Redundant cell identification method, apparatus and device, and readable medium

Abstract

The invention discloses a redundant cell identification method, apparatus and device, and a readable medium. Working parameter data and MDT data are collected; constructing sampling point level features, grid level features, cell level features and cluster and region level features according to the work parameter data and the MDT data; screening out to-be-judged grids with high redundancy potential from the grids of the whole network; clustering the grids to be judged through an HDBSCAN algorithm, and obtaining a high-redundancy sector area through clustering merging and edge cluster expansion; by taking the high-redundancy sector area as an analysis range, calculating a metadata updating result and a related statistical index after any cell in the area is closed; constructing a multi-objective function including a coverage quality objective, a capacity objective and an economical efficiency objective, and setting constraint conditions; and optimizing chromosome codes through an initialization strategy, strengthening the result of the design evolution operator and the elite retention strategy, and outputting a redundant cell list. According to the scheme of the invention, the accuracy of redundant cell screening can be improved.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a method, apparatus, device, and readable medium for identifying redundant cells. Background Technology

[0002] With the continuous development of the communications industry and the increasing number of 5G sites, in dense network scenarios, multiple sectors overlap, leading to redundancy in some sectors. This means that the coverage and capacity of these cells can be shared by other cells after they are shut down. Identifying and shutting down these cells can reduce co-channel interference, save power, and facilitate relocation and reuse, achieving cost reduction and efficiency improvement.

[0003] Current methods for assessing and screening redundant cells mainly rely on expert experience, using traffic volume and coverage indicators for evaluation. If a cell's traffic volume remains at a low level for a long period, below a certain threshold, it may be identified as a redundant cell. If a cell has strong coverage in its neighboring cells, or if the coverage radius of the cell is significantly larger than the actual coverage required, it can be considered a cell to be considered redundant.

[0004] However, relying solely on experience to make judgments can easily lead to discrepancies with the actual coverage situation in complex scenarios. This can result in inaccurate assessments of a particular sector due to either overestimating or underestimating its geographical coverage impact. Summary of the Invention

[0005] This invention provides a method, apparatus, device, and readable medium for identifying redundant cells, thereby improving the accuracy of redundant cell screening.

[0006] According to one aspect of the present invention, a method for identifying redundant cells is provided, comprising:

[0007] The system collects engineering parameter data and MDT data. The engineering parameter data includes the base station name, unique identifier ID, longitude and latitude, coverage type, antenna suspension height, antenna azimuth angle, scene type, transmit power, and minimum access level. The MDT data includes the unique identifier ID of the grid, the unique identifier ID of the sampling point, the cell identifier CI of the primary serving cell, the frequency band of the primary serving cell, the reference received signal power RSRP of the primary serving cell, the signal-to-interference-plus-noise ratio SINR of the primary serving cell, the cell identifier CI of the neighboring cells, the frequency band of the neighboring cells, and the reference received signal power RSRP of the neighboring cells.

[0008] Based on the aforementioned engineering parameter data and MDT data, sampling point-level features, grid-level features, cell-level features, and cluster and region-level features are constructed.

[0009] From the entire network of rasters, rasters with high redundancy potential are selected for decision-making; the rasters to be decided are clustered using the HDBSCAN algorithm, and high redundancy sector regions are obtained through cluster merging and edge cluster expansion;

[0010] Using the high-redundancy sector region as the analysis scope, calculate the metadata update results and related statistical indicators after shutting down any cell in the region;

[0011] Construct a multi-objective function that includes coverage quality, capacity, and economic objectives, and set constraints; output a list of redundant cells by initialization strategy, chromosome encoding optimization, strengthening the design of evolutionary operators, and elite retention strategy.

[0012] Optionally, the sampling point-level features are obtained by rasterizing and filling the sampling points with the engineering parameter data and MDT data; the raster-level features are obtained by aggregating the data of all sampling points within a raster; the cell-level features include raster-dominant cell-level features and region-dominant cell-level features, wherein the raster-dominant cell-level features are centered on the dominant cell and aggregate the sampling point feature data of all associated rasteres within the coverage area of ​​the dominant cell; the region-dominant cell-level features are centered on the dominant cell and aggregate the sampling point feature data of all associated rasteres within the coverage area of ​​the dominant cell; the cluster and region-level features are based on the target area for redundancy analysis and aggregate the raster-level features of all rasteres and the cell-level features of all cells within the target area.

[0013] Optionally, the step of selecting grid cells to be judged from the entire grid network that exhibit high redundancy potential includes:

[0014] Determine the number of neighboring cells for each grid, sort the grids according to the number of neighboring cells, and determine the quantiles of the number of neighboring cells;

[0015] Based on the scene type of the target analysis area, select the corresponding quantile ratio R, and filter out the grids with more than the number of neighboring cells than the quantile ratio R; wherein, the scene type is divided into urban and county scene and suburban area scene;

[0016] For the grid obtained after quantile filtering, a threshold T for the total number of neighboring cells is preset based on the scene type of the target analysis area. This threshold T is used to remove grids with too few neighboring cells and low redundancy potential.

[0017] The grid cells to be determined are obtained by removing those whose total number of neighboring cells is less than the threshold T.

[0018] Optionally, the step of clustering the raster to be determined using the HDBSCAN algorithm includes:

[0019] From the raster-level features, the geographic features, redundancy features, and basic identifier of the raster to be judged are extracted; wherein, the geographic features include the longitude and latitude of the raster center, which are used to reflect the spatial location of the raster; the redundancy features are the number of neighboring cells in the raster, which are used to reflect the redundancy potential of the raster; the basic identifier is the raster ID; the geographic features, redundancy features, and basic identifier are integrated to form a clustered dataset;

[0020] The cluster stability index, the noise point ratio, and the relative clustering effectiveness index are set. The cluster stability index is used to measure the structural stability of the clusters formed by clustering. The noise point ratio is used to measure the proportion of grids identified as noise during the clustering process out of the total number of grids to be identified. The relative clustering effectiveness index is used to verify the rationality of the density distribution of the clustering results.

[0021] The initial parameters of the HDBSCAN algorithm are determined as follows: clustering distance, core grid distance, minimum cluster size, and minimum number of neighbors. The clustering distance refers to the distance between the center points of any two grids, calculated using the great circle distance method combined with the longitude and latitude of the grids to construct a grid-to-grid distance matrix for clustering. The core grid distance is the minimum distance required for each grid to find its minimum number of neighboring grids within a set radius. The minimum cluster size is set based on the average number of grid cells in the region. The minimum number of neighbors is set based on the average distance between stations in the region.

[0022] Clustering is performed using the HDBSCAN algorithm. The iteration is terminated when the clustering quality index simultaneously meets the preset standard or the number of iterations reaches the preset maximum number of iterations, thus obtaining the preliminary clustering results.

[0023] Optionally, obtaining highly redundant sector regions through cluster merging and edge cluster expansion includes:

[0024] Based on the cluster center coordinates of each cluster in the preliminary clustering results, calculate the distance between any two cluster centers; determine whether two clusters belong to the same type of cluster, the criterion being whether the distribution of the number of neighboring cells in the raster within the cluster is similar; if the center distance between two clusters of the same type is less than a preset value, merge the two clusters into one cluster, a high-redundancy sector region.

[0025] For all the merged clusters, sort them in descending order according to the number of dominant cells within the cluster, and remove all clusters with fewer than a preset number of dominant cells;

[0026] The entire network of graticles is filtered, and the filtering criteria are that the main serving cell or neighboring cells of the graticle belong to the list of dominant cells in the preliminary clustering results; the filtered graticles are included in the preliminary clustering results to form the final high redundancy sector region.

[0027] Optionally, the step of using the high-redundancy sector region as the analysis scope to calculate the metadata update results and related statistical indicators after shutting down any cell in the region includes:

[0028] Set a corresponding threshold θ according to the different network standards; for each sampling point within the original coverage area of ​​the closed cell, find the strongest neighboring cell with the largest RSRP among the neighboring cells of the sampling point. If the RSRP of the strongest neighboring cell is not less than θ, update the new level value of the sampling point to the RSRP of the strongest neighboring cell; if the RSRP of the strongest neighboring cell is less than θ, mark the sampling point as a missing sampling point.

[0029] The number of actual sampling points is calculated, and the statistical range is all the grids originally covered by the closed cell. For each grid, the number of sampling points in the grid is multiplied by the result of the indicator function, and then the calculation results of all grids are summed to obtain the total number of lost sampling points.

[0030] Calculate the number of sampling points in the remaining sectors. For each remaining sector, the new number of sampling points is equal to the original number of sampling points in that sector plus the sampling point increment.

[0031] The statistical indicators include the proportion of degraded gratings reflecting the degree of coverage quality deterioration and the proportion of coverage holes reflecting the proportion of uncovered areas. Specifically, the criteria for judging degraded gratings are determined, and the proportion of degraded gratings is calculated by dividing the number of degraded gratings in the analysis area by the total number of gratings in that area. The criteria for judging coverage holes are determined by determining that if no strongest neighbor cell is found for all sampling points in a certain grating after the cell is closed, then the grating is a coverage hole grating. The proportion of coverage holes is calculated by dividing the number of coverage hole gratings in the analysis area by the total number of gratings in that area.

[0032] Optionally, the construction of a multi-objective function covering quality, capacity, and economic objectives, and the setting of constraints, includes:

[0033] The coverage quality objective function is obtained by multiplying the first dynamic weight factor by the ratio of the degraded grid increment to the total number of grids, and adding the second dynamic weight factor by the ratio of the number of newly added coverage hole grids to the total number of grids; wherein, the degraded grid increment refers to the number of degraded grids added after the cell is closed, and the number of newly added coverage hole grids refers to the number of newly added coverage hole grids after the cell is closed, and the first dynamic weight factor and the second dynamic weight factor are adjusted according to the priority of network optimization;

[0034] For all remaining sectors, calculate the difference between the sampling point load and the congestion threshold for each sector, divide it by the congestion threshold, and take the maximum value among all calculation results to obtain the capacity objective function; wherein, the sampling point load of the sector is the number of sampling points in that sector, and the congestion threshold is the maximum number of sampling points preset according to the capacity carrying capacity of the sector.

[0035] The economic objective function is obtained by dividing the number of closed cells by the total number of candidate redundant cells; wherein, the candidate redundant cells refer to cells with redundancy potential that have been initially screened.

[0036] The constraints include: the proportion of degraded grids exceeds a preset threshold; the proportion of coverage holes exceeds a preset threshold; the total number of lost sampling points exceeds a preset threshold; and the number of sampling points in any remaining sector exceeds a preset threshold.

[0037] Optionally, the initialization strategy includes: determining the calculation factors for the initialization weights, including the number of neighboring cells of the candidate cell, the average distance between neighboring cells, and the current load.

[0038] Construct an initial weight model and calculate the initial weight for each candidate cell: divide the number of neighboring cells of the cell by the average distance between neighboring cells, and multiply by the natural exponential function;

[0039] The normalized weights are obtained by dividing the initial weight of each cell by the sum of the initial weights of all candidate cells.

[0040] An initial population is generated using binary encoding, with each chromosome corresponding to a redundant decision scheme and each gene position on the chromosome corresponding to a candidate cell. For each gene position, a uniformly distributed random number between 0 and 1 is generated. If the random number is less than the normalized weight of the corresponding cell, the gene position is set to 1; otherwise, it is set to 0.

[0041] Optionally, the optimization of chromosome coding includes:

[0042] Determine the coverage redundancy and load pressure of a cell; wherein, the coverage redundancy refers to the proportion of grid cells within the cell's coverage area that are occupied by other cells; the load pressure refers to the ratio of the cell's current load to the congestion threshold, which reflects the cell's capacity carrying capacity.

[0043] The dynamic weight of each candidate cell is calculated based on the coverage redundancy and the load pressure.

[0044] The dynamic weights of the gene loci in all candidate cells are sorted to determine the quantiles, and the weight value corresponding to the quantile is used as the screening threshold. For each gene locus in a candidate cell, if its dynamic weight is greater than the screening threshold, the gene locus is set to 1; otherwise, it is set to 0.

[0045] Optionally, the enhanced design evolution operator includes:

[0046] Extract the set of spatially adjacent cells, construct a crossover mask, and perform crossover on the parent chromosome according to the mask, preserving the continuity of geographical distribution;

[0047] Based on the sensitivity to coverage degradation, the mutation probability is calculated, and mutations are preferentially implemented in low-sensitivity, high-load areas to optimize population diversity.

[0048] Optionally, the elite retention strategy includes:

[0049] A three-dimensional elite pool is constructed, stratified by the number of closures, coverage degradation rate, and load balancing degree. Elite individuals with the highest quantiles in the three dimensions are retained in the non-dominated solutions. When the elite pool is full, similar solutions are eliminated based on the crowding distance.

[0050] According to another aspect of the present invention, a redundant cell identification device is provided, comprising:

[0051] The data acquisition unit is used to collect engineering parameter data and MDT data. The engineering parameter data includes the base station name, unique identifier ID, longitude and latitude, coverage type, antenna suspension height, antenna azimuth angle, scene type, transmit power, and minimum access level. The MDT data includes the unique identifier ID of the grid, the unique identifier ID of the sampling point, the cell identifier CI of the primary serving cell, the frequency band of the primary serving cell, the reference received signal power RSRP of the primary serving cell, the signal-to-interference-plus-noise ratio SINR of the primary serving cell, the cell identifier CI of the neighboring cell, the frequency band of the neighboring cell, and the reference received signal power RSRP of the neighboring cell.

[0052] The feature construction unit is used to construct sampling point-level features, grid-level features, cell-level features, and cluster and region-level features based on the engineering parameter data and MDT data.

[0053] The grid filtering unit is used to filter out grids with high redundancy potential from the entire grid; the grids to be judged are clustered by the HDBSCAN algorithm, and high redundancy sector regions are obtained by cluster merging and edge cluster expansion;

[0054] The calculation unit is used to calculate the metadata update results and related statistical indicators after shutting down any cell in the high redundancy sector region, taking the high redundancy sector region as the analysis scope;

[0055] The output unit is used to construct a multi-objective function that includes coverage quality, capacity, and economic objectives and set constraints; it outputs a list of redundant cells by initializing strategies, optimizing chromosome coding, strengthening the design of evolutionary operators, and implementing elite retention strategies.

[0056] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0057] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the redundant cell identification method according to any embodiment of the present invention.

[0058] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the redundant cell identification method according to any embodiment of the present invention.

[0059] According to another aspect of the present invention, a computer program product is provided, the computer program product comprising a computer program that, when executed by a processor, implements the redundant cell identification method according to any embodiment of the present invention.

[0060] The technical solution of this invention uses MDT data to identify highly redundant sector regions through HDBSCAN clustering, and then uses the NSGA-II algorithm to process the multi-objective trade-off problem to obtain a list of redundant cells, thereby improving the accuracy of redundant cell identification.

[0061] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0062] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is a flowchart of a redundant cell identification method provided in Embodiment 1 of the present invention;

[0064] Figure 2 This is a flowchart of a grid filtering method to be determined provided in Embodiment 2 of the present invention;

[0065] Figure 3 This is a flowchart of a clustering method using the HDBSCAN algorithm provided in Embodiment 3 of the present invention;

[0066] Figure 4 This is a flowchart of a method for calculating metadata update results and related statistical indicators provided in Embodiment 4 of the present invention;

[0067] Figure 5 This is a schematic diagram of the structure of a redundant cell identification device provided in Embodiment 5 of the present invention;

[0068] Figure 6 This is a schematic diagram of the structure of an electronic device that implements the redundant cell identification method of the present invention. Detailed Implementation

[0069] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0070] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0071] Example 1

[0072] Figure 1 This is a flowchart of a redundant cell identification method provided in Embodiment 1 of the present invention. This embodiment is applicable to the identification of redundant cells. The method can be executed by a redundant cell identification device, which can be implemented in hardware and / or software and can be configured in an electronic device. Figure 1 As shown, the method includes:

[0073] S110, collect engineering parameter data and MDT data.

[0074] Engineering parameter data includes fields such as base station name, base station ID, longitude, latitude, coverage type, antenna height, antenna azimuth angle, scene type, transmit power, and minimum access level; MDT data includes relevant information such as grid ID, sampling point ID, primary cell CI, primary cell frequency band, primary cell RSRP, primary cell SINR, neighboring cell CI, neighboring cell frequency band, and neighboring cell RSRP.

[0075] Preferably, the collected data is processed through the following methods: 1. Missing value handling: Sampling points with missing RSRP in the main cell are directly removed; 2. Outlier filtering: Sampling points with RSRP > -40dBm or RSRP < -156dBm (< -140dBm for 4G) are considered outliers and their records are closed; 3. Neighbor cell matching: Neighbor cell CIs are completed using engineering parameter tables (PCI / CGI, latitude and longitude). If matching fails or the neighbor cell RSRP is missing, the neighbor cell record is closed.

[0076] S120. Based on the engineering parameter data and MDT data, construct sampling point-level features, grid-level features, cell-level features, and cluster and region-level features.

[0077] The feature classification into four levels—sampling point level, raster level, cell level, and cluster / region level—is designed to address the problems of excessively coarse data granularity, imprecise evaluation, and inaccurate regional division in existing technologies, and to adapt to the different needs of the entire redundant cell identification process. The sampling point level provides the most microscopic foundation, recording the location and signal of individual sampling points based on MDT data, providing raw, fine-grained data for redundancy analysis and avoiding the coarseness of traditional sector-level data. The raster level aggregates sampling points into geographic raster cells, eliminating accidental differences in individual sampling points and forming overall raster coverage quality indicators (such as coverage rate and number of neighboring cells), which are directly used for high-redundancy raster screening and HDBSCAN clustering. The cell level focuses on the cell itself, evaluating cell coverage and degradation ratio from both the raster-level (raster-dominant cell level) and regional-level (region-dominant cell level) perspectives, providing direct evidence for evaluating the role of redundant cells. The cluster / region level reflects the overall situation of high-redundancy clusters or target areas at a macroscopic level, supporting the judgment of the integrity of high-redundancy areas and the regional constraints of NSGA-II multi-objective decision-making.

[0078] In this embodiment of the invention, the sampling point-level features are obtained by rasterizing and filling the sampling points with engineering parameter data and MDT data; the raster-level features are obtained by aggregating the data of all sampling points within the raster; the cell-level features include raster-dominant cell-level features and region-dominant cell-level features, wherein the raster-dominant cell-level features are centered on the dominant cell and aggregate the sampling point feature data of all associated rasteres within the coverage area of ​​the dominant cell; the region-dominant cell-level features are centered on the dominant cell and aggregate the sampling point feature data of all associated rasteres within the coverage area of ​​the dominant cell; the cluster and region-level features are based on the target area for redundancy analysis and aggregate the raster-level features of all rasteres and the cell-level features of all cells within the target area.

[0079] The sampling point level features are obtained by rasterizing the sampling points using the MDT platform and live network data parameters, resulting in a sampling point dataset (P). Features include raster ID, sampling point ID, sampling point longitude, sampling point latitude, primary cell CI, and neighboring cell CI.

[0080] Raster-level features are aggregated from sampling point-level features (S) for subsequent raster coverage quality assessment. All sampling point data within a raster are aggregated to form a raster-level indicator, which mainly includes raster center longitude, raster center latitude, raster coverage rate, dominant cell CI details, and neighboring cell CI details.

[0081] Cell-level features are aggregated at the cell level (C). For ease of subsequent statistical calculations, when aggregating features at the sampling point level, two types of features are distinguished: grid-dominant cell-level features (SC) and region-dominant cell-level features (QC). Grid-dominant cell-level features aggregate sampling point features within the same grid, based on the dominant cell. These primarily include the dominant cell's CI, longitude, latitude, number of sampling points, and the corresponding neighboring cell CIs, their corresponding number of sampling points, and the average RSRP of the neighboring cells. Within a cluster or the entire region, region-level features are aggregated based on dominant cell-related information. Indicators include cell CI, cell longitude, cell latitude, number of occupied grids, and proportion of degraded grids.

[0082] Cluster and region-level features are defined on a cluster or entire region basis, aggregating the overall statistical features of the clusters to be judged and the region as a whole. These include information such as the number of raster cells, the proportion of degraded raster cells, coverage rate, list of dominant cells, and list of neighboring cells.

[0083] S130. Select grid cells with high redundancy potential from the entire grid; cluster the grid cells to be determined using the HDBSCAN algorithm, and obtain high redundancy sector regions through cluster merging and edge cluster expansion.

[0084] The main purpose of identifying high-redundancy sector regions is to filter out graticles with a large number of neighboring cells and cluster them, with the resulting clusters being designated as high-redundancy regions. The selection of graticles to be judged is based on the number of neighboring cells as the core indicator of high redundancy potential. First, graticles are filtered by scenario quantile, then graticles with a total number of neighboring cells below the scenario threshold are removed, retaining graticles with high neighboring cell overlap and high redundancy potential.

[0085] HDBSCAN clustering takes the latitude and longitude of the raster to be judged and the number of neighboring cells as input. By dynamically adjusting the clustering parameters and combining indicators such as cluster stability and noise ratio, the clustering quality is evaluated to obtain a preliminary high-redundancy cluster.

[0086] Cluster merging and edge expansion merge clusters of the same type that are close to each other, and remove micro-clusters with fewer dominant cells to de-fragment; then, based on the list of dominant cells in the cluster, select grids in the entire network that contain these dominant cells to include in the cluster, complete the edge coverage, and finally form a high-redundancy sector area.

[0087] S140. Taking the high-redundancy sector area as the analysis scope, calculate the metadata update results and related statistical indicators after shutting down any cell in the area.

[0088] When a sector is shut down, some sampling points are transferred to the strongest neighboring cell, while others suffer coverage loss because the strongest signal level is below a threshold. The signal level of the sampling points transferred to the strongest neighboring cell is the same as that of the original strongest neighboring cell. Signal migration after cell shutdown is simulated using grid-level MDT data. The availability of coverage is determined by whether the neighboring cell RSRP (Recovery Surrounding Performance Ratio) meets the standard. Furthermore, network impact is quantified using grid aggregation indicators to ensure that coverage quality remains within acceptable limits after redundant cells are shut down.

[0089] S150. Construct a multi-objective function that includes coverage quality, capacity, and economic objectives, and set constraints. Output a list of redundant cells by initializing the strategy, optimizing chromosome coding, strengthening the design of evolutionary operators, and implementing elite retention strategies.

[0090] The purpose of coverage quality objectives is to avoid a significant drop in coverage after cell shutdown; the purpose of capacity objectives is to minimize the overload of remaining sectors and prevent congestion in neighboring cells due to load sharing; the purpose of economic objectives is to maximize the ratio of shutdown cells to candidate cells, achieving cost reduction and efficiency improvement. A termination threshold is set, and the decision is terminated when any condition is met, thus preventing network performance degradation.

[0091] The initialization strategy builds a weighted model based on the number of neighboring cells, average distance between neighboring cells, and current load of candidate cells, and generates an initial population according to probability to ensure that the initial solution takes into account both high redundancy and low load, thus avoiding coverage gaps and congestion from the source. Chromosome coding optimization uses coverage redundancy and load pressure to calculate dynamic weights, allowing only cells with high redundancy and low load to participate in the decision-making process, and eliminating cells with no redundancy value. The evolutionary operator strengthens population diversity and evolutionary efficiency through geographical association crossover and directed mutation. The elite retention strategy establishes a three-dimensional elite pool to retain high-quality individuals from the non-dominated solutions of each generation. After the pool is full, similar solutions are eliminated according to the crowding distance to ensure the diversity and optimality of the solutions.

[0092] Through iterative evolution using the above steps, the final non-dominated solution in the elite pool is the optimal redundancy scheme. Cells marked as to be closed in the solution are then organized to form a list of redundant cells.

[0093] Example 2

[0094] Figure 2 This is a flowchart of a grid filtering method according to Embodiment 2 of the present invention. This embodiment further explains and illustrates the method based on the above embodiments. Figure 2 As shown, the method includes:

[0095] S210. Determine the number of neighboring cells for all grids, sort the grids according to the number of neighboring cells, and determine the quantiles of the number of neighboring cells.

[0096] Filtering graticules above the scene quantile: By calculating the quantile of the number of neighboring cells, graticules above the quantile are filtered out. Based on the base station density and user situation in different scenarios, and based on experience, the graticules with the highest number of neighboring cells (R, recommended value: 25%) are selected as the graticules to be judged.

[0097] S220. Based on the scene type of the target analysis area, select the corresponding quantile ratio R and filter out the grids with more neighboring cells than the quantile ratio R; wherein, the scene type is divided into urban and county scene and suburban area scene.

[0098] Urban and county towns: High site density, large overlapping coverage areas, and a large number of neighboring cells within a grid; a recommended R value of 25% is suggested. The percentage can be adjusted appropriately based on the total number of cells and the expected closure of redundant cells.

[0099] Suburban areas: Site density is low, and there are fewer areas with concentrated sites. A recommended R value is 20%. The percentage can be adjusted appropriately based on the total number of cells and the expected closure of redundant cells.

[0100] S230. For the raster obtained after quantile filtering, combined with the scene type of the target analysis area, a preset threshold T for the total number of neighboring cells is set. This threshold T is used to remove raster cells with too few neighboring cells and low redundancy potential.

[0101] S240. Remove grid cells whose total number of neighboring cells is less than the threshold T to obtain the grid cell to be determined.

[0102] Grids with fewer than a threshold number of neighboring cells are removed. The threshold T is set according to different scenarios, and grids with fewer than the threshold number of neighboring cells are removed.

[0103] Example 3

[0104] Figure 3 This is a flowchart of a clustering method using the HDBSCAN algorithm provided in Embodiment 3 of the present invention. This embodiment further explains and illustrates the method based on the above embodiments. Figure 3 As shown, the method includes:

[0105] S310. Extract the geographic features, redundant features, and basic identifier of the raster to be judged from the raster-level features. The geographic features include the longitude and latitude of the raster center, which are used to reflect the spatial location of the raster. The redundant features are the number of neighboring cells in the raster, which are used to reflect the redundancy potential of the raster. The basic identifier is the raster ID. Integrate the geographic features, redundant features, and basic identifier to form a clustering dataset.

[0106] From the raster-level features, geographic features, namely the longitude / latitude of the raster center, are selected to determine the spatial location of the raster, used to calculate the distance between rasteres; redundancy features, namely the number of neighboring cells within a raster (more neighboring cells indicate higher coverage overlap and greater redundancy potential); and a basic identifier, i.e., the raster ID, is used for subsequent data association and result traceability. These three types of information are correlated to form a structured dataset, ensuring that the spatial location, redundancy potential, and identifier of each raster correspond, providing suitable input data for the subsequent clustering of the HDBSCAN algorithm.

[0107] S320. Set the cluster stability index, noise point ratio, and relative clustering effectiveness index; where the cluster stability index is used to measure the structural stability of the clusters formed by clustering; the noise point ratio is used to measure the proportion of grids that are judged as noise during the clustering process to the total number of grids to be judged; the relative clustering effectiveness index is used to verify the rationality of the clustering results in terms of density distribution.

[0108] Cluster stability index (λ-stability):

[0109] Where, ρ core ρ represents the core density of the sample. reach This represents the cross-reach density of samples.

[0110] Noise point ratio threshold: η_threshold is set to 5%-15%; the relative clustering validity index (DBCV) is validated using the density peak criterion.

[0111] S330. Determine the initial parameters of the HDBSCAN algorithm: clustering distance, core grid distance, minimum cluster size, and minimum number of neighbors. Clustering distance refers to the distance between the center points of any two grids, calculated using the great circle distance method, combined with the longitude and latitude of the grids, to construct a grid-to-grid distance matrix for clustering. Core grid distance refers to the minimum distance required for each grid to find its minimum number of neighboring grids within a set radius. The minimum cluster size is set based on the average number of grid cells in the region. The minimum number of neighbors is set based on the average distance between stations in the region.

[0112] Clustering distance: A distance matrix is ​​constructed for clustering by calculating the distance between the center points of each grid according to the latitude and longitude of the grid center based on the great circle distance of the Earth. Each element in the matrix is ​​the pairwise distance between grids.

[0113] Core grid distance: For each grid cell, calculate its core grid distance, which is the minimum distance required to find at least min_samples neighbors within a given radius.

[0114] min_cluster_size: Based on the average number of cell grids in the region, the total number of grids in the region is divided by the number of cells.

[0115] min_samples: The station spacing varies in different scenarios, and the average station spacing in the region is used as the base value.

[0116] In addition, a dynamic parameter adjustment strategy is included to adjust the initial parameters mentioned above:

[0117] min_cluster_size iteration rules:

[0118]

[0119] Where α = 0.8 and β = 1.2 are empirical coefficients.

[0120] Adaptive formula for min_samples:

[0121]

[0122] Spatial weight factor optimization: Introducing a feature space scaling factor:

[0123]

[0124] Where σ_cluster is the standard deviation of latitude and longitude within the cluster, and σ_total is the global standard deviation.

[0125] S340. Clustering is performed using the HDBSCAN algorithm. The iteration is terminated when the clustering quality index reaches the preset standard or the number of iterations reaches the preset maximum number of iterations, and the preliminary clustering results are obtained.

[0126] The iterative process terminates when any of the following conditions are met:

[0127] (1) The quality indicators simultaneously meet:

[0128] λ stab ≥0.75&DBCV≥0.6&η noise ≤0.15;

[0129] (2) Reach the maximum number of iterations (it is recommended to set k_max = 5).

[0130] In this embodiment of the invention, a highly redundant sector region is obtained through cluster merging and edge cluster expansion, including:

[0131] Based on the cluster center coordinates of each cluster in the preliminary clustering results, calculate the distance between any two cluster centers; determine whether two clusters belong to the same type of cluster, the criterion being whether the distribution of the number of neighboring cells in the raster within the cluster is similar; if the center distance between two clusters of the same type is less than a preset value, merge the two clusters into one cluster, a high-redundancy sector region.

[0132] For all the merged clusters, sort them in descending order according to the number of dominant cells within the cluster, and remove all clusters with fewer than a preset number of dominant cells;

[0133] The entire network of graticles is filtered, and the filtering criteria are that the main serving cell or neighboring cells of the graticle belong to the list of dominant cells in the preliminary clustering results; the filtered graticles are included in the preliminary clustering results to form the final high redundancy sector region.

[0134] Preferably, clusters of the same type with a distance of less than 200m are merged. The indicators of the clusters to be judged are statistically analyzed, sorted by the number of occupied graticles and the number of dominant cells. Micro-clusters with less than 5 dominant cells are removed. Ensure that the complete area of ​​the main coverage or sidelobe coverage of decommissioned cells is included in the analysis scope to avoid omitting edge graticles. Expand the area of ​​the clusters to be judged by using the list of dominant cells in the clusters. Screen graticles containing redundant cells to be judged from the entire network graticles and include them in the judgment area.

[0135] Example 4

[0136] Figure 4This is a flowchart illustrating a method for calculating metadata update results and related statistical indicators according to Embodiment 4 of the present invention. This embodiment further explains and illustrates the method based on the above embodiments. Figure 4 As shown, the method includes:

[0137] S410. Set the corresponding threshold θ according to the different network standards; for each sampling point within the original coverage area of ​​the closed cell, find the strongest neighboring cell with the largest RSRP among the neighboring cells of the sampling point. If the RSRP of the strongest neighboring cell is not less than θ, update the new level value of the sampling point to the RSRP of the strongest neighboring cell; if the RSRP of the strongest neighboring cell is less than θ, mark the sampling point as a missing sampling point.

[0138] Update the sampling point level. If there exists a strongest neighboring cell Cneighbor and its RSRPneighbor ≥ (where RSRP is the threshold for no coverage), then:

[0139] RSRP new (p) = RSRP neighbor ;

[0140] RSRP neighbor If the value is less than θ, then p is marked as a missing sampling point:

[0141]

[0142] S420. Calculate the number of actual sampling points. The statistical range is all the grids originally covered by the closed cell. For each grid, multiply the number of sampling points in the grid by the result of the indicator function, and then sum the calculation results of all grids to obtain the total number of lost sampling points.

[0143] Calculation of total lost sampling points:

[0144]

[0145] This is an indicator function that takes the value 1 if the condition is true, and 0 otherwise.

[0146] S430. Calculate the number of sampling points in the remaining sectors. For each remaining sector, the new number of sampling points is equal to the original number of sampling points in that sector plus the sampling point increment.

[0147] The number of sampling points in the remaining sector is the sum of the original number of sampling points in the neighboring sector and the number of sampling points transferred from the closed neighboring sector:

[0148]

[0149] Increment of sector sampling points:

[0150]

[0151] S440. Statistical indicators include the proportion of degraded gratings reflecting the degree of coverage quality deterioration and the proportion of coverage holes reflecting the proportion of uncovered areas. Specifically, the criteria for judging degraded gratings are determined, and the proportion of degraded gratings is calculated by dividing the number of degraded gratings in the analysis area by the total number of gratings in that area. The criteria for judging coverage holes are determined by determining that if no strongest neighbor cell is found for all sampling points in a certain grating after the cell is closed, then the grating is a coverage hole grating. The proportion of coverage holes is calculated by dividing the number of coverage hole gratings in the analysis area by the total number of gratings in that area.

[0152] Degraded raster percentage: The percentage of raster cells within a region that meet the criteria for a degraded raster. Raster g is a degraded raster if it meets any of the following conditions:

[0153] Condition 1: Raster coverage is less than 80%;

[0154] Condition 2: Average RSRP (raster level) is below -110dBm:

[0155]

[0156] Condition 1: The average RSRP (raster level) is calculated by averaging the RSRP (including surrogate values) of all sampled points within the raster.

[0157]

[0158] Where Ng is the total number of sampling points within grid g.

[0159] Condition 2: Raster coverage, 1 minus the proportion of all weakly covered and missing sampling points within the raster to the total sampling points:

[0160]

[0161] Where Nweak(g) is the number of weak coverage sampling points within grid g, N missing (g) represents the number of missing sampling points within grid g.

[0162] If, after a cell is closed, none of the sampling points within the grid are replaced by the strongest neighboring cell, resulting in the grid having no primary coverage cell, then the grid is considered a grid without primary coverage.

[0163]

[0164] In this embodiment of the invention, a multi-objective function covering quality, capacity, and economic objectives is constructed and constraints are set, including:

[0165] The coverage quality objective function is obtained by multiplying the first dynamic weighting factor by the ratio of the degraded grid increment to the total number of grids, and adding the second dynamic weighting factor by the ratio of the number of newly added coverage hole grids to the total number of grids. Here, the degraded grid increment refers to the number of degraded grids added after the cell is closed, and the number of newly added coverage hole grids refers to the number of newly added coverage hole grids after the cell is closed. The first and second dynamic weighting factors are adjusted according to the priority of network optimization.

[0166] For all remaining sectors, calculate the difference between the sampling point load and the congestion threshold for each sector, divide it by the congestion threshold, and take the maximum value among all calculation results to obtain the capacity objective function; where the sampling point load of a sector is the number of sampling points in that sector, and the congestion threshold is the maximum number of sampling points preset according to the sector's capacity carrying capacity.

[0167] The economic objective function is obtained by dividing the number of closed cells by the total number of candidate redundant cells; where candidate redundant cells refer to cells with redundancy potential that have been initially selected.

[0168] The constraints include: the proportion of degraded grids exceeds a preset threshold; the proportion of coverage holes exceeds a preset threshold; the total number of lost sampling points exceeds a preset threshold; and the number of sampling points in any remaining sector exceeds a preset threshold.

[0169] The coverage quality target introduces a composite coverage degradation index:

[0170]

[0171] Wherein, ΔG 劣化 To degrade the raster increment, H 空洞 The number of newly added coverage holes is represented by ω1 and ω2, which are dynamic weighting factors.

[0172] Capacity target defines sector load balancing:

[0173]

[0174] Among them, S i S represents the sampling point load of sector i. th The congestion threshold;

[0175] Economic Target: Percentage of Closed Communities

[0176] The constraint strengthening process terminates if any one of the following four conditions is met:

[0177] Coverage category: Termination is based on the expected degree of coverage degradation.

[0178] Condition 1: The proportion of degraded rasters exceeds a threshold (e.g., 10%);

[0179] Condition 2: The percentage of holes in the coverage exceeds a threshold (e.g., 5%).

[0180] Capacity-based: Termination occurs when sampling points are lost or sectors are congested.

[0181] Condition 3: The total number of lost sampling points exceeds the threshold (to avoid losing sampling points).

[0182] Condition 4: The number of sampling points in any remaining sector exceeds the threshold (to avoid congestion).

[0183] In this embodiment of the invention, the initialization strategy includes: determining the calculation factors for the initialization weight, including the number of neighboring cells of the candidate cell, the average distance between neighboring cells, and the current load.

[0184] Construct an initial weight model and calculate the initial weight for each candidate cell: divide the number of neighboring cells of the cell by the average distance between neighboring cells, and multiply by the natural exponential function;

[0185] The normalized weights are obtained by dividing the initial weight of each cell by the sum of the initial weights of all candidate cells.

[0186] An initial population is generated using binary encoding, with each chromosome corresponding to a redundant decision scheme and each gene position on the chromosome corresponding to a candidate cell. For each gene position, a uniformly distributed random number between 0 and 1 is generated. If the random number is less than the normalized weight of the corresponding cell, the gene position is set to 1; otherwise, it is set to 0.

[0187] Heuristic initialization using coverage redundancy and geographical distribution ensures a balanced initial population across coverage and capacity dimensions. An initialization weight model is established by calculating the number of neighboring cells, average neighbor distance, and load of candidate cells. After normalization, an initial solution satisfying the constraints is generated probabilistically. This model prioritizes cells with reasonable coverage redundancy (moderate number and even distribution of neighboring cells) and load pressure below the congestion threshold, avoiding network coverage gaps and other cell overload risks from the initial stage, thus laying a high-quality population foundation for multi-objective optimization.

[0188]

[0189] Where, N neigh (i) represents the number of neighboring cells of cell i; S is the average distance between neighboring cells; i S represents the current load of cell i. th This is the congestion threshold.

[0190] The initialization process includes:

[0191] (1) Calculate the initial weights of all candidate cells.

[0192] (2) Normalized weights:

[0193] (3) Press The initial population is generated probabilistically, with each chromosome gene locus satisfying:

[0194]

[0195] In this embodiment of the invention, the chromosome encoding is optimized, including:

[0196] Determine the coverage redundancy and load pressure of a cell; where coverage redundancy refers to the proportion of grid cells within the cell's coverage area that are occupied by other cells; load pressure refers to the ratio of the cell's current load to the congestion threshold, and this parameter reflects the cell's capacity carrying capacity.

[0197] The dynamic weight of each candidate cell is calculated based on coverage redundancy and load pressure.

[0198] The dynamic weights of all candidate sites are sorted to determine quantiles, and the weight value corresponding to the quantile is used as the screening threshold. For each candidate site, if its dynamic weight is greater than the screening threshold, the site is set to 1; otherwise, it is set to 0.

[0199] The chromosome coding employs a dynamic weighting mechanism, using coverage redundancy (the proportion of grid cells dominated by other cells) and load pressure as dual core parameters. A threshold is used to select cells with high coverage redundancy (top 25%) and low load pressure to participate in the shutdown decision. Only cells with high coverage redundancy and low load pressure are allowed to participate in the shutdown decision.

[0200] Calculation of dynamic weights for gene loci:

[0201]

[0202] Where, N overlap (i) represents the number of cells within the coverage area of ​​cell i that are dominated by other cells; N grid (i) represents the total number of coverage grids in cell i.

[0203] Encoding rules:

[0204]

[0205] Sort all weight values ​​and use the Q quantile as the threshold.

[0206] In this embodiment of the invention, the enhanced design evolution operator includes:

[0207] Extract the set of spatially adjacent cells, construct a crossover mask, and perform crossover on the parent chromosome according to the mask, preserving the continuity of geographical distribution;

[0208] Based on the sensitivity to coverage degradation, the mutation probability is calculated, and mutations are preferentially implemented in low-sensitivity, high-load areas to optimize population diversity.

[0209] The geographic association cross operator constructs a cross mask by clustering spatially adjacent cells, preserving the continuity of geographical distribution and avoiding regional coverage gaps caused by cell closures. The targeted mutation mechanism dynamically adjusts the mutation probability based on coverage degradation sensitivity and load pressure, prioritizing mutations in low-sensitivity, high-load areas.

[0210] (1) Geographic association crossover operator;

[0211] Input the parent chromosome and extract the set of spatially adjacent cells:

[0212] C adj ={(i,j)|||loc i -loc j || <R cluster};

[0213] Among them, R cluster Let be the cluster radius.

[0214] Construct the cross mask:

[0215]

[0216] Perform masked crossover:

[0217] (2) Directed mutation mechanism;

[0218] Mutation probability calculation:

[0219]

[0220] in, To cover degradation sensitivity; k = 0.5 is the shape factor.

[0221] In this embodiment of the invention, the elite retention strategy includes:

[0222] A three-dimensional elite pool is constructed, stratified by the number of closures, coverage degradation rate, and load balancing degree. Elite individuals with the highest quantiles in the three dimensions are retained in the non-dominated solutions. When the elite pool is full, similar solutions are eliminated based on the crowding distance.

[0223] The three-dimensional elite pool is stratified and archived according to the number of shutdowns, coverage degradation rate, and load balancing, retaining the top 10% of elite individuals in each of the three dimensions of the multi-objective non-dominated solution. A congestion distance elimination mechanism avoids excessive concentration of solutions, ensuring the diversity of the Pareto front. The significance of this strategy lies in simultaneously optimizing the three objectives of network shutdown scale, coverage quality assurance, and capacity. The specific method is as follows:

[0224] (1) Hierarchical archiving:

[0225] First dimension: sorted in descending order by the number of closed items;

[0226] Second dimension: Sorted in ascending order of degradation rate;

[0227] Third dimension: By load balancing degree (1-max(S) i / S th Arrange in descending order.

[0228] (2) Elite Selection Rules:

[0229] Each generation retains the top 10% of solutions in each of the three dimensions of the non-dominated solutions;

[0230] When the elite pool is full, similar solutions are eliminated based on crowding distance.

[0231] Furthermore, dynamic optimization management methods can be incorporated, with an adaptive penalty function dynamically adjusting the coverage degradation penalty coefficient and load balancing weights. Initially, the focus is on increasing the number of shutdowns, while in the mid-to-late stages, coverage quality constraints are strengthened. A weight migration mechanism shifts the algorithm from coarse-grained capacity release to fine-grained coverage redundancy maintenance. By balancing the focus of different stages of network optimization, dynamic adaptation between capacity expansion and redundancy assurance is ensured, preventing global optimization from getting trapped in local optima.

[0232] (1) Adaptive penalty function:

[0233]

[0234] Dynamically adjust parameters:

[0235]

[0236] (2) Dynamic weight transfer:

[0237]

[0238] When evaluating redundant cells, the Pareto optimization function integrates three indicators: coverage loss rate, capacity gain, and redundancy adjustment amount, to select the non-dominated solution set. Sensitivity analysis quantifies the impact of shutting down each cell on the overall network through a shutdown sensitivity index, providing a risk-controlled shutdown plan for the final decision. The physical significance of this module lies in achieving scientific decision-making under multi-objective trade-offs, ensuring that network redundancy is optimized while maintaining capacity and coverage performance within acceptable thresholds.

[0239] The non-dominated solution set in the elite pool represents the optimal redundant cell shutdown scheme under different optimization objectives, i.e., the reliable redundant cell list.

[0240] (1) The elite keywords generated in the elite retention strategy are used to form the list of candidate communities.

[0241] (2) The list of cells that satisfy the Pareto optimization function and the shutdown sensitivity index in the evaluation and decision-making method is the final decision redundant cell list.

[0242] Example 5

[0243] Figure 5 This is a schematic diagram of a redundant cell identification device provided in Embodiment 5 of the present invention. Figure 5 As shown, the device includes:

[0244] The data acquisition unit 510 is used to acquire engineering parameter data and MDT data. The engineering parameter data includes the base station name, unique identifier ID, longitude and latitude, coverage type, antenna suspension height, antenna azimuth angle, scene type, transmit power and minimum access level. The MDT data includes the unique identifier ID of the grid, the unique identifier ID of the sampling point, the cell identifier CI of the primary serving cell, the frequency band of the primary serving cell, the reference received signal power RSRP of the primary serving cell, the signal-to-interference-plus-noise ratio SINR of the primary serving cell, the cell identifier CI of the neighboring cell, the frequency band of the neighboring cell and the reference received signal power RSRP of the neighboring cell.

[0245] The feature construction unit 520 is used to construct sampling point-level features, grid-level features, cell-level features, and cluster and region-level features based on engineering parameter data and MDT data.

[0246] The grid filtering unit 530 is used to filter out grids with high redundancy potential from the entire grid; the grids to be judged are clustered using the HDBSCAN algorithm, and high redundancy sector regions are obtained through cluster merging and edge cluster expansion;

[0247] The calculation unit 540 is used to calculate the metadata update results and related statistical indicators after shutting down any cell in a high-redundancy sector region, with the analysis scope being a high-redundancy sector region.

[0248] Output unit 550 is used to construct a multi-objective function that includes coverage quality, capacity and economic objectives and set constraints; and outputs a list of redundant cells by initialization strategy, chromosome coding optimization, strengthening the design of evolutionary operators and elite retention strategy.

[0249] The redundant cell identification device provided in the embodiments of the present invention can execute the redundant cell identification method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method.

[0250] Example 6

[0251] Figure 6 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0252] like Figure 6 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0253] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0254] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as redundant cell identification methods.

[0255] In some embodiments, the redundant cell identification method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the redundant cell identification method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the redundant cell identification method by any other suitable means (e.g., by means of firmware).

[0256] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0257] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0258] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0259] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0260] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0261] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0262] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0263] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for identifying redundant cells, characterized in that, include: The system collects engineering parameter data and MDT data. The engineering parameter data includes the base station name, unique identifier ID, longitude and latitude, coverage type, antenna suspension height, antenna azimuth angle, scene type, transmit power, and minimum access level. The MDT data includes the unique identifier ID of the grid, the unique identifier ID of the sampling point, the cell identifier CI of the primary serving cell, the frequency band of the primary serving cell, the reference received signal power RSRP of the primary serving cell, the signal-to-interference-plus-noise ratio SINR of the primary serving cell, the cell identifier CI of the neighboring cells, the frequency band of the neighboring cells, and the reference received signal power RSRP of the neighboring cells. Based on the aforementioned engineering parameter data and MDT data, sampling point-level features, grid-level features, cell-level features, and cluster and region-level features are constructed. From the entire network of rasters, rasters with high redundancy potential are selected for decision-making; the rasters to be decided are clustered using the HDBSCAN algorithm, and high redundancy sector regions are obtained through cluster merging and edge cluster expansion; Using the high-redundancy sector region as the analysis scope, calculate the metadata update results and related statistical indicators after shutting down any cell in the region; Construct a multi-objective function that includes coverage quality, capacity, and economic objectives, and set constraints; output a list of redundant cells by initialization strategy, chromosome encoding optimization, strengthening the design of evolutionary operators, and elite retention strategy.

2. The method according to claim 1, characterized in that, The sampling point-level features are obtained by rasterizing and filling the sampling points with the engineering parameter data and MDT data; the raster-level features are obtained by aggregating the data of all sampling points within a raster; the cell-level features include raster-dominant cell-level features and region-dominant cell-level features, wherein the raster-dominant cell-level features are centered on the dominant cell and aggregate the sampling point feature data of all associated rasteres within the coverage area of ​​the dominant cell; the region-dominant cell-level features are centered on the dominant cell and aggregate the sampling point feature data of all associated rasteres within the coverage area of ​​the dominant cell; the cluster and region-level features are based on the target area for redundancy analysis and aggregate the raster-level features of all rasteres and the cell-level features of all cells within the target area.

3. The method according to claim 1, characterized in that, The process of selecting grid cells with high redundancy potential from the entire grid includes: Determine the number of neighboring cells for each grid, sort the grids according to the number of neighboring cells, and determine the quantiles of the number of neighboring cells; Based on the scene type of the target analysis area, select the corresponding quantile ratio R, and filter out the grids with more than the number of neighboring cells than the quantile ratio R; wherein, the scene type is divided into urban and county scene and suburban area scene; For the grid obtained after quantile filtering, a threshold T for the total number of neighboring cells is preset based on the scene type of the target analysis area. This threshold T is used to remove grids with too few neighboring cells and low redundancy potential. The grid cells to be determined are obtained by removing those whose total number of neighboring cells is less than the threshold T.

4. The method according to claim 1, characterized in that, The clustering of the grid cells to be determined using the HDBSCAN algorithm includes: From the raster-level features, the geographic features, redundancy features, and basic identifier of the raster to be judged are extracted; wherein, the geographic features include the longitude and latitude of the raster center, which are used to reflect the spatial location of the raster; the redundancy features are the number of neighboring cells in the raster, which are used to reflect the redundancy potential of the raster; the basic identifier is the raster ID; the geographic features, redundancy features, and basic identifier are integrated to form a clustered dataset; The cluster stability index, the noise point ratio, and the relative clustering effectiveness index are set. The cluster stability index is used to measure the structural stability of the clusters formed by clustering. The noise point ratio is used to measure the proportion of grids identified as noise during the clustering process out of the total number of grids to be identified. The relative clustering effectiveness index is used to verify the rationality of the density distribution of the clustering results. The initial parameters of the HDBSCAN algorithm are determined as follows: clustering distance, core grid distance, minimum cluster size, and minimum number of neighbors. The clustering distance refers to the distance between the center points of any two grids, calculated using the great circle distance method combined with the longitude and latitude of the grids to construct a grid-to-grid distance matrix for clustering. The core grid distance is the minimum distance required for each grid to find its minimum number of neighboring grids within a set radius. The minimum cluster size is set based on the average number of grid cells in the region. The minimum number of neighbors is set based on the average distance between stations in the region. Clustering is performed using the HDBSCAN algorithm. The iteration is terminated when the clustering quality index simultaneously meets the preset standard or the number of iterations reaches the preset maximum number of iterations, thus obtaining the preliminary clustering results.

5. The method according to claim 4, characterized in that, The high-redundancy sector region obtained through clustering and edge cluster expansion includes: Based on the cluster center coordinates of each cluster in the preliminary clustering results, calculate the distance between any two cluster centers; determine whether two clusters belong to the same type of cluster, the criterion being whether the distribution of the number of neighboring cells in the raster within the cluster is similar; if the center distance between two clusters of the same type is less than a preset value, merge the two clusters into one cluster, a high-redundancy sector region. For all the merged clusters, sort them in descending order according to the number of dominant cells within the cluster, and remove all clusters with fewer than a preset number of dominant cells; The entire network of graticles is filtered, and the filtering criteria are that the main serving cell or neighboring cells of the graticle belong to the list of dominant cells in the preliminary clustering results; the filtered graticles are included in the preliminary clustering results to form the final high redundancy sector region.

6. The method according to claim 1, characterized in that, The analysis scope is defined as the highly redundant sector region. The metadata update results and related statistical indicators after shutting down any cell in the region are calculated, including: Set a corresponding threshold θ according to the different network standards; for each sampling point within the original coverage area of ​​the closed cell, find the strongest neighboring cell with the largest RSRP among the neighboring cells of the sampling point. If the RSRP of the strongest neighboring cell is not less than the θ, update the new level value of the sampling point to the RSRP of the strongest neighboring cell. If the RSRP of the strongest neighboring cell is less than θ, the sampling point is marked as a missing sampling point; The number of actual sampling points is calculated, and the statistical range is all the grids originally covered by the closed cell. For each grid, the number of sampling points in the grid is multiplied by the result of the indicator function, and then the calculation results of all grids are summed to obtain the total number of lost sampling points. Calculate the number of sampling points in the remaining sectors. For each remaining sector, the new number of sampling points is equal to the original number of sampling points in that sector plus the sampling point increment. The statistical indicators include the proportion of degraded gratings reflecting the degree of coverage quality deterioration and the proportion of coverage holes reflecting the proportion of uncovered areas. Specifically, the criteria for judging degraded gratings are determined, and the proportion of degraded gratings is calculated by dividing the number of degraded gratings in the analysis area by the total number of gratings in that area. The criteria for judging coverage holes are determined by determining that if no strongest neighbor cell is found for all sampling points in a certain grating after the cell is closed, then the grating is a coverage hole grating. The proportion of coverage holes is calculated by dividing the number of coverage hole gratings in the analysis area by the total number of gratings in that area.

7. The method according to claim 6, characterized in that, The construction of a multi-objective function covering quality, capacity, and economic objectives, and the setting of constraints, includes: The coverage quality objective function is obtained by multiplying the first dynamic weight factor by the ratio of the degraded grid increment to the total number of grids, and adding the second dynamic weight factor by the ratio of the number of newly added coverage hole grids to the total number of grids; wherein, the degraded grid increment refers to the number of degraded grids added after the cell is closed, and the number of newly added coverage hole grids refers to the number of newly added coverage hole grids after the cell is closed, and the first dynamic weight factor and the second dynamic weight factor are adjusted according to the priority of network optimization; For all remaining sectors, calculate the difference between the sampling point load and the congestion threshold for each sector, divide it by the congestion threshold, and take the maximum value among all calculation results to obtain the capacity objective function; wherein, the sampling point load of the sector is the number of sampling points in that sector, and the congestion threshold is the maximum number of sampling points preset according to the capacity carrying capacity of the sector. The economic objective function is obtained by dividing the number of closed cells by the total number of candidate redundant cells; wherein, the candidate redundant cells refer to cells with redundancy potential that have been initially screened. The constraints include: the proportion of degraded grids exceeds a preset threshold; the proportion of coverage holes exceeds a preset threshold; the total number of lost sampling points exceeds a preset threshold; and the number of sampling points in any remaining sector exceeds a preset threshold.

8. The method according to claim 7, characterized in that, The initialization strategy includes: determining the calculation factors for the initialization weights, including the number of neighboring cells of the candidate cell, the average distance between neighboring cells, and the current load. Construct an initial weight model and calculate the initial weight for each candidate cell: divide the number of neighboring cells of the cell by the average distance between neighboring cells, and multiply by the natural exponential function; The normalized weights are obtained by dividing the initial weight of each cell by the sum of the initial weights of all candidate cells. An initial population is generated using binary encoding, with each chromosome corresponding to a redundant decision scheme and each gene position on the chromosome corresponding to a candidate cell. For each gene position, a uniformly distributed random number between 0 and 1 is generated. If the random number is less than the normalized weight of the corresponding cell, the gene position is set to 1; otherwise, it is set to 0.

9. The method according to claim 8, characterized in that, The optimization of chromosome coding includes: Determine the coverage redundancy and load pressure of a cell; wherein, the coverage redundancy refers to the proportion of grid cells within the cell's coverage area that are occupied by other cells; the load pressure refers to the ratio of the cell's current load to the congestion threshold, which reflects the cell's capacity carrying capacity. The dynamic weight of each candidate cell is calculated based on the coverage redundancy and the load pressure. The dynamic weights of the gene loci in all candidate cells are sorted to determine the quantiles, and the weight value corresponding to the quantile is used as the screening threshold. For each gene locus in a candidate cell, if its dynamic weight is greater than the screening threshold, the gene locus is set to 1; otherwise, it is set to 0.

10. The method according to claim 9, characterized in that, The enhanced design evolution operator includes: Extract the set of spatially adjacent cells, construct a crossover mask, and perform crossover on the parent chromosome according to the mask, preserving the continuity of geographical distribution; Based on the sensitivity to coverage degradation, the mutation probability is calculated, and mutations are preferentially implemented in low-sensitivity, high-load areas to optimize population diversity.

11. The method according to claim 10, characterized in that, The elite retention strategy includes: A three-dimensional elite pool is constructed, stratified by the number of closures, coverage degradation rate, and load balancing degree. Elite individuals with the highest quantiles in the three dimensions are retained in the non-dominated solutions. When the elite pool is full, similar solutions are eliminated based on the crowding distance.

12. A redundant cell identification device, characterized in that, include: The data acquisition unit is used to collect engineering parameter data and MDT data. The engineering parameter data includes the base station name, unique identifier ID, longitude and latitude, coverage type, antenna suspension height, antenna azimuth angle, scene type, transmit power, and minimum access level. The MDT data includes the unique identifier ID of the grid, the unique identifier ID of the sampling point, the cell identifier CI of the primary serving cell, the frequency band of the primary serving cell, the reference received signal power RSRP of the primary serving cell, the signal-to-interference-plus-noise ratio SINR of the primary serving cell, the cell identifier CI of the neighboring cell, the frequency band of the neighboring cell, and the reference received signal power RSRP of the neighboring cell. The feature construction unit is used to construct sampling point-level features, grid-level features, cell-level features, and cluster and region-level features based on the engineering parameter data and MDT data. The grid filtering unit is used to filter out grids with high redundancy potential from the entire grid; the grids to be judged are clustered by the HDBSCAN algorithm, and high redundancy sector regions are obtained by cluster merging and edge cluster expansion; The calculation unit is used to calculate the metadata update results and related statistical indicators after shutting down any cell in the high redundancy sector region, taking the high redundancy sector region as the analysis scope; The output unit is used to construct a multi-objective function that includes coverage quality, capacity, and economic objectives and set constraints; it outputs a list of redundant cells by initializing strategies, optimizing chromosome coding, strengthening the design of evolutionary operators, and implementing elite retention strategies.

13. An electronic device, characterized in that, The electronic device includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the redundant cell identification method according to any one of claims 1-11.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the redundant cell identification method according to any one of claims 1-11.

15. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the redundant cell identification method according to any one of claims 1-11.

Images