5G site planning and regional clustering method based on continuous clustering algorithm

Abstract

The invention relates to a 5G site planning and regional clustering method based on a continuous clustering algorithm, and the method comprises the following steps: 1, employing a 5G site agile planning method of the continuous clustering algorithm based on grid density, rapidly recognizing a weak coverage region, and giving a reasonable station adding suggestion and an initial work parameter planning suggestion through the continuous clustering algorithm; 2, analyzing the geometric distance relation between the weak coverage points and the base stations, finding out the optimal construction base station group of the weak coverage points under the judgment of the construction types of the base stations, and carrying out the reasonable clustering of the weak coverage point group; step 3, performing iterative operation on each parameter of the 5G base station planning site by adopting a particle swarm iterative algorithm, selecting key factors such as base station coverage area, base station traffic and construction cost as sub-objective functions, and finally determining a total objective function of the 5G optimal site; the method has the advantages that the continuous clustering algorithm based on the grid density is adopted, the site planning accuracy is improved, and the weak coverage point groups are reasonably clustered.

Description

Technical Field

[0001] The present invention belongs to the technical field of 5G planning, and specifically relates to a 5G site planning and regional clustering method based on a continuous clustering algorithm. Background Art

[0002] With the rapid development of 5G technology, the demand for base station construction is increasing. The planning of mobile communication networks, especially the site selection, is becoming more and more difficult. Therefore, the layout planning of base station sites is particularly important. Although 5G can share a large number of existing 4G sites, due to the increase in 5G-NR transmission power, the application of beamforming, 3D-MIMO antenna technology and anti-interference technology, the NR coverage performance is completely different from LTE. In addition, the new 5G business requires increased site deployment density. Therefore, 5G cannot be built completely at the existing site 1:1. Some sites still need to be shared and some new sites need to be built. The current 5G site planning method It is mainly based on manual planning or simple density clustering algorithm output. Although the manual planning method has high accuracy, it requires a lot of human resources and relies heavily on the experience and knowledge of planning and design personnel. It is time-consuming and cumbersome. Therefore, the manual planning method can only be applied to site planning in a small range (filling points) and cannot be applied to site planning in a large range or the entire network of 5G. Therefore, it is very necessary to provide a 5G site planning and regional clustering method based on a continuous clustering algorithm based on grid density, which improves the accuracy of site planning and reasonably clusters weak coverage point groups. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art and to provide a 5G site planning and regional clustering method based on a continuous clustering algorithm that adopts a grid density-based continuous clustering algorithm to improve the accuracy of site planning and reasonably cluster weak coverage point groups.

[0004] The object of the present invention is achieved by: a 5G site planning and regional clustering method based on a continuous clustering algorithm, the method comprising the following steps:

[0005] Step 1: Use the 5G site agile planning method based on the continuous clustering algorithm of grid density to quickly identify the weak coverage area, and give reasonable site addition suggestions and initial engineering parameter planning suggestions through the continuous clustering algorithm;

[0006] Step 2: Analyze the geometric distance relationship between the weak coverage points and the base stations, and find the optimal base station group for the weak coverage points based on the type of base station construction, and reasonably cluster the weak coverage point groups;

[0007] Step 3: Use the particle swarm iterative algorithm to iteratively calculate the parameters of the 5G base station site planning, select the base station coverage area, base station traffic volume and construction cost key factors as sub-objective functions, and finally determine the overall objective function of the optimal 5G site.

[0008] Step 1 specifically includes the following steps:

[0009] Step 1.1: Pre-processing of weak coverage raster data;

[0010] Step 1.2: DBSCAN density clustering;

[0011] Step 1.3: K-Means clustering;

[0012] Step 1.4: Output of planning site;

[0013] Step 1.5: Initial engineering parameter planning of the community.

[0014] The weak coverage raster data pre-processing in step 1.1 specifically includes the following steps:

[0015] Step 1.11: Filter the grid data whose RSRP_0 is less than the weak coverage threshold n;

[0016] Step 1.12: Filter the grid data whose distance from the weak coverage grid to the master cell (Antenna_0) is greater than the weak coverage distance threshold d;

[0017] Step 1.13: After collaborative filtering, the RSRP_0 at different heights of the same grid is averaged.

[0018] The DBSCAN density clustering in step 1.2 specifically includes the following steps:

[0019] Step 1.21: Input data source set (mainly involving 5G site initialization data) is denoted as D;

[0020] Step 1.22: Mark all objects in set D as unvisited;

[0021] Step 1.23: Randomly select an unvisited object in set D as D;

[0022] Step 1.24: Preset the parameter eps to the specified circle radius and parameter MinPts. If there are at least MinPts objects in the neighborhood of eps with p as the center, a new cluster C is created;

[0023] Step 1.25: Let N be the set of objects in the eps neighborhood of p. Repeat step (4) for each point in N. All objects that meet the conditions form the same cluster C′.

[0024] Step 1.26: If there are still objects that do not belong to any cluster after the above steps, they are marked as noise points, and the density clustering results with parameters eps and MinPts as constraints are finally output.

[0025] The K-Means clustering in step 1.3 specifically includes the following steps:

[0026] Step 1.31: Input the data source, data set D and the preset number of clusters K;

[0027] Step 1.32: Randomly select K objects in the data set D as the initial clustering centers;

[0028] Step 1.33: Calculate the distance between each object and each seed cluster center, and assign each object to the cluster center closest to it; the cluster centers and the objects assigned to them are a cluster group;

[0029] Step 1.34: Each time a sample is assigned, the cluster center is recalculated based on the existing objects in the cluster. This process will be repeated until no object is reassigned to a different cluster or no cluster center changes;

[0030] Step 1.35: According to the above steps, the second K-Means clustering result can be finally obtained.

[0031] The base station coverage area sub-objective function in step 3 is specifically: the base station coverage range can be expressed as the target value f1: Where λ is the blind area loss adjustment coefficient; S is the total area of ​​the base station optimization area; S i Represents the coverage area of ​​a base station; C ij represents the overlapping coverage area between base stations; m represents the number of co-sited stations; and n represents the number of new-sited stations.

[0032] The base station traffic quantum objective function in step 3 is specifically: the variable is inversely proportional to the target value, and the traffic capacity is placed in the denominator, so the objective function of the traffic capacity is corrected to: In the formula, α represents the correction value; A Mi represents the original average daily total traffic of LTE at the co-sited station; A Ni Represents the average daily total flow of stations around the new site.

[0033] The construction cost sub-objective function in step 3 is specifically: The objective function is: In the formula, C Mi C represents the cost of building a co-location site; Ni Represents the construction cost of the new site.

[0034] The overall objective function of the 5G optimal site in step 3 is specifically as follows: the three sub-objective functions are normalized by using the hierarchical analysis method, and finally the three sub-objective function variables are converted into values ​​under the same metric; finally, the overall optimization objective function is: F = ω1f1′+ω2f2′+ω3f3′, where ω i Represents the weight of each sub-objective function; f i ' represents the normalized value.

[0035] Beneficial effects of the present invention: The present invention is a 5G site planning and regional clustering method based on a continuous clustering algorithm. In use, the present invention uses a clustering algorithm combined with map information to quickly identify weak coverage areas, and uses a continuous clustering algorithm to give reasonable suggestions for adding sites and initial parameter planning suggestions, thereby making up for the shortcomings of low efficiency of existing manual planning and poor accuracy of single clustering; the present invention analyzes the geometric distance relationship between weak coverage points and base stations, and under the judgment of the type of base station construction, finds the optimal construction base station group that makes the total business volume of the weak points reach 90%, and reasonably clusters the weak coverage point group; the present invention uses a particle swarm iterative optimization algorithm to iteratively calculate the parameters of the 5G base station planning site, selects three key factors of base station coverage area, base station traffic volume and site construction cost as sub-objective functions, and finally determines the total objective function of the 5G optimal site; the present invention has the advantages of using a continuous clustering algorithm based on grid density, improving the accuracy of site planning, and reasonably clustering the weak coverage point group. . BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a general 5G site planning method and flow chart of the present invention.

[0037] Figure 2 This is an improved method and flow chart for 5G site planning of the present invention.

[0038] Figure 3 This is a schematic diagram of site optimization based on density clustering of the present invention.

[0039] Figure 4 This is a schematic diagram of the initial cell engineering parameter planning based on K-Means of the present invention.

[0040] Figure 5 The clustering algorithm flow chart of the present invention. DETAILED DESCRIPTION

[0041] The present invention will be further described below in conjunction with the accompanying drawings.

[0042] Example 1

[0043] like Figure 1-5 As shown, a 5G site planning and regional clustering method based on a continuous clustering algorithm comprises the following steps:

[0044] Step 1: Use the 5G site agile planning method based on the continuous clustering algorithm of grid density to quickly identify the weak coverage area, and give reasonable site addition suggestions and initial engineering parameter planning suggestions through the continuous clustering algorithm;

[0045] ① Pre-processing of weak coverage grid data: The weak coverage grid processing mainly performs collaborative filtering on the weak coverage grids in the 4G / 5G joint simulation or 5G real results by presetting the weak coverage threshold n and the distance threshold d (grid points closer to the base station should be given priority for network optimization rather than adding base station coverage), and further provides the basic input data source for the subsequent clustering algorithm, such as Figure 2 The specific processing flow is as follows:

[0046] (1) Filter the grid data whose RSRP_0 is less than the weak coverage threshold n;

[0047] (2) Filter the grid data whose distance from the weak coverage grid to the master cell (Antenna_0) is greater than the weak coverage distance threshold d;

[0048] (3) After collaborative filtering, the RSRP_0 at different heights of the same grid is averaged.

[0049] ②DBSCAN density clustering: Select the grid after weak coverage filtering, and use the first DBSCAN density clustering method to cluster multiple adjacent coverage grids into one category based on the grid's longitude and latitude, such as Figure 3 As shown; the density clustering optimization process is as follows:

[0050] (1) The input data source set (mainly involving 5G site initialization data) is denoted as D;

[0051] (2) Mark all objects in set D as unvisited;

[0052] (3) Randomly select an unvisited object in set D as D;

[0053] (4) The preset parameter eps is the specified circle radius and the parameter MinPts. If there are at least MinPts objects in the neighborhood of eps with p as the center of the circle, a new cluster C is created;

[0054] (5) Let N be the set of objects in the eps neighborhood of p. Repeat step (4) for each point in N. All objects that meet the conditions form the same cluster C′.

[0055] (6) If there are still objects that do not belong to any cluster after the above steps, they are marked as noise points, and the final output is the density clustering result with parameters eps and MinPts as constraints.

[0056] ③K-Means clustering: Statistics are collected for the results of the first DBSCAN density clustering. If the first clustering range is too large to allow a base station to be fully covered, K-means is used to perform a second clustering of the larger clusters. This can yield relatively small weak coverage clusters that enable a single base station to be fully covered. The K-Means clustering process is as follows:

[0057] (1) Input the data source, data set D, and the preset number of clusters K;

[0058] (2) Randomly select K objects in the data set D as the initial cluster centers;

[0059] (3) Calculate the distance between each object and each seed cluster center, and assign each object to the cluster center closest to it; the cluster centers and the objects assigned to them form a cluster group;

[0060] (4) Each time a sample is assigned, the cluster center is recalculated based on the existing objects in the cluster, and this process is repeated until no (or a minimum number of) objects are reassigned to different clusters or no cluster center changes;

[0061] (5) According to the above steps, the second K-Means clustering result can be finally obtained.

[0062] ④ Site planning output: Comprehensively process the results of the two clusterings and output a list of required site requirements, which mainly includes the following methods and steps:

[0063] (1) According to the results of the two clusterings, the categories with particularly small number of weak coverage grids are merged; if there are no nearby weak coverage grids that can be merged, the category is directly deleted;

[0064] (2) Output the center point of the weak coverage area;

[0065] (3) Obtain the type of terrain and building height of the center point based on map information (GIS) to determine whether the weak coverage center point is available (excluding the influence of water areas and super-high buildings). If the current center point is not available, find a nearby available grid as the weak coverage center point;

[0066] (4) Calculate the distance between each type of weak coverage center point and surrounding sites. If the weak coverage center point is very close to surrounding base stations, the center point is invalid.

[0067] (5) Save the center point and output the site planning list.

[0068] ⑤ Initial cell engineering parameter planning: The initial cell engineering parameter planning mainly uses the K-Means clustering algorithm with a K value of 3 to cluster the weak coverage areas covered by a single base station into three categories, and obtain the center points of the three categories of areas. The normal direction of the cell antenna is aligned with the weak coverage center area, and the initial cell engineering parameters are calculated based on the location and altitude of the base station and the center point. Compared with the traditional manual setting of engineering parameters or the 4G inheritance method, it can solve the coverage problem more effectively and accurately. Figure 4 shown.

[0069] Step 2: Analyze the geometric distance relationship between the weak coverage points and the base stations, and find the optimal base station group for the weak coverage points based on the type of base station construction, and reasonably cluster the weak coverage point groups;

[0070] Step 3: Use the particle swarm iterative algorithm to iteratively calculate the parameters of the 5G base station site planning, select the base station coverage area, base station traffic volume and construction cost key factors as sub-objective functions, and finally determine the overall objective function of the optimal 5G site.

[0071] The present invention is a 5G site planning and regional clustering method based on a continuous clustering algorithm. During use, the continuous clustering algorithm based on grid density of the present invention uses the 5G simulation result as the data source input, pre-processes the weak coverage grid, outputs preliminary effective grid data after the first DBSCAN density clustering, and again quotes the K-Means clustering algorithm to output the 5G planned site. Finally, the K-Means clustering algorithm with K-Value equal to 3 is used to complete the initial engineering parameter planning, which better improves the accuracy and efficiency of the planning of 5G site and engineering parameters. The present invention has the advantages of adopting a continuous clustering algorithm based on grid density, improving the accuracy of site planning, and reasonably clustering weak coverage point groups.

[0072] Example 2

[0073] like Figure 1-5 As shown, a 5G site planning and regional clustering method based on a continuous clustering algorithm comprises the following steps:

[0074] Step 1: Use the 5G site agile planning method based on the continuous clustering algorithm of grid density to quickly identify the weak coverage area, and give reasonable site addition suggestions and initial engineering parameter planning suggestions through the continuous clustering algorithm;

[0075] Step 2: Analyze the geometric distance relationship between the weak coverage points and the base stations, and find the optimal base station group for the weak coverage points based on the type of base station construction, and reasonably cluster the weak coverage point groups;

[0076] First, preprocess the data and calculate the distance between all weak coverage points and the existing website address according to the two-norm formula.

[0077] Clean the coordinates of existing weak coverage points. The formula is as follows: Analyze the weak coverage points that are not within the threshold of the existing website address, sort them in descending order according to the amount of business volume, and use the points with the largest business volume as the pre-selected points for the construction of new base stations; from the set of all weak coverage points that are not within the threshold of the existing website address, select points greater than 10 and less than or equal to 30 as one category, and points less than or equal to 10 as another category, and add up the business volumes of the weak coverage points belonging to one category.

[0078] Remember C i = traffic(T1)-a×traffic(T1), where a is a limit value that needs to be optimized. Now it is just assumed to be 10; if C i > 0, choose to build a macro base station; if C i ≤0, then build a micro base station.

[0079] After the selection is completed, the points within the range of the new base station are eliminated, and then the next point with the highest traffic volume (the next largest value) is cycled; it should be noted that with the expected construction of new base stations, if the weak coverage point and the threshold of a new base station are within a range of less than 10, the weak coverage point will be skipped; the secondary cycle ends when the ratio of the sum of the traffic volume of the covered points to the sum of the traffic volume of the points not within the threshold of the existing site is greater than or equal to 0.9.

[0080] Assume that the cost of building different base stations is the total traffic volume of weak coverage points. The objective function is: in, Weak coverage points within the existing site address threshold cannot be used as base station construction locations. The coordinates of all weak coverage points are set to be the coordinates of weak coverage points that are not within the existing site address threshold; therefore, the weak coverage points are divided into two types of data: ||p″-p0||≤10, ||p′-p0||>10.

[0081] Assume the coordinates of the planned base stations. In order to achieve 90% of the total weak coverage traffic volume covered by the planned base stations, the points with large traffic volume are given priority as the locations for base station construction. Therefore, the following solution is available: If ||pL-pL′||>10, record T i =1, that is, build a base station at this point, otherwise T i =0; then traverse all weak coverage points; if ||p-pL||<10, record the total business volume within this range, recorded as tF1; if 10≤||p-pL||≤30, record the total business volume within this range, recorded as tF2.

[0082] Let α be the limit value, and assume that the limit value is α = 10; α is an assumed value at this time. After the algorithm is fully constructed, it needs to be interpolated and optimized to find the optimal value; and remove the covered points in p', and record the total business of the planned base station covered points as tF'; important constraints: From the above model, we can get: when the limit value α = 10, it can be concluded that the cost of planning the base station is 60.46 million yuan; but since this limit value α is an assumed value, there may be better points. When the model remains unchanged, α is interpolated and optimized; therefore, MATLAB is used to find the lowest point of the construction cost of the planned base station, that is, 238 macro base stations and 3060 micro base stations are built. The boundary value α is 85.28. The traffic volume of weak coverage points covered by the planned base stations is tF′=6350929.30, accounting for the total traffic volume of weak coverage points: Meet the requirements.

[0083] In actual work, weak coverage points with close distances are clustered into one category to obtain weak coverage areas. In this way, different weak coverage areas can be managed separately to better solve the weak coverage problem. If the distance between two weak coverage points is not greater than 20, then these two weak coverage points should be clustered into one category, and the clustering property is considered to be transitive, that is, if point A and point B are in the same category, point B and point C are in the same category, then points A, B and C are in the same category. First, select points with a distance of no more than 20 and use the average value point to replace them. Select a clustering algorithm to cluster the weak coverage points and calculate the time complexity. The flow chart of the clustering algorithm is as follows Figure 5 shown.

[0084] Next, we need to analyze the weak coverage points again and classify the weak coverage points that are close to each other; then we need to segment and cluster the regions, and calculate and compare the time complexity in each region.

[0085] DBI, also known as classification accuracy index, is an indicator for evaluating the quality of clustering algorithms. The smaller the DBI value, the lower the dispersion and the better the classification effect. The specific definition is as follows: Where, DBI is the DBI index value; is the average Euclidean distance from the i-th class sample to its class center; ||w i -w j ||2 is the Euclidean distance between the class centers of the i-th and j-th classes.

[0086] Step 3: Use the particle swarm iterative algorithm to iteratively calculate the parameters of the 5G base station site planning, select the base station coverage area, base station traffic volume and construction cost key factors as sub-objective functions, and finally determine the overall objective function of the optimal 5G site.

[0087] The present invention is a 5G site planning and regional clustering method based on a continuous clustering algorithm. During use, the present invention gives priority to building base stations at weak coverage points with large business volume, thereby ensuring that the business volume coverage in the area can reach the expected 90%. If the model-derived base station data transmission speed changes with the change of the region, it can be ensured that the points with large business volume must be within the coverage area, and the larger the area, the faster the data transmission speed; by standardizing the numerical value of the boundary value, a reference is provided for the type of planned base station construction, and a reference threshold is provided for the type of planned base station with irregular coverage area in reality; the present invention has the advantages of adopting a continuous clustering algorithm based on grid density, improving the accuracy of site planning, and reasonably clustering weak coverage point groups.

[0088] Example 3

[0089] like Figure 1-5 As shown, a 5G site planning and regional clustering method based on a continuous clustering algorithm comprises the following steps:

[0090] Step 1: Use the 5G site agile planning method based on the continuous clustering algorithm of grid density to quickly identify the weak coverage area, and give reasonable site addition suggestions and initial engineering parameter planning suggestions through the continuous clustering algorithm;

[0091] Step 2: Analyze the geometric distance relationship between the weak coverage points and the base stations, and find the optimal base station group for the weak coverage points based on the type of base station construction, and reasonably cluster the weak coverage point groups;

[0092] Step 3: Use the particle swarm iterative algorithm to iteratively calculate the parameters of the 5G base station site planning, select the base station coverage area, base station traffic volume and construction cost key factors as sub-objective functions, and finally determine the overall objective function of the optimal 5G site.

[0093] ① Sub-objective function 1—coverage range: The coverage range of a base station is related to multiple factors, such as transmission frequency, device loss, transmission power, antenna gain, antenna height, azimuth, and regional obstruction. These factors cannot be quantified one by one, but the coverage range of the base station can be basically determined based on the following link budget or through the Infovista5G simulation software. That is, the coverage range of these 80 base stations can be expressed as the target value f1: Where λ is the blind area loss adjustment coefficient; S is the total area of ​​the base station optimization area; S i represents the coverage area of ​​a certain base station, which can be calculated using the above formula or simulation software; C ij represents the overlapping coverage area between base stations; m represents the number of co-sited stations; and n represents the number of new-sited stations.

[0094] ②Sub-objective function 2—traffic capacity: The traffic capacity of the base station can be represented by the average monthly Internet traffic DOU per household; the LTE traffic size of the site can represent the size of future 5G traffic, so the average daily total LTE traffic can be used to represent the traffic capacity of the site; this value can be found in the network background analysis system. In order to make the variable inversely proportional to the target value, the traffic capacity is placed in the denominator, so the objective function of the traffic capacity is corrected to: In the formula, α represents the correction value, which is 1 by default; A Mi represents the original average daily total LTE traffic of the co-located site (in GB); A Ni Represents the average daily total flow of the six stations around the new site station.

[0095] ③Sub-objective function 3—construction cost: The construction cost of a base station should take into account multiple factors, including the main equipment and installation costs and supporting investment costs such as poles, power supplies, and external power. The current status of each site's machine room and antenna feeder is different, and the construction cost varies greatly. The total investment can be estimated based on the actual status of each site. In general, the cost of a new site is higher than that of a co-location transformation. The objective function is: In the formula, C Mi C represents the cost of building a co-location site; Ni Represents the construction cost of the new site.

[0096] ④ Overall optimization objective function: The three sub-objective functions are normalized by using the hierarchical analysis method, and finally the three sub-objective function variables are converted into values ​​under the same metric; finally, the overall optimization objective function is: F = ω1f1′+ω2f2′+ω3f3′, where ω i Represents the weight of each sub-objective function; f i ′ represents the normalized value; the smaller the optimization target value F is, the better; the variable values ​​in the above formula are very reliable, which can ensure the reliability of the final target value, such as the coverage range using software simulation or link budget; the traffic capacity query background system data is real and reliable; the construction cost can be accurately estimated based on the latest budget estimate.

Claims

1. A 5G site planning and regional clustering method based on a continuous clustering algorithm, characterized by: The method comprises the following steps: Step 1: Use the 5G site agile planning method based on the continuous clustering algorithm of grid density to quickly identify the weak coverage area, and give reasonable site addition suggestions and initial engineering parameter planning suggestions through the continuous clustering algorithm; Step 2: Analyze the geometric distance relationship between the weak coverage points and the base stations, and find the optimal base station group for the weak coverage points based on the type of base station construction, and reasonably cluster the weak coverage point groups; Step 3: Use the particle swarm iterative algorithm to iteratively calculate the parameters of the 5G base station site planning, select the base station coverage area, base station traffic volume and construction cost key factors as sub-objective functions, and finally determine the overall objective function of the optimal 5G site.

2. The 5G site planning and regional clustering method based on the continuous clustering algorithm according to claim 1, characterized in that: The step 1 specifically comprises the following steps: Step 1.1: Pre-processing of weak coverage raster data; Step 1.2: DBSCAN density clustering; Step 1.3: K-Means clustering; Step 1.4: Output of planning site; Step 1.5: Initial engineering parameter planning of the community.

3. The 5G site planning and regional clustering method based on the continuous clustering algorithm according to claim 2, characterized in that: The weak coverage raster data pre-processing in step 1.1 specifically includes the following steps: Step 1.11: Filter the grid data whose RSRP_0 is less than the weak coverage threshold n; Step 1.12: Filter the grid data whose distance from the weak coverage grid to the master cell (Antenna_0) is greater than the weak coverage distance threshold d; Step 1.13: After collaborative filtering, the RSRP_0 at different heights of the same grid is averaged.

4. The 5G site planning and regional clustering method based on the continuous clustering algorithm according to claim 2, characterized in that: The DBSCAN density clustering in step 1.2 specifically includes the following steps: Step 1.21: Input data source set (mainly involving 5G site initialization data) is denoted as D; Step 1.22: Mark all objects in set D as unvisited; Step 1.23: Randomly select an unvisited object in set D as D; Step 1.24: Preset the parameter eps to the specified circle radius and parameter MinPts. If there are at least MinPts objects in the neighborhood of eps with p as the center, a new cluster C is created; Step 1.25: Let N be the set of objects in the eps neighborhood of p. Repeat step (4) for each point in N. All objects that meet the conditions form the same cluster C′. Step 1.26: If there are still objects that do not belong to any cluster after the above steps, they are marked as noise points, and the density clustering results with parameters eps and MinPts as constraints are finally output.

5. The 5G site planning and regional clustering method based on the continuous clustering algorithm according to claim 2, characterized in that: The K-Means clustering in step 1.3 specifically includes the following steps: Step 1.31: Input the data source, data set D and the preset number of clusters K; Step 1.32: Randomly select K objects in the data set D as the initial clustering centers; Step 1.33: Calculate the distance between each object and each seed cluster center, and assign each object to the cluster center closest to it; the cluster centers and the objects assigned to them are a cluster group; Step 1.34: Each time a sample is assigned, the cluster center is recalculated based on the existing objects in the cluster. This process will be repeated until no object is reassigned to a different cluster or no cluster center changes; Step 1.35: According to the above steps, the second K-Means clustering result can be finally obtained.

6. The 5G site planning and regional clustering method based on the continuous clustering algorithm according to claim 1, characterized in that: The base station coverage area sub-objective function in step 3 is specifically: the base station coverage range can be expressed as the target value f1: Where λ is the blind area loss adjustment coefficient; S is the total area of ​​the base station optimization area; S i Represents the coverage area of ​​a base station; C ij represents the overlapping coverage area between base stations; m represents the number of co-sited stations; and n represents the number of new-sited stations.

7. The 5G site planning and regional clustering method based on the continuous clustering algorithm according to claim 6, characterized in that: The base station traffic quantum objective function in step 3 is specifically: the variable is inversely proportional to the target value, and the traffic capacity is placed in the denominator, so the objective function of the traffic capacity is corrected to: In the formula, α represents the correction value; A Mi represents the original average daily total traffic of LTE at the co-sited station; A Ni Represents the average daily total flow of stations around the new site.

8. The 5G site planning and regional clustering method based on the continuous clustering algorithm according to claim 7, characterized in that: The construction cost sub-objective function in step 3 is specifically: The objective function is: In the formula, C Mi C represents the cost of building a co-location site; Ni Represents the construction cost of the new site.

9. The 5G site planning and regional clustering method based on the continuous clustering algorithm according to claim 8, characterized in that: The overall objective function of the 5G optimal site in step 3 is specifically as follows: the three sub-objective functions are normalized by using the hierarchical analysis method, and finally the three sub-objective function variables are converted into values ​​under the same metric; finally, the overall optimization objective function is: F = ω1f1′+ω2f2′+ω3f3′, where ω i represents the weight of each sub-objective function; f i ' represents the normalized value.

Images