high Internet speed. In real life, due to the different coverage ranges of base stations and the different business volumes between weak coverage points, base stations face many problems in the site selection process. For example, the coverage range of base stations is circular or sector-shaped. Different coverage ranges will lead to different site selection plans for base stations, and the current site selection is based on covering as many areas as possible with most weak coverage points without considering the total business volume. In fact, the business volume of many weak coverage points is so small that these weak coverage points can be abandoned to reduce construction costs. In addition, due to the characteristics between weak coverage points, such as transitivity, or being relatively scattered and not transitive, it will also affect the site selection of base stations.
[0003] At present, the base station site selection method based on a single model is still the mainstream of base station planning, such as using a 0-1 model without site screening, or a K-means model, but without a comprehensive planning method based on the coverage characteristics of base stations in the region or the distribution characteristics of weak coverage points. Therefore, a base station site planning method based on multiple model optimization is needed to solve the above technical problems. Summary of the invention
[0004] The purpose of this application is to provide a base station site planning method and system based on multiple model optimization to solve the technical problems that the existing single model base station site selection accuracy is not high and cannot realize multi-scenario application.
[0005] The purpose of this application can be achieved through the following technical solutions:
[0006] According to a first aspect of the present application, a base station site planning method based on multiple model optimization is provided, comprising:
[0007] Step S1: Divide the planning area into A, B, and C according to the base station coverage type and the weak coverage point distribution characteristics;
[0008] Step S2: If the base station coverage is circular, it is area A. Data preprocessing is performed on weak coverage points and candidate base stations in area A, and weak coverage points with a traffic volume of less than 10 are eliminated. The candidate base stations are preliminarily screened using existing base stations and threshold conditions. The coordinates of the newly built base station are used as decision variables, the minimum total cost is used as the objective function, and the distance between base stations is greater than the threshold, the distance between the weak coverage point and the macro base station is less than 30, the distance between the weak coverage point and the micro base station is less than 10, and 90% of the total traffic volume of the weak coverage point is covered as constraints. A 0-1 planning model is established, and a genetic algorithm is used to solve the 0-1 planning model, and finally the site coordinates of the macro base station and the micro base station are obtained respectively;
[0009] Step S3: If the coverage range of the base station is sector-shaped and there is no transitivity between the weak coverage points, it is zone B. The weak coverage points within the planned base station range are cloned according to their traffic volume for clustering. The cloned weak coverage points are clustered using the K-means algorithm, and the number of cluster centers k=3 is set. The direction of the obtained cluster center is used as the main direction of each base station sector. The angle of the main direction of the sector is solved by polar coordinate conversion, and the sectors with the main direction angle of adjacent sectors less than 45° are refined and clustered, so as to obtain the optimal sector angle of each base station. Finally, the total traffic volume of the coverage point is calculated. When the total traffic volume meets a certain value, the optimal site coordinates and sector angle configuration are output;
[0010] Step S4: If the base station coverage is sector-shaped and the weak coverage points are transitive, it is zone C. A DBSCAN clustering model is established to cluster the weak coverage points. The neighborhood radius is set to 20 and the minimum number of neighbors is set to 1. All weak coverage points are clustered to form clusters. The coordinates of the center points of each cluster are output for base station site planning.
[0011] Step S5: Comprehensively analyze the base station site selection plans in the three areas A, B, and C to obtain the optimal base station site selection plan for the entire area.
[0012] The planning and site selection scheme is to directly adopt the planning and site selection schemes of areas A, B, and C to form a comprehensive and optimal base station site selection planning scheme.
[0013] Ideally, the coverage of a base station is usually circular, but in reality, some base stations will present a sector-shaped coverage. If the site selection calculation is still performed according to the circular coverage, it will inevitably lead to an increase in weak coverage points, and some areas cannot be covered. At the same time, under the sector-shaped coverage condition, the distribution of weak coverage points will show scattered or concentrated characteristics. When the neighborhood radius between weak coverage points does not exceed a value, such as 20, it is considered that there is transitivity between the weak coverage points. If the neighborhood radius between the weak coverage points exceeds the set neighborhood radius, it is considered that there is no transitivity between the weak coverage points and they are relatively scattered. According to the transitivity characteristics of the weak coverage points, two models are used to cluster them respectively, which can more accurately perform site selection planning calculations and meet the base station site selection planning in different scenarios. Therefore, when performing base station site selection planning for a certain area, it can be partitioned according to the base station coverage type and weak coverage point characteristics of the area, and the base station planning of each area can be calculated separately, and finally the base station planning of the entire area is obtained to form the optimal base station site selection planning scheme.
[0014] Furthermore, let the set of candidate base stations be T, and the selection of each candidate base station i be v i ,in :
[0015] (2)
[0016] Thus, the candidate base station state matrix is formed:
[0017] (11)
[0018] V is the base station state matrix, n is the number of sites, Indicates the selection status of each site, for example:
[0019] (12)
[0020] Indicates that the first site is selected and the rest are not selected.
[0021] The base station parameters contain the coordinates and coverage of the candidate base stations, and are encoded in the form of a real two-dimensional matrix, where each row is the parameter of a base station, and the number of matrix rows is the number of base stations, forming a base station parameter matrix, as shown below:
[0022] (13)
[0023] in, is the base station data row vector of candidate base station i, including the x-coordinate, y-coordinate and the coverage range of the base station.
[0024] Furthermore, the fitness function of the genetic algorithm is:
[0025] (17)
[0026] in For the case where candidate base station i is used as a new base station, For the case where the newly built base station i is a macro base station, In the case where the newly built base station i is a micro base station, the cost of establishing a macro base station is 10, and the cost of establishing a micro base station is 1.
[0027] Furthermore, the genetic algorithm uses arithmetic crossover, and the specific formula is:
[0028] (20)
[0029] (twenty one)
[0030] in and represents 2 parent chromosomes, and represents the daughter chromosomes produced by crossover, is a random number between 0 and 1, 0 <i<N。
[0031] Furthermore, in region B, the coverage varies with the deflection angle The change is as follows linearly:
[0032] (twenty two)
[0033] R is the radiation radius of the main direction of the sector attenuated to the weak coverage point with the deflection angle, R0 is the coverage radius of the main direction of the sector, is the deflection angle (in degrees) relative to the main direction, which deflects from the main direction to both sides. The maximum value of the angle of the base station signal deviating from the main direction is ,Right now .
[0034] Furthermore, the specific steps of clustering using the K-means algorithm are:
[0035] A1: The number of clusters k=3, three center points are randomly selected, and the distance from each sample point to the center point is calculated;
[0036] B1: Assign each sample to the nearest center point;
[0037] C1: Update the center point according to the clustering results;
[0038] D1: Repeat B1 and reallocate each sample point to the nearest center point;
[0039] E1: Repeat C1 and update the center point;
[0040] F1: Repeat D1 to determine whether the clustering result is consistent with the previous result. If it is consistent, the algorithm terminates, otherwise it returns to E1 and continues to iterate.
[0041] Furthermore, let the total traffic volume of the area covered by the newly built base station i be , the total business volume that the new base station can cover is:
[0042] (twenty three)
[0043] in, The coverage relationship existence parameter between the newly built base station i and the weak coverage point h, It is the traffic volume of a weak coverage point within the coverage area of the base station.
[0044] In a newly built base station i, for any weak coverage point h that it can cover within the circular area, the calculation formula for the deflection angle relative to the main direction is:
[0045] (twenty four)
[0046] In the above formula, There are three main directions of sectors, is the deflection angle of the line connecting the weak coverage point h and the newly built base station i. Substituting the relative main direction deflection angle into formula (25) can obtain the radiation range of the newly built base station in the direction of the weak coverage point:
[0047] (25)
[0048] R ih is the radiation radius of the main direction of the sector i of the newly built base station attenuated to the weak coverage point h with the deflection angle, R0 is the coverage radius of the sector in the main direction, is the deflection angle of the weak coverage point h within the coverage range of the newly built base station i relative to the main direction of the sector, then the coverage relationship existence parameter between the newly built base station i and the weak coverage point h The value of is:
[0049] (26)
[0050] in, is the coverage relationship existence parameter between the newly built base station i and the weak coverage point h, is the Euclidean distance between the newly built base station i and the weak coverage point h. The value of can be used to determine whether the newly built base station i can cover the weak coverage point h. Substituting it back into formula (23), the size and proportion of the total coverage business volume can be obtained through further calculation.
[0051] Furthermore, the steps of clustering operation using DBSCAN model are as follows:
[0052] A2: Randomly select a weak coverage point;
[0053] B2: With the weak coverage point as the center, draw a circle with a neighborhood radius of Eps=20. When the number of sample points in the circle is greater than or equal to the minimum number of neighbors (MinPts), take the point as the core point and cluster the sample points in the circle into one category.
[0054] C2: Continue to use the other sample points in the circle with the core point as the center and perform the B2 operation, and finally connect all the core points to form a cluster and group them into one category;
[0055] D2: Repeat the above three steps and finally cluster all weak coverage points.
[0056] According to a second aspect of the present application, a base station site planning system based on multiple model optimization is provided, including:
[0057] The regional classification unit divides the planning area into A, B, and C according to the base station coverage type and the weak coverage point distribution characteristics, and sends the corresponding data to the corresponding regional planning calculation unit;
[0058] The planning calculation unit of area A performs data preprocessing on weak coverage points and candidate base stations in area A, removes weak coverage points with a traffic volume of less than 10, uses existing base stations and threshold conditions to preliminarily screen candidate base stations, takes the coordinates of the newly built base station as the decision variable, takes the minimum total cost as the objective function, takes the distance between base stations greater than the threshold, the distance between the weak coverage point and the macro base station less than 30, the distance between the weak coverage point and the micro base station less than 10, and the total traffic volume of the weak coverage point is 90% covered as constraints, establishes a 0-1 planning model, uses a genetic algorithm to solve the 0-1 planning model, and finally obtains the site coordinates of the macro base station and the micro base station respectively;
[0059] The B-area planning calculation unit clones the weak coverage points within the planned base station range according to their traffic volume for clustering, clusters the cloned weak coverage points using the K-means algorithm, sets the number of cluster centers k=3, takes the obtained cluster center direction as the main direction of each base station sector, solves the sector main direction angle through polar coordinate conversion, and refines the clustering of sectors with main direction angles of adjacent sectors less than 45°, thereby obtaining the optimal sector angle of each base station. Finally, the total traffic volume of the coverage points is calculated. When the total traffic volume meets a certain value, the optimal site coordinates and sector angle configuration are output;
[0060] In the C-area planning calculation unit, a DBSCAN clustering model is established to cluster the weak coverage points. The neighborhood radius is set to 20 and the minimum number of neighbors is set to 1. All weak coverage points are clustered to form clusters, and the coordinates of the center points of each cluster are output for base station site selection planning.
[0061] The comprehensive planning unit receives the base station site planning results of the planning calculation unit in area A, the planning calculation unit in area B, and the planning calculation unit in area C, and comprehensively processes the base station site planning in the three areas to obtain the optimal base station site planning solution for the entire area.
[0062] The planning and site selection scheme is to directly adopt the planning and site selection schemes of areas A, B, and C to form a comprehensive and optimal base station site selection planning scheme.
[0063] Compared with the prior art, the beneficial effects of this application are:
[0064] (1) This application uses a genetic algorithm for site screening under the 0-1 model. The genetic algorithm has significant advantages in base station planning, mainly in the following aspects: First, it can efficiently search and optimize the base station location by simulating the natural evolution process, and find a better base station location distribution; second, the genetic algorithm significantly improves the efficiency of the algorithm by using coding and operations, such as efficient selection and optimization through the 0-1 planning model; in addition, it can adapt to complex constraints and integrate constraints into the objective function to meet various planning needs; through crossover and mutation operations, the genetic algorithm strikes a balance between global search and local optimization to improve the planning effect; the optimization of the fitness function helps to gradually approach the optimal layout. In practical applications, the genetic algorithm has a significant effect on solving base station planning problems. For example, in this study, through genetic algorithm optimization, the final result is to establish 2144 macro base stations and 850 micro base stations, and the coverage business volume reaches 90.98% of the total business volume. In summary, the genetic algorithm plays an important role in the complex base station location selection problem and significantly improves network coverage and service quality.
[0065] (2) This application uses the K-means clustering algorithm to clone the weak coverage points. The advantage of this operation is that the clustering effect not only considers the spatial distribution of the weak coverage points, but also takes the attribute of the traffic volume of the weak coverage points into consideration;
[0066] (3) This application uses the DBSCAN algorithm as the main calculation idea. The time complexity of this algorithm is lower than that of common clustering algorithms such as K-means, and the clustering accuracy of this algorithm is relatively high. At the same time, the DBSCAN algorithm's characteristic of classification based on spatial density makes it have a certain degree of transitivity, and the parameter setting enables this algorithm to successfully solve the spatial limitation requirements of this application;
[0067] (4) The area is divided into zones according to the base station coverage type and weak coverage point characteristics, and the base station planning of each area is calculated separately. Finally, the base station planning of the entire area is obtained, which can significantly improve the accuracy of base station planning in the entire area and reduce costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0068] To facilitate understanding by those skilled in the art, the present application is further described below in conjunction with the accompanying drawings.
[0069] Figure 1 This is a map of the types of base station construction areas for this application;
[0070] Figure 2 Establish a 0-1 model planning base station flow chart for this application;
[0071] Figure 3 This is a 0-1 integer programming diagram for this application;
[0072] Figure 4 Genetic algorithm flow chart for this application;
[0073] Figure 5 Create a new base station visualization for this application;
[0074] Figure 6 This is a schematic diagram of the base station coverage area for this application;
[0075] Figure 7 This is a schematic diagram of the K-means algorithm process for this application;
[0076] Figure 8 This is a schematic diagram of clustering of weak coverage points within the coverage area of the base station of this application;
[0077] Fig. 9 This is a schematic diagram of the DBSCAN clustering algorithm for this application;
[0078] Fig.10 A visualization of the weak coverage points of this application. DETAILED DESCRIPTION
[0079] The technical solution of the present application will be described clearly and completely in conjunction with the embodiments below. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application.
[0080] According to an embodiment of the present application, there is a base station site planning method based on multiple model optimization.
[0081] First, the planning area is divided according to the base station coverage type and the distribution characteristics of weak coverage points, and recorded as A, B, and C respectively. If the base station coverage is circular, it is area A; if the base station coverage is sector-shaped and there is no transitivity between weak coverage points, it is area B; if the base station coverage is sector-shaped and there is transitivity between weak coverage points, it is area C. Figure 1 shown.
[0082] (I) When the base station coverage is circular, such as Figure 2 As shown, for area A, a 0-1 model can be used to plan the site. For each candidate base station, two decisions need to be made: building a site or not. If building a site, two situations need to be chosen: building a macro base station or building a micro base station. Therefore, a 0-1 planning model is established, with the candidate base station as the decision variable, the minimum cost as the objective function, and the threshold constraint, total traffic constraint, and coverage condition constraint as the constraint conditions. The model can be solved using a genetic algorithm to obtain several sets of optimal site planning solutions. The specific steps are as follows:
[0083] (a) Preprocessing the data of weak coverage points
[0084] In the process of selecting a base station site, two issues need to be considered: 1) How to select a site address so that the base station covers 90% of the total traffic volume of the weak coverage point; 2) How to plan the site selection of the two base stations to minimize the cost. The site selection of the base station does not necessarily cover all weak coverage points as much as possible, but to cover as much traffic as possible. Therefore, in the site selection process, priority should be given to covering some weak coverage points with large traffic volume. Under the condition that the new station covers 90% of the total traffic volume of the weak coverage points, the weak coverage points with very small traffic volume can be left uncovered. That is, in the data of all weak coverage points in Table 1, the traffic volume of each weak coverage point is first preprocessed. In Table 1, the sum of column C (traffic volume data) is obtained, and the total traffic volume of all weak coverage points is 7056230.115, and the at least traffic volume to be covered is 6350607.
[0085] By sorting column C, we find that the highest business volume of the weak coverage point is close to 50,000, while the business volume of some weak coverage points is even less than 1, and there are quite a few such weak coverage points.
[0086] Table 1 Service data of weak coverage points
[0087]
[0088] According to the above analysis, we delete the points whose business volume is less than or equal to i. After the deletion, the total business volume is j, and the number of weak coverage points remaining after the deletion is k.
[0089] Table 2 Deletion of coverage points with relatively small traffic volume
[0090]
[0091] According to Table 2, we found that when deleting all weak coverage points below 10, the total business volume is only reduced by 300,000. It is obvious that the impact of deleting coverage points is very small compared to the total business volume of 7 million, and the remaining total business volume after deletion is always higher than 90%. Therefore, deleting all weak coverage points below 10 in Table 2 has little impact on site planning. In this application, we delete all weak coverage points with a business volume less than 10 and obtain the remaining 61,395 weak coverage points. The error of the final business volume analysis calculation before and after data deletion is 0.05, and the confidence level is 95%.
[0092] (b) Data processing of candidate base stations
[0093] The distance between the new base station site and the existing site cannot be less than or equal to a given threshold. Assume that the given threshold is d=10. Therefore, for the existing base station The Euclidean distance between the newly built base station P and the new base station P shall not be less than or equal to 10, that is: (1)
[0094] For new base stations, if a new base station is built within the range of all grid points in the 2500×2500 area, the threshold restriction must be considered: the distance between the new base station and the existing base station cannot be less than or equal to 10. Based on the data of existing sites in Table 3, the area range of the new base station is defined with this constraint, and some data is cleaned. In formula (1), is the location coordinate of the existing base station.
[0095] Table 3 Coordinates of existing base station sites
[0096]
[0097] After data processing, 380799 coordinates that did not meet the constraints were cleaned up, and the deleted grid points within 2500×2500 were marked to form a scatter image.
[0098] A total of 5869201 candidate coordinates that can be used as new base points are obtained, and each candidate coordinate point is numbered sequentially. The numbering is shown in Table 4:
[0099] Table 4 Preprocessed candidate base station data
[0100]
[0101] (c) Establishment of 0-1 planning model
[0102] In order to solve the problem of new station site selection, this application establishes a 0-1 integer programming model:
[0103] like Figure 3As shown, the coordinates of the points in the white block area in the above scatter image are There are three ways to build a base station: not building a base station here, building a macro base station, and building a micro base station. The three options correspond to three different values. The decision that needs to be made is to find a suitable site for the new base station.
[0104] Decision variables: coordinates of the new base station .
[0105] As long as a base station is built, a certain cost must be incurred. The cost of a macro base station is 10, and the cost of a micro base station is 1. Whether to build a base station depends on the total business volume that can be covered after the base station is built, and the type of base station to be built depends on the distance between the base station and the weak coverage point and the cost of building the base station.
[0106] Based on this, this application introduces 0,1 variables , build a site to cover more business volume, and plan whether to build a site in this area:
[0107] (2)
[0108] Introduce two 0-1 variables , plan whether to choose macro base stations or micro base stations so that 90% of the total traffic at weak coverage points is covered by the planned base stations; (3)
[0109] Indicates that a macro base station is being built. It means that a micro base station is being established.
[0110] The criterion for measuring whether a decision is optimal is to minimize the cost of building a website, so the minimum cost of building a website is used as the objective function. The objective function is:
[0111] (4)
[0112] Constraints:
[0113] (5)
[0114] (6)
[0115] (7)
[0116] (8)
[0117] (9)
[0118] Among the above constraints, formula (5) indicates that for a candidate base station i selected as a new base station, it must choose between the macro base station and the micro base station. Formula (6) indicates that the distance between two different new base station sites should be greater than 10. Formula (7) needs to satisfy , that is, if the weak coverage point s is covered by the newly built macro base station i, then the Euclidean distance between the weak coverage point and the macro base station is less than or equal to 30. Formula (8) needs to satisfy , that is, if the weak coverage point s can be covered by the newly built micro base station i, then the Euclidean distance between the weak coverage point and the micro base station is less than or equal to 10. Formula (9) is the total traffic volume of the weak coverage points that can be covered by the newly built base station, which must be greater than or equal to 6350607. H is the set of weak coverage points that can be covered by the newly built base station, is the traffic volume of the hth weak coverage point .
[0119] Genetic Algorithm for Solving Mathematical Programming Models
[0120] like Figure 4 The genetic algorithm process shown in the figure mainly considers five basic elements in the design of the genetic algorithm: (1) the encoding method, that is, the expression method of chromosomes; (2) the method of generating the initial population; (3) the determination of the fitness function; (4) the determination of appropriate genetic operators, mainly including selection, crossover and mutation operators; (5) the setting of algorithm parameters, mainly including population size, evolutionary generations, crossover probability, mutation probability, etc.
[0121] The solution process of the genetic algorithm is:
[0122] 1) Encoding method
[0123] All operations of the genetic algorithm act on the code. A good code can greatly improve the efficiency of the algorithm. For the specific site problem, each candidate base station has only two situations: selected and not selected. Therefore, according to the above 0-1 planning model, let the candidate base station set be T, and the selection of each candidate base station i is as follows, where :
[0124] (10)
[0125] Thus, the candidate base station state matrix is formed:
[0126] (11)
[0127] V is the base station state matrix, n is the number of sites, Indicates the selection status of each site, for example:
[0128] (12)
[0129] Indicates that the first site is selected and the rest are not selected.
[0130] The parameters of the base station include the coordinates and coverage of the candidate base stations. This application uses the encoding form of a real two-dimensional matrix. Each row is the parameter of a base station, and the number of matrix rows is the number of base stations, forming a base station parameter matrix, as shown below:
[0131] (13)
[0132] in, is the base station data row vector of candidate base station i, including the x-coordinate, y-coordinate and the coverage range of the base station.
[0133] For example, a parameter matrix is:
[0134] (14)
[0135] So, The base station coordinates are (355, 123), the base station coverage is 30, and it is a macro base station. Multiply the candidate base station state formula (11) with the base station parameter matrix formula (14) to obtain the antibody encoding matrix D:
[0136] (15)
[0137] in After the operation, the remaining elements in matrix D are are the x-coordinate, y-coordinate and coverage range of the selected base station.
[0138] (16)
[0139] in is the x coordinate of the selected base station, is the y coordinate of the selected base station, is the coverage area of the selected base station.
[0140] 2) Initialization of population and individual selection method
[0141] The genetic algorithm requires an initial population at the beginning. This application selects several candidate base points from the candidate base points as the initial population, that is, randomly selects the initial population. The length of the chromosome, that is, the optimal number of base stations, cannot be known. There are many methods for selecting individuals. This application adopts the most commonly used roulette selection method, and in order to ensure the continuity of excellent individuals in the population, a gene pool is established, and all offspring and their parents generated by each crossover mutation are placed in it, and then sorted according to the fitness value. Assuming that the population size is NP, the first NP chromosomes are selected as the parents of the next iteration.
[0142] 3) Determination of fitness function and handling of constraints
[0143] The fitness function is generally designed based on the objective function. In order to reduce the cost of building new base stations, a cost objective function for site selection is established, which is the fitness function of the genetic algorithm:
[0144] (17)
[0145] in For the case where candidate base station i is used as a new base station, For the case where the newly built base station i is a macro base station, Indicates the case where the newly built base station i is a micro base station. The cost of establishing a macro base station is 10, and the cost of establishing a micro base station is 1.
[0146] According to the characteristics of the constraints, a penalty term is constructed and then added to the objective function to transform it into an unconstrained problem.
[0147] When the coordinates fall into the infeasible region that does not meet the constraints, a penalty function is added Keep it away from the optimal solution.
[0148] The specific form of the penalty function is as follows: Assume that the objective function is f(D), where D is the decision variable vector configured by the base station, and the penalty function It is used to integrate the constraints into the objective function and construct the overall objective function g(D) as follows:
[0149] g(D)=f(D)+P(x) (18)
[0150] Among them, P(x) is the penalty term. When the solution x violates the constraint, P(x) will impose a penalty to move away from the optimal solution. In base station planning, considering the constraints such as the minimum distance between base stations and coverage range, the penalty term can be expressed as:
[0151] (19)
[0152] in, is the distance between base stations i and j. is the minimum distance between base stations. is the coverage range of base station i. is the maximum coverage allowed. and is the penalty coefficient, which is used to control the influence of different constraints.
[0153] By adding the penalty function P(x) to the objective function f(D), the genetic algorithm can take into account both the optimization goal and the constraint conditions during the search process, thereby obtaining a more reasonable and effective base station planning solution.
[0154] 4) Crossover and mutation
[0155] Crossover determines the global search capability of the genetic algorithm. It operates on two chromosomes at the same time, combining the characteristics of the two to produce new offspring. The performance of the genetic algorithm depends largely on the performance of the crossover operation used. Whether the chromosomes of the parents cross is controlled by the crossover rate. Since the chromosomes are real number coded, in order to increase the degree of population dispersion, this application adopts the method of arithmetic crossover. The specific formula is as follows:
[0156] (20)
[0157] (twenty one)
[0158] in and represents 2 parent chromosomes, and represents the daughter chromosomes produced by crossover, is a random number between 0 and 1, 0 <i<N。
[0159] Mutation is a random change in chromosome genes, which determines the local search capability of the genetic algorithm. Mutation can provide genes that the population does not contain, which is controlled by the mutation rate. The mutation rate determines the percentage of the number of mutated genes in the kinship factor. Conventional mutation methods include basic position mutation and uniform mutation. This application adopts basic position mutation. After the chromosome mutation rate PM is judged, if it is less than PM, a gene is randomly selected for re-assignment, and the assignment must be randomly selected within the planning area.
[0160] The final result obtained by solving all the above algorithms and models is: 2144 macro base stations and 850 micro base stations were established, covering 6420087 business volumes, accounting for 90.98% of the total business volume. The site coordinates of each new base station are shown in Table 5. The site selection of some base stations is shown in Table 5:
[0161] Table 5 Site selection of some base stations
[0162]
[0163] like Figure 5 As shown, by visualizing the newly built base stations in the area and their coverage areas, a distribution image of the newly built base stations can be obtained.
[0164] (ii) When the base station coverage is sector-shaped and there is no transitivity between weak coverage points, it is recorded as area B. Figure 6 As shown, the signal radiation intensity of each station also changes with the change of spatial distance. For each base station O, its coverage of the surrounding space signals is mainly in three main directions. Near each main direction, the signal coverage intensity is attenuated and the actual coverage distance radius is reduced. Figure 6 OA is a main radiation coverage direction of base station O, considering the direction deviating to the left The signal strength changes when the angle changes When , the length that can be covered is reduced to 1 / 2 of the original length. Figure 6 ,like And O is a macro base station, then OA=30, OB=15, as The change of coverage length is roughly in the shape of curve ACDB, and the lengths of OA, OB, OC, and OD, that is, the signal coverage range, decrease at a linear rate. It is easy to conclude that the coverage range changes with the deflection angle. The change is as follows:
[0165] (twenty two)
[0166] R is the radiation radius of the main direction of the sector attenuated to the weak coverage point with the deflection angle, R0 is the coverage radius of the main direction of the sector, is the deflection angle (in degrees) relative to the main direction, which deflects from the main direction to both sides. In this case, the maximum angle of the base station signal deviating from the main direction is ,Right now .
[0167] Based on the actual conditions, the coverage area of the base station is no longer a perfect circle, but has three main directions and the coverage area is divided into Figure 6 The situation shown scales down linearly.
[0168] Furthermore, the selection of the main direction cannot only point to the direction where the weak coverage points are densely distributed, but also needs to consider the difference in business volume at different weak coverage points, and it is necessary to give priority to covering the weak coverage points with large business volume.
[0169] According to the above analysis, the fastest way to find areas with densely distributed weak coverage points is to cluster the weak coverage points. If the difference in business volume between coverage points is to be considered, a single point with a large business volume needs to be cloned into multiple points. The direction of the cluster center point obtained in this way can obviously cover more business volume, and the newly built base station is more likely to cover 90% of the business volume of the weak coverage points.
[0170] At this time, K-means is used for clustering. 1) The specific steps are:
[0171] Figure 7 This is a schematic diagram of the K-means algorithm process. First, the distance between each sample point and each center point is calculated, and the sample point is assigned to the category represented by the nearest center point. When one iteration is completed, the center point of each category is updated according to the clustering result, and then the previous operation is repeated again until there is no difference between the two classification results.
[0172] A1: The number of clusters k=3, randomly select 3 center points, and calculate the distance from each sample point to the center point.
[0173] B1: Assign each sample to the nearest center point.
[0174] C1: Update the center point according to the clustering results.
[0175] D1: Repeat B1 and reallocate each sample point to the nearest center point.
[0176] E1: Repeat C1 and update the center point.
[0177] F1: Repeat D1 to determine whether the clustering result is consistent with the previous result. If it is consistent, the algorithm terminates, otherwise it returns to E1 and continues to iterate.
[0178] 2) Clustering of weak coverage points to determine the main direction of the sector:
[0179] With the help of MATLAB programming language, the coordinate data of the newly built base stations solved by the 0-1 model are imported, and the K-means algorithm is used to perform cluster analysis on the data for the weak coverage points covered by each newly built base station. Since the newly built base station actually covers the signal from three main directions, the number of clusters k is set to 3.
[0180] Take a macro base station as an example for cluster analysis. When clustering, pay attention to the following three aspects:
[0181] (1) The points in each cluster should be relatively concentrated, so that the sector can cover as many weak coverage points as possible, thereby increasing the base station traffic.
[0182] (2) The number of clusters needs to be set in advance. Since each base station has only three main coverage directions, the k value is 3.
[0183] (3) The newly generated points in the cloning process should also be considered within the clustering range, so that weak coverage points with large business volume can be given priority.
[0184] The weak coverage points within the coverage area of the macro base station are clustered, and the clustering results are visualized with the help of MATLAB software.
[0185] Figure 8(a) is a clustering diagram of weak coverage points within the coverage range of the base station, where the asterisk represents the macro base station, three different colored points represent three types of cluster points, the number of clusters is 3, the SC silhouette coefficient is 0.90681, and the cluster center points of the three types of cluster points are also marked in the figure. The circular area is the coverage range of the macro base station under ideal conditions.
[0186] After cluster analysis, select each cluster Figure 3 The line connecting the center point and the base station is used as the initial main direction, and sectors are established in a linear manner from the main direction to both sides according to the relationship shown in formula (22). The decision plan with an angle between the main directions less than 45 degrees is further refined into clusters to adjust the main direction angle. This division of clusters and main directions can ensure that the number of coverage points and the volume of business are maximized as much as possible.
[0187] 3) Sector angle results
[0188] The results of the site construction plan in Area A are adopted. For the sector division of each base station point, the east direction is 0 degrees, and the counterclockwise direction is the positive direction, which represents the main direction angle of the three sectors. The detailed sector angle results are shown in Table 6. Table 6 shows the sector angles of some micro base stations:
[0189] Table 6 Sector angles of some micro base stations
[0190]
[0191] Figure 8 (b) is a schematic diagram of the main direction of a sector of a macro base station:
[0192] 4) Calculation of maximum coverage volume
[0193] Suppose the total traffic volume of the area covered by the newly built base station i is , can be calculated by summing up the traffic volume of weak coverage points in the sector area. Then the total traffic volume that the new base station can cover is:
[0194] (twenty three)
[0195] in, is the existence parameter of the coverage relationship between the newly built base station i and the weak coverage point h, is the traffic volume of a weak coverage point within the coverage area of the base station, and H is the total coverage points within the coverage area.
[0196] In a newly built base station i, for any weak coverage point h that it can cover within the circular area, the calculation formula for the deflection angle relative to the main direction is:
[0197] (twenty four)
[0198] In the above formula, There are three main directions of sectors, is the deflection angle of the line connecting the weak coverage point h and the newly built base station i. Substituting the relative main direction deflection angle into formula (25), the radiation range of the newly built base station in the direction of the weak coverage point can be obtained:
[0199] (25)
[0200] is the radiation radius of the main direction of the sector i of the newly built base station attenuated to the weak coverage point h with the deflection angle, R0 is the coverage radius of the sector in the main direction, is the deflection angle of the weak coverage point h within the coverage range of the newly built base station i relative to the main direction of the sector, then the coverage relationship existence parameter between the newly built base station i and the weak coverage point h The value of is:
[0201] (26)
[0202] in, is the coverage relationship existence parameter between the newly built base station i and the weak coverage point h, is the Euclidean distance between the newly built base station i and the weak coverage point h. The value of can be used to determine whether the newly built base station i can cover the weak coverage point h. Substituting it back into formula (23), the size and proportion of the total coverage business volume can be obtained through further calculation.
[0203] It should be noted that there are overlapping circular coverage areas between newly built base stations. The same weak coverage point may be covered by multiple base stations, but it is not necessarily within the sector coverage of multiple base stations. Therefore, in the solution process, it is necessary to Weak coverage points are moved directly into the covered sequence , the weak coverage points in the sequence will not be covered by other base stations.
[0204] Based on the above analysis, the newly built base station data and weak coverage point data in area A were imported, and the K-means clustering method was used to solve the optimal sector direction data. It was calculated that the newly built base station site can cover up to 69.27% of the total business volume of all weak coverage points under the optimal sector conditions.
[0205] (III) When the base station coverage is sector-shaped and the weak coverage points are transitive, it is recorded as region C and the DBSCAN algorithm is used for regional clustering.
[0206] Compared with traditional clustering methods, the DBSCAN clustering algorithm can effectively process clusters of arbitrary shapes and sizes, and can automatically identify and exclude noise points.
[0207] The K-means algorithm needs to set the number of clusters in advance. The starting center point generated according to the number of clusters will also greatly affect the clustering results. In this article, it is impossible to determine how many categories the weak coverage points are divided into. Therefore, the K-means clustering algorithm cannot achieve accurate clustering of weak coverage points. In contrast, the DBSCAN algorithm uses the results of clustering the data itself to display the number of clusters, which is suitable for clustering problems of weak coverage points with a large number of data points.
[0208] The DBSCAN clustering algorithm requires two parameters to be determined
[0209] (1) Neighborhood radius (EPS): EPS is the radius of the circle around a sample point. If the distance between any two weak coverage points in the clustered weak coverage point set is less than or equal to 20, the EPS value is 20.
[0210] (2) Minimum number of neighbors (MinPts): MinPts is the minimum number of sample points in the circular neighborhood of a sample point (including the sample point). Because the weak coverage points need to be compared with each other, there may be a sample point whose distance from the other sample points is greater than 20. In this case, the point is regarded as a separate cluster. Therefore, the value of MinPts is 1.
[0211] The specific steps of the algorithm are:
[0212] A2: Randomly select a weak coverage point.
[0213] B2: Draw a circle with the weak coverage point as the center according to Eps = 20. When the number of sample points in the circle is greater than or equal to MinPts, take the point as the core point and cluster the sample points in the circle into one category.
[0214] C2: Continue to use the other sample points in the circle with the core point as the center and perform the B2 operation. Finally, all the core points are connected to form a cluster and grouped into one category.
[0215] D2: Repeat the above three steps and finally cluster all weak coverage points.
[0216] like Fig. 9 As shown in the figure, a core point is first selected, and a circle with a radius of 20 is drawn with the core point as the center, and the points in the circle are clustered into one category. Then the points in the circle except the center are taken as core points, and a circle with a radius of 20 is drawn with the core point as the center, and further clustering is performed until all points are clustered. All points in the circle can be clustered into one category, and the points outside the circle are outliers.
[0217] DBSCAN algorithm solves weak coverage point clustering
[0218] Based on the DBSCAN clustering algorithm analyzed above, the MATLAB programming language was used to import the data of 182,807 weak coverage points in Table 1 and perform clustering analysis on the weak coverage points.
[0219] The DBSCAN algorithm is used to cluster the data and all weak coverage points are clustered into 898 categories. The total time complexity of the method used in the present invention is .
[0220] Table 7 Examples of clustering of weak coverage points
[0221]
[0222] Table 7 shows the clustering of some weak coverage points.
[0223] like Fig.10 As shown in the figure, by visualizing different types of weak coverage points with different colors, the approximate clustering of various sample points can be observed more intuitively. Due to the limited number of colors in MATLAB, the same color may be used between different clusters. It is worth noting that in a large area of the same color, there will be individual dots of another color; in a circle with a radius of 20 and a center of this point, there are no other weak coverage points except itself, so it forms a cluster alone; therefore, this phenomenon is a normal clustering situation that meets the clustering requirements.
[0224] The above three methods are used to plan the base station sites in areas A, B, and C in the region respectively, and then integrated into the entire region to form the optimal base station site planning scheme, that is, S=S A +S B +S C , S is the optimal base station site planning scheme, S A is the site planning scheme for base stations in area A, S B is the site planning scheme for base stations in area B, S C This is the site planning scheme for base stations in area C. There can be multiple A, B or C in one area, or only one or two of them. Generally speaking, this method takes into account the coverage characteristics of each base station and the characteristics of weak coverage points, so different methods are used to plan the base station sites, and finally the optimal base station site planning scheme is formed.
[0225] The above are only preferred specific implementation methods of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can make equivalent substitutions or changes within the technical scope disclosed in the present application according to the technical solution and application concept of the present application, which should be covered by the protection scope of the present application.
Claims
1. A base station site planning method based on multiple model optimization, characterized in that: include: Step S1: Divide the planning area into A, B, and C according to the base station coverage type and the weak coverage point distribution characteristics; Step S2: If the base station coverage is circular, it is area A. Data preprocessing is performed on weak coverage points and candidate base stations in area A, and weak coverage points with a traffic volume of less than 10 are eliminated. The candidate base stations are preliminarily screened using existing base stations and threshold conditions. The coordinates of the newly built base station are used as decision variables, the minimum total cost is used as the objective function, and the distance between base stations is greater than the threshold, the distance between the weak coverage point and the macro base station is less than 30, the distance between the weak coverage point and the micro base station is less than 10, and 90% of the total traffic volume of the weak coverage point is covered as constraints. A 0-1 planning model is established, and a genetic algorithm is used to solve the 0-1 planning model, and finally the site coordinates of the macro base station and the micro base station are obtained respectively; Step S3: If the coverage range of the base station is sector-shaped and there is no transitivity between the weak coverage points, it is zone B. The weak coverage points within the planned base station range are cloned according to their traffic volume for clustering. The cloned weak coverage points are clustered using the K-means algorithm, and the number of cluster centers k=3 is set. The direction of the obtained cluster center is used as the main direction of each base station sector. The angle of the main direction of the sector is solved by polar coordinate conversion, and the sectors with the main direction angle of adjacent sectors less than 45° are refined and clustered, so as to obtain the optimal sector angle of each base station. Finally, the total traffic volume of the coverage point is calculated. When the total traffic volume meets a certain value, the optimal site coordinates and sector angle configuration are output; Among them, a new base station is built i The total business volume in the covered area is , the total business volume that the new base station can cover is: (23) In the formula, For new base stations i With weak coverage points h Override relation existence parameters, For new base stations i Weak coverage points within the coverage area h The business volume, H is the total coverage points within the coverage area; Building a new base station i For any weak coverage point that can be covered in the circular area h , the calculation formula of the deflection angle relative to the main direction is: (24) In the above formula, There are three main directions of sectors, Weak Cover Point h With new base station i The deflection angle of the connecting line, substituting the relative main direction deflection angle into formula (25) can obtain the radiation range of the newly built base station in the direction of the weak coverage point: (25) For new base stations i The main direction of the sector decays to a weak coverage point with the deflection angle h The radiation radius of the location, R0 is the coverage radius of the sector in the main direction, For new base stations i Weak coverage points within the coverage area h The deflection angle relative to the main direction of the sector, then the new base station i With weak coverage points h Override relationship existence parameters The value of is: (26) in, For new base stations i With weak coverage points h Override relation existence parameters, For new base stations i With weak coverage points h The Euclidean distance between Step S4: If the base station coverage is sector-shaped and the weak coverage points are transitive, it is zone C. A DBSCAN clustering model is established to cluster the weak coverage points. The neighborhood radius is set to 20 and the minimum number of neighbors is set to 1. All weak coverage points are clustered to form clusters. The coordinates of the center points of each cluster are output for base station site planning. Step S5: Comprehensively analyze the base station site selection plans in the three areas A, B, and C to obtain the optimal base station site selection plan for the entire area.
2. A method for base station site planning based on multiple model optimization according to claim 1, characterized in that: Assume that the candidate base station set is T , each candidate base station i The selection is ,in : (2) Thus, the candidate base station state matrix is formed: (11) V is the base station state matrix, n is the number of sites, Indicates the selection status of each site; The base station parameters contain the coordinates and coverage of the candidate base stations, and are encoded in the form of a real two-dimensional matrix, where each row is the parameter of a base station, and the number of matrix rows is the number of base stations, forming a base station parameter matrix, as shown below: (13) in, Candidate base station i The base station data row vector includes x coordinate, y Coordinates and coverage of the base station.
3. The method for base station site planning based on multiple model optimization according to claim 2, characterized in that: Genetic algorithm fitness function: (17) in, Candidate base station i As for the case of a newly built base station, In the case where the newly built base station is a macro base station, In the case where the newly built base station is a micro base station, the cost of establishing a macro base station is 10, and the cost of establishing a micro base station is 1.
4. The method for base station site planning based on multiple model optimization according to claim 3, characterized in that: The genetic algorithm adopts arithmetic crossover, and the specific formula is: (20) (21) in, represents 2 parent chromosomes, and represents the daughter chromosomes produced by crossover, is a random number between 0 and 1, 0 <i<N。 5. The method for base station site planning based on multiple model optimization according to claim 1, characterized in that: In zone B, the coverage varies with the deflection angle. The change is as follows: (22) R is the radiation radius of the main direction of the sector that decays to the weak coverage point as the deflection angle changes, and R0 is the coverage radius of the main direction of the sector. It is the deflection angle relative to the main direction, deflecting to both sides from the main direction.
6. The method for base station site planning based on multiple model optimization according to claim 5, characterized in that: The specific steps of clustering using the K-means algorithm are: A1: Number of clusters k= 3. Randomly select 3 center points and calculate the distance from each sample point to the center point; B1: Assign each sample to the nearest center point; C1: Update the center point according to the clustering results; D1: Repeat B1 and reallocate each sample point to the nearest center point; E1: Repeat C1 and update the center point; F1: Repeat D1 to determine whether the clustering result is consistent with the previous result. If it is consistent, the algorithm terminates; otherwise, return to E1 and continue iterating.
7. The method for base station site planning based on multiple model optimization according to claim 1, characterized in that: The steps for clustering using the DBSCAN model are: A2: Randomly select a weak coverage point; B2: With the weak coverage point as the center, draw a circle with a neighborhood radius of Eps=20. When the number of sample points in the circle is greater than or equal to the minimum number of neighbors, take the point as the core point and cluster the sample points in the circle into one category. C2: Continue to use the other sample points in the circle with the core point as the center and perform the B2 operation, and finally connect all the core points to form a cluster and group them into one category; D2: Repeat the above three steps and finally cluster all weak coverage points.
8. A system corresponding to the base station site planning method based on multiple model optimization according to any one of claims 1 to 7, characterized in that: include: The regional classification unit divides the planning area into A, B, and C according to the base station coverage type and the weak coverage point distribution characteristics, and sends the corresponding data to the corresponding area planning calculation unit; The planning calculation unit of area A performs data preprocessing on weak coverage points and candidate base stations in area A, removes weak coverage points with a traffic volume of less than 10, uses existing base stations and threshold conditions to preliminarily screen candidate base stations, takes the coordinates of the newly built base station as the decision variable, takes the minimum total cost as the objective function, takes the distance between base stations greater than the threshold, the distance between the weak coverage point and the macro base station less than 30, the distance between the weak coverage point and the micro base station less than 10, and the total traffic volume of the weak coverage point is 90% covered as constraints, establishes a 0-1 planning model, uses a genetic algorithm to solve the 0-1 planning model, and finally obtains the site coordinates of the macro base station and the micro base station respectively; The B-area planning calculation unit clones the weak coverage points within the planned base station range according to their traffic volume for clustering, clusters the cloned weak coverage points using the K-means algorithm, sets the number of cluster centers k=3, takes the obtained cluster center direction as the main direction of each base station sector, solves the sector main direction angle through polar coordinate conversion, and refines the clustering of sectors with main direction angles of adjacent sectors less than 45°, thereby obtaining the optimal sector angle of each base station. Finally, the total traffic volume of the coverage points is calculated. When the total traffic volume meets a certain value, the optimal site coordinates and sector angle configuration are output; In the C area planning calculation unit, a DBSCAN clustering model is established to cluster the weak coverage points. The neighborhood radius is set to 20, the minimum number of neighbors is set to 1, all the weak coverage points are clustered to form clusters, and the coordinates of the center points of each cluster are output for base station site selection planning; The comprehensive planning unit receives the base station site planning results of the planning calculation unit in area A, the planning calculation unit in area B, and the planning calculation unit in area C, and comprehensively processes the base station site planning in the three areas to obtain the optimal base station site planning solution for the entire area.
Citation Information
Patent Citations
Base station deployment method and base station deployment device
CN109429234A
Cost optimization unmanned aerial vehicle base station deployment method based on an improved genetic algorithm
CN109819453A
