A coverage prediction method for planning stations based on no AOA

Through the planning station coverage prediction method based on AOA parameters, using MR and existing network LTE cell industrial parameter data, combined with search progressive algorithm and wireless environment real propagation model, the accuracy and high cost of coverage prediction in the mature stage of LTE network construction are solved. , improves user satisfaction and coverage prediction efficiency.

CN115866623BActive Publication Date: 2024-11-19CHINA TELECOM CORP LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202211514131.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-11-30
Publication Date
2024-11-19
Estimated Expiration
2042-11-30

AI Technical Summary

Technical Problem

In the mature stage of LTE network construction, traditional simulation tools and road testing methods are difficult to accurately predict the coverage of planning stations, resulting in large workload, high cost, low user satisfaction, and the inability to fully tap potential poor coverage areas, resulting in customer churn.

Method used

The planning station coverage prediction method based on AOA parameters is adopted, and the accurate prediction and optimization of the planning station coverage state is achieved through MR and existing network LTE cell industrial parameter data, combined with search progressive algorithm, direction angle determination and wireless environment real propagation model.

Benefits of technology

It improves the efficiency and accuracy of the LTE network planning station solution, reduces costs, avoids user loss, realizes comprehensive coverage prediction in various environmental scenarios, and simplifies the indicator data and calculation rules in actual work.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115866623B_ABST
    Figure CN115866623B_ABST
Patent Text Reader

Abstract

The present invention discloses a coverage prediction method for a planned station based on no AOA, and relates to the technical field of wireless communication data analysis. The present invention is based on two types of data, MR without AOA parameters and existing LTE cell engineering parameters, and realizes coverage prediction and evaluation of the planned station by searching for progressive algorithms, azimuth determination algorithms, and real propagation models of wireless environments. The present invention can realize intelligent planning and coverage prediction of planned stations in the mature stage of network construction, can comprehensively obtain the site and azimuth, station height, and downtilt angle of the planned station, can accurately and reliably predict the wireless coverage status of the planned station, and the prediction process can effectively adapt to various environmental scenarios. At the same time, it is a long-term development LTE network site planning method to ensure the efficiency of network resource investment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of wireless communication data analysis, and in particular to a planning station coverage prediction method based on no AOA (Angle-of-Arrival, signal arrival angle). Background Art

[0002] In the field of mobile communications, with the gradual development and maturity of LTE networks, LTE network coverage has achieved continuous coverage, but LTE network users' demands for network quality have also increased. In order to provide LTE network users with better service quality, it is necessary to explore potential LTE network problems, find LTE network coverage problems faster and more accurately, and propose planning and site construction plans, which have become the focus of current work tasks.

[0003] In the early stage of LTE network construction, simulation was mainly carried out through simulation tools. This method is suitable for large continuous areas. At the current mature stage of LTE network, the disadvantages of this method are: first, the workload is large, and a large amount of parameters and geographical environment information need to be input; second, the accuracy is low, and the simulation results can only show the basic coverage, and cannot fully reflect the actual situation of the LTE network; third, it is not applicable. At the current mature stage of LTE network, this method is no longer applicable for a small number of discontinuous locations.

[0004] In the middle and late stages of LTE construction, planning and site construction plans are proposed mainly through road testing and user complaint information for individual problems. This approach also has some disadvantages: first, the investment cost of personnel, vehicles, test equipment, etc. is high and the efficiency is low, and the potential coverage-poor areas are not fully explored; second, the network planning and site construction method based on user complaint information is in a passive state of LTE network construction and optimization, resulting in low user satisfaction with the LTE network, which can easily cause customer churn. Summary of the invention

[0005] In view of this, the present invention proposes a coverage prediction method for a planned station based on no AOA. The present invention is based on two data sources, namely, MR (Measurement Report) without AOA parameters and engineering parameters of existing LTE cells. The method can comprehensively obtain the site and azimuth angle, station height, downtilt angle, etc. of the planned station, and can accurately and reliably predict the wireless coverage status of the planned station. The prediction process can effectively adapt to various environmental scenarios.

[0006] In order to achieve the above object, the technical solution provided by the present invention is:

[0007] A planning station coverage prediction method based on no AOA includes the following steps:

[0008] Step 1: Extract the MR information of all cells in the city, and divide the MR information into two parts: the first part is the MR information with longitude and latitude, and the second part is the MR information without longitude and latitude; rasterize the first part of the MR information, obtain the RSRP coverage value of each grid, cluster the weak coverage grids based on the search progressive method to generate clusters, obtain the range of the weak coverage area, and calculate the area of ​​the cluster. i , Perimeter i And the center of gravity Focus i (cx',cy'):

[0009] Area i =∑s[i]

[0010] Perimeter i =∑l[i]

[0011] Focus i (cx',cy')=((∑cx[i]*s[i]) / ∑s[i] / 3, (∑cy[i]*s[i]) / ∑s[i] / 3)

[0012] Where s[i] is the area of ​​the i-th triangle that divides the cluster, l[i] is the length of the i-th side of the cluster, and (cx[i], cy[i]) is the centroid coordinates of the i-th triangle that divides the cluster;

[0013] Focus i (cx',cy') is the location of the planned station. i , while focusing on the center of gravity i (cx',cy') is the center of the circle, with the north direction as 0 degrees, calculate its positive and negative angle bandwidth The number of weakly covered grids Grid0 within the angle range of degrees, and calculate the positive and negative Angle in the clockwise direction of 1 degree bandwidth The number of weakly covered grids in the angle range of degrees is Grid1, and so on. Finally, the positive and negative Angles in the clockwise direction of 359 degrees are calculated. bandwidth Number of weakly covered grids within the angle range Grid 359 ; Sort by the number of weakly covered grids, and select the first K grids with the largest number of weakly covered grids n The corresponding angle Bearing n Azimuth as the direction angle of the planned station area i-n , the angle between the direction angles of the planned station cells is greater than the minimum threshold angle Angle min ;

[0014] Based on the planning station location iTake the center of the circle as the center, obtain the surrounding adjacent base stations, and extract the location of the corresponding base stations whose coverage direction is not facing the planned station location. i The cell and its MR information; obtain the engineering parameter information of the surrounding adjacent cells, and use the average station height and average downtilt angle of the surrounding adjacent cells as the station height of the planned station i Downtilt i-n , thus obtaining the initial plan for the planning station; among them:

[0015]

[0016]

[0017] In the formula, Hight t and Downtilt t Identify the site height and downtilt angle of the tth adjacent cell respectively;

[0018] Step 2: Based on the MR information of the neighboring cells around the planning station, the first part of the MR information is cleaned and transformed to obtain two parts of feature vectors. The first part is the original MR feature vector, and the second part is the feature vector after cleaning and deformation based on the original feature vector. The importance of the feature vector in the model analysis is analyzed by using the pipeline model Pipeline. According to the score of the feature vector importance, the top N most important feature vectors are selected as the basis for modeling.

[0019] The selected feature vectors are normalized to reduce the influence of individual feature vectors on modeling. Then, the model is trained based on the random forest method. The grid search method GridSearchCV is combined to select the optimal combination of the three important parameters n_estimators, max_depth, and min_samples_split in the random forest method, so as to finally generate the MR direction angle prediction model.

[0020] The second part of MR information is input into the MR direction angle prediction model to obtain the direction angle Downtilt of the second part of MR information. j , combined with the corrected TADV time advance in the MR information, the MR longitude and latitude (Lon j ,Lat j ), thereby achieving full MR coverage of the existing network cells; MR longitude and latitude (Lon j ,Lat j )The calculation function is as follows:

[0021] (Lon j ,Lat j )=getLocation(lon cell ,latcell ,Direction j ,TADV j )

[0022] Among them, lon cell ,lat cell ,Direction j ,TADV j They are respectively the longitude of the cell, the latitude of the cell, the predicted direction angle of the jth MR without AOA parameters and excluding longitude and latitude, and the correction value of TADV reported by the jth MR without AOA parameters and excluding longitude and latitude;

[0023] Step 3: Based on the full MR of the surrounding adjacent cells, the actual wireless propagation model of the scene environment around the planning station is obtained through support vector machine and cross-validation method:

[0024] Best_model=getmodel(gpid,df)

[0025] Among them, gpid is the label of the cluster, df is the set of all MRs of the surrounding adjacent cells;

[0026] Combine the antenna horizontal pattern and antenna vertical pattern, and calculate the coverage radius Distance of the planned station. i 、Hight i 、Azimuth i-n , Downtilt i-n , predict the grid RSRP coverage within the coverage radius of the planned station; RSRP calculation function is: RSRP_value=cal(gpid,Best_model,center_lon,center_lat,azmuth s,tilts,lon_lst,lat_lst,height)

[0027] Among them, gpid is the label of the cluster, Best_model is the real wireless propagation model of the surrounding environment of the planning station, center_lon is the longitude of the planning station, center_lat is the latitude of the planning station, azmuths is the direction angle set of the planning station cell, tilts is the downtilt angle set of the planning station cell, lon_lst is the grid longitude within the coverage range of the planning station, lat_lst is the grid latitude within the coverage range of the planning station, and height is the height of the planning station;

[0028] Adjust the height t , Downtilt i-n, and then merge it with the full MR of the surrounding adjacent existing LTE cells, take the cell corresponding to the maximum RSRP value in the same grid as the main coverage cell, and obtain the coverage prediction result after the planned station is opened.

[0029] It can be seen from the above description that the beneficial effects of the technical solution of the present invention are:

[0030] 1. Solve the problem of the inapplicability of using traditional simulation tools to predict the coverage of planned stations in the mature stage of LTE network construction. In the current mature stage of LTE network, this method is no longer applicable for a small number of discontinuous locations. First, the workload is large and a large amount of parameters and geographical environment information needs to be input. Second, the accuracy is low. The simulation results can only show the basic coverage and cannot fully reflect the actual situation of the LTE network.

[0031] 2. It not only improves the efficiency of LTE network planning station solution output, but also solves the problem of high cost of personnel, vehicles, test equipment, etc. in exploring LTE network coverage through road testing during the mature stage of LTE network construction, and the problem of incomplete exploration of potential areas with poor coverage.

[0032] 3. Change from passive to active, avoid user churn, and solve the problem of passive network construction through user complaint information during the mature stage of LTE network construction, which leads to low user satisfaction with LTE network and easily causes customer churn.

[0033] 4. Solve the problem of low accuracy of calculating MR position by conventional triangulation and fingerprint positioning methods when there are no AOA parameters in MR and little information about measured neighboring areas. Achieve accurate, reliable and comprehensive prediction of wireless coverage of planning stations in various environmental scenarios.

[0034] 5. The algorithm of the present invention has low complexity and fast running speed. The present invention is based on only two types of data, namely, MR without AOA parameters and existing LTE cell engineering parameters. By searching for progressive algorithms, directional angle determination algorithms, and real propagation models of wireless environments, the coverage prediction and evaluation of planned stations is achieved.

[0035] 6. Good applicability. The present invention only needs to combine the above two types of data and can realize the output of the planning station plan by calling the relevant algorithm module through the program. It does not require too much indicator data and complex calculation rules, and can be better used in practical work.

[0036] In summary, the present invention is based on two types of data, MR without AOA parameters and existing LTE cell engineering parameters, and realizes coverage prediction and evaluation of planned stations by searching progressive algorithms, directional angle determination algorithms, and real propagation models of wireless environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to describe this patent more clearly, several drawings are provided below, which are intended to assist in explaining the background technology, technical principles and / or certain specific implementation plans of this patent.

[0038] Figure 1 A basic flow chart for implementing the method of the present invention in an embodiment of the present invention.

[0039] Figure 2 Basic data preprocessing process in an embodiment of the present invention.

[0040] Figure 3 The embodiment of the present invention searches for and gradually mines a potential coverage problem process.

[0041] Figure 4 The initial plan process of the planning station in the embodiment of the present invention.

[0042] Figure 5 The embodiment of the present invention provides a coverage prediction process based on a wireless propagation model in an actual environment.

[0043] Figure 6 An example of calculating the centroid and area of ​​a cluster in an embodiment of the present invention.

[0044] Figure 7 An example of the effect of the initial plan of the planning station in the embodiment of the present invention.

[0045] Figure 8 Feature selection in the embodiment of the present invention.

[0046] Fig. 9 An example of the effect of MR direction angle prediction deviation in an embodiment of the present invention.

[0047] Fig.10 An example of a real environment wireless propagation model around a planning station in an embodiment of the present invention.

[0048] Fig.11 An example of the coverage prediction effect of planning stations in an embodiment of the present invention.

[0049] Fig.12 An example of the overall grid coverage effect of the planned station and the existing network cells in the embodiment of the present invention. DETAILED DESCRIPTION

[0050] In order to facilitate technical personnel in this field to understand the technical solution of this patent, and at the same time, to make the technical purpose, technical solution and beneficial effects of this patent clearer and to fully support the scope of protection of the claims, the technical solution of this patent is further and more detailedly explained in the form of specific cases.

[0051] The following will be combined with the drawings of the embodiments of the present invention to fully and clearly describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all embodiments obtained by ordinary technicians in this field without making creative work belong to the protection scope of the present invention.

[0052] A planning station coverage prediction method based on no AOA has the following four main purposes:

[0053] Objective 1: To solve the problem that the use of simulation tools for planning station coverage prediction is not applicable during the mature stage of LTE network construction.

[0054] The second purpose is to solve the problem of using road testing to explore LTE network coverage during the mature stage of LTE network construction, which is costly and inefficient in terms of personnel, vehicles, and test equipment, and the incomplete exploration of potential areas with poor coverage.

[0055] The third purpose is to solve the problem that in the mature stage of LTE network construction, network construction and optimization are passively carried out through user complaint information, resulting in low user satisfaction with the LTE network, which easily causes customer churn.

[0056] Objective 4: To solve the problem of low accuracy of calculating MR position by conventional triangulation and fingerprint positioning methods when there are no AOA parameters in MR and less neighboring area measurement information. Through the algorithm of the present invention, MR without AOA parameters can not only realize fast and efficient mining of potential coverage problems of LTE network and output planning schemes, but also realize accurate, reliable and comprehensive prediction of wireless coverage of planning stations in various environmental scenarios, ensuring the efficiency of network resource investment.

[0057] This method first divides MR into two parts. The first part is MR with longitude and latitude without AOA parameters, and the second part is MR without longitude and latitude without AOA parameters. This part cannot be presented geographically intuitively and cannot be used for regional coverage assessment directly. i And the planned area direction angle Azimuth i-n , Coverage radius Distance i, where i represents the location of the i-th planned station, and n represents the n-th cell of the i-th planned station. At the same time, by cleaning and converting the first part of the MR without AOA parameters and containing longitude and latitude, two feature vectors are obtained. The first part is the original feature vector of MR, and the second part is the feature vector after cleaning and deformation based on the original feature vector. The pipeline model Pipeline is used to analyze the importance of features in the model. According to the score of feature importance, the top N most important features are selected as the basis for modeling. Then, the model is trained based on the random forest algorithm. Combined with the grid search algorithm GridSearchCV, the optimal combination of the three important parameters n_estimators, max_depth, and min_samples_split in the random forest algorithm is selected, so as to finally generate the MR direction angle prediction model. The second part of the MR without longitude and latitude is input into the prediction model to obtain the direction angle Direction of MR j , combined with the modified TADV to calculate the MR longitude and latitude (Lon j , Lat j ), so as to obtain the comprehensive coverage distribution of the existing LTE cells, facilitate intuitive geographical presentation of cell coverage, better coverage analysis, and output the coverage prediction of each cell in the planning station, so as to achieve a comprehensive evaluation of the planning station, where j represents the jth MR without AOA parameters and longitude and latitude under the existing LTE cell, and the corrected TADV is the TADV reported by the MR mr The conversion expression is as follows:

[0058] TADV=(TADV mr +1)*78+random.randint(-39,39)

[0059] The method specifically comprises the following steps:

[0060] Step 1: Extract the MR information of all cells in the city, taking the city as the unit. Since the collected MR does not have AOA parameters and the measured neighboring area information is less, it is impossible to accurately locate the MR through the normal triangulation positioning algorithm and fingerprint positioning algorithm. In order to better and more accurately locate the MR, it is necessary to establish a model to predict the position of the MR without longitude and latitude. Divide the MR into two parts, the first part is the MR with longitude and latitude, and the second part is the MR without longitude and latitude. By rasterizing the first part of the MR, the RSRP coverage value of each grid is obtained, and the weak coverage grids are clustered based on the search progressive algorithm to generate clusters, and the range of the weak coverage area is obtained, and the area of ​​the cluster is calculated at the same time. i , Perimeter i And the center of gravity Focus i (cx',cy'). Where Areai 、Perimeter i 、Focus i The calculation formulas for (cx',cy') are:

[0061] Area i =∑s[i]

[0062] Perimeter i =∑l[i]

[0063] Focus i (cx',cy')=((∑cx[i]*s[i]) / ∑s[i] / 3, (∑cy[i]*s[i]) / ∑s[i] / 3)

[0064] Where s[i] is the area of ​​the i-th triangle that divides the cluster, l[i] is the length of the i-th side of the cluster, for example l[1] is the length from vertex A1 to vertex A2, l[2] is the length from vertex A2 to vertex A3, and so on. (cx[i],cy[i]) are the centroid coordinates of the i-th triangle that divides the cluster.

[0065] Focus i (cx',cy') is the location of the planned station. i , while focusing on the center of gravity i (cx',cy') is the center of the circle, with the north direction as 0 degrees, calculate its positive and negative angle bandwidth The number of weakly covered grids within the angle range of degrees is Grid0, and the positive and negative Angles in the 1 degree direction are calculated clockwise. bandwidth The number of weakly covered grids in the angle range of degrees is Grid1, and so on, and finally the positive and negative Angle in the 359 degree direction is calculated. bandwidth Number of weakly covered grids within the angle range Grid 359 . Sort by the number of weakly covered grids and select the first K grids with the largest number of weakly covered grids. n The corresponding angle Bearing n Azimuth as the direction angle of the planned station area i-n To ensure the rationality of the planning scheme, the angle between the direction angles of the planned station cells is greater than the minimum threshold angle Angle min .

[0066] Planning station location Location i Take the center of the circle as the center, obtain the surrounding adjacent base stations, and extract the location of the corresponding base stations whose coverage direction is not facing the planned station location. iObtain the engineering parameter information of the surrounding adjacent cells, and use the average station height and average downtilt angle of the surrounding adjacent cells as the station height of the planned station. i , Downtilt i-n , thus obtaining the initial plan for the planning station. The calculation formula is as follows:

[0067]

[0068]

[0069] Among them Hight t and Downtilt t Identify the site height and downtilt angle of the tth adjacent cell respectively.

[0070] Step 2: Based on the MR of the neighboring cells around the planning station, two feature vectors are obtained by cleaning and transforming the first part of the MR. The first part is the original MR feature vector, and the second part is the feature vector after cleaning and deformation based on the original feature vector. The importance of the feature in the model analysis is analyzed by using the pipeline model Pipeline. According to the score of the feature importance, the top N most important features are selected as the basis for modeling.

[0071] To ensure the accuracy of the prediction model, the features need to be normalized before modeling to reduce the impact of individual features on modeling. Then, the model is trained based on the random forest algorithm. The grid search algorithm GridSearchCV is combined to select the optimal combination of the three important parameters n_estimators, max_depth, and min_samples_split in the random forest algorithm, so as to finally generate the MR direction angle prediction model. The second part MR is input into the prediction model to obtain the direction angle Direction of the second part MR. j , combined with the corrected TADV value in MR, the MR longitude and latitude (Lon j , Lat j ), thus achieving full MR coverage of the existing network cells. MR longitude and latitude (Lon j , Lat j )The calculation function is as follows:

[0072] (Lon j , Lat j )=getLocation(lon cell ,lat cell ,Direction j ,TADV j )

[0073] where loncell ,lat cell ,Direction j ,TADV j They are respectively the longitude of the cell, the latitude of the cell, the predicted direction angle of the j-th MR without AOA parameters and excluding longitude and latitude, and the correction value of TADV reported by the j-th MR without AOA parameters and excluding longitude and latitude.

[0074] Step 3: Based on the full MR of the surrounding adjacent cells, the actual wireless propagation model of the scene environment around the planning station is obtained through the support vector machine algorithm and the cross-validation algorithm, avoiding the problem that the traditional propagation model is not suitable for the specific environment. The effect is as follows: Fig.10 As shown, the function expression is as follows:

[0075] Best_model=getmodel(gpid,df)

[0076] Among them, gpid is the label of the cluster, and df is the set of all MRs of the surrounding adjacent cells.

[0077] At the same time, combined with the antenna horizontal radiation pattern and the antenna vertical radiation pattern, according to the planned station coverage radius Distance i 、Hight i 、Azimuth i-n , Downtilt i-n , predict the grid RSRP coverage within the coverage radius of the planned station, and the RSRP calculation function expression is:

[0078] RSRP_value=cal(gpid,Best_model,center_lon,center_lat,azmuths,tilts,lon_lst,lat_lst,height)

[0079] Among them, gpid is the label of the cluster, Best_model is the real wireless propagation model of the surrounding environment of the planning station, center_lon is the longitude of the planning station, center_lat is the latitude of the planning station, azmuths is the direction angle set of the planning station cell, tilts is the downtilt angle set of the planning station cell, lon_lst is the grid longitude within the coverage range of the planning station, lat_lst is the grid latitude within the coverage range of the planning station, and height is the height of the planning station.

[0080] To ensure that the coverage of the planned station can achieve the expected effect, the station height can be adjusted. i , Downtilt i-nThen, the full MR of the surrounding adjacent existing LTE cells is merged, and the cell corresponding to the maximum RSRP value in the same grid is taken as the main coverage cell, so as to finally obtain the coverage prediction result after the planned station is opened.

[0081] Here is a more specific example:

[0082] like Figure 1 As shown, a planning station coverage prediction method based on no AOA is implemented based on the following modules:

[0083] The basic data acquisition and preprocessing module 101 is used to obtain the full amount of MR and LTE network cell engineering parameter information of the existing network cell without AOA parameters from the MR server and the LTE network cell engineering parameter network management, obtain the relevant key fields after parsing and cleaning, and import them into the Oracle database

[0084] The search progressive method to mine potential coverage problems module 102 rasterizes the MR containing longitude and latitude in the first part, obtains the RSRP coverage value of each grid, clusters the weak coverage grids based on the search progressive algorithm, obtains the range of the weak coverage area, and calculates the area of ​​the cluster. i , Perimeter i And the center of gravity Focus i (cx',cy'), where area and perimeter are used to determine the construction priority of different planning station schemes.

[0085] Planning station initial plan output module 103, with the center of gravity of the weak coverage cluster Focus i (cx',cy') is the location of the planned station. i , to plan the station location i The distribution of weak coverage grids in the cluster is calculated, and the first K directions where weak coverage grids are concentrated are selected as the direction angle Azimuth of the planned station cell. i-n , and ensure that the angle between the direction angles of the planned station cells is greater than the minimum threshold angle Angle min . At the same time, the planning station location Location i Take the center of the circle as the center, obtain the surrounding adjacent base stations, and extract the location of the corresponding base stations whose coverage direction is not facing the planned station location. i Combined with the engineering parameter information of the surrounding cells, the average station height and average downtilt angle of the surrounding adjacent cells are used as the station height of the planned station. i , Downtilt i-n , thus obtaining the initial plan of the planning station.

[0086] The module 104 predicts the coverage of the planned station based on the actual environment wireless propagation model. Based on the MR of the adjacent cells around the planned station, the MR direction angle prediction model is obtained by training and modeling the first part of the MR. The second part of the MR is input into the prediction model to obtain the direction angle Direction of the second part of the MR. j , combined with the corrected TADV value, the MR latitude and longitude (Lon j , Lat j ), thus achieving full MR coverage of the existing cell. Based on the full MR of the surrounding adjacent cells, the actual wireless propagation model of the scene environment around the planned station can be obtained, avoiding the problem that the traditional propagation model is not applicable to the specific environment. At the same time, combined with the antenna horizontal radiation pattern and the antenna vertical radiation pattern, according to the coverage radius Distance of the planned station i 、Hight i 、Azimuth i-n Downtilt i-n , predict the grid RSRP coverage within the coverage radius of the planned station. To ensure that the coverage of the planned station can achieve the expected effect, the station height can be adjusted. i Downtilt i-n Then, the full MR of the surrounding adjacent existing LTE cells is merged, and the cell corresponding to the maximum RSRP value in the same grid is taken as the main coverage cell, so as to finally obtain the coverage prediction result after the planned station is opened.

[0087] In this embodiment, Figure 2 As shown, basic data acquisition and preprocessing 101 includes two parts: MR without AOA parameters and LTE network cell working parameters.

[0088] Step 10101 is used to obtain the full amount of MR and LTE network cell engineering parameter information of the existing cell without AOA parameters from the MR server and the LTE network cell engineering parameter network management respectively.

[0089] Step 10102 is used to parse the MR and LTE network cell engineering parameter data without AOA parameters, and import them into the Oracle database after cleaning.

[0090] The main key fields of the MR without AOA parameters after analysis and cleaning are shown in the following table.

[0091]

[0092]

[0093] The key fields of the LTE network cell working parameters after analysis and cleaning are as follows:

[0094] Fields type illustrate Example CITY NVARCHAR2(10) City Cangzhou ENB INT Base station identification 95874 CELLID INT Cell ID 50 BAND INT Frequency band 1800 DLEARFCN INT Downlink frequency 1850 PCI INT Physical cell identification 61 LONGITUDE FLOAT longitude 116.4362 LATITUDE FLOAT latitude 38.03946 AZIMUTH FLOAT Direction 150 HEIGHT FLOAT Standing high 35 MACHINE_TITLE FLOAT Mechanical down tilt 3 ELECTRICAL_TILTE FLOAT Electronic down tilt 4 COVER_TYPE NVARCHAR2(5) Base station type Macro Station

[0095] In this embodiment, Figure 3 As shown, the search progressive method to mine potential coverage problem module 102, by rasterizing the first part of the MR containing longitude and latitude, obtains the RSRP coverage value of each grid, clusters the weak coverage grids based on the search progressive algorithm to generate clusters, obtains the range of the weak coverage area, and calculates the area of ​​the cluster Area i , Perimeter i And the center of gravity Focus i (cx',cy'), including the following specific steps:

[0096] Step 10201, based on the first part MR without AOA parameters, calculate the grid RSRP mean, if the grid RSRP mean is less than the threshold RSRP THR , it is marked as a weakly covered grid, and through loop iteration, the set of all weakly covered grids is finally obtained.

[0097] Step 10202, set the search tag SID and cluster tag CID for each grid in the weak coverage grid set, and initialize them to 0 to facilitate subsequent loop iteration judgment. The search tag SID indicates whether the surrounding weak coverage grids have been searched with the grid as the center. Its value range is 0 or 1. 0 means that the surrounding weak coverage grids have not been searched with the grid as the center, and 1 means that the surrounding weak coverage grids have been searched with the grid as the center. The cluster tag CID is used to identify a unique value of the cluster. Its value is expressed as a string of milliseconds of the current time, in the format of yyyymmddHHmissffffff, and an example is 20220318094216101631.

[0098] Step 10203, based on the result of the above step 10202, process each weak coverage grid in turn, read its retrieval tag SID and determine whether the surrounding weak coverage grids have been searched with the grid as the center. If the retrieval tag SID is 0, search the surrounding weak coverage grids through step 10204. If the retrieval tag SID is 1, read the next grid for judgment.

[0099] Step 10204: If the search tag ID read in step 10203 is 0, first modify the value to 1, and set the value of the cluster tag CID to the current time string accurate to milliseconds, in a format such as yyyymmddHHmissffffff, and an example such as 20220318094216101631. The weak coverage grid position is used as the search center, and the threshold R is used as the radius to search for weak coverage grids within its range, and the CID of the searched weak coverage grid is set to the same tag as it.

[0100] Step 10205, determine whether the number of grid sets in the cluster increases, the purpose is to determine whether the cluster converges, if the number of grids in the cluster increases, it means that the cluster has not converged, there are still reachable grids that have not been divided into the cluster, at this time, further iterative processing is required through step 10206. If there are no new grids in the cluster, then exit this iteration, and read the next new grid data from the weak coverage grid set table through step 10203 for processing.

[0101] Step 10206, taking the reachable grid searched in step 10204 as the center, perform iterative loop processing according to steps 10203 to 10205.

[0102] Step 10207: finally all weak coverage clusters are obtained through the above steps. Examples of weak coverage clusters are as follows: Figure 7 As shown in the figure, the closed area is the boundary of the weak coverage cluster, and the points in the closed area are the coverage grid. For each weak coverage cluster, calculate its area Area i , Perimeter i And the center of gravity Focus i (cx',cy'), providing an important reference for network construction.

[0103] like Figure 6 As shown, assume that the boundary of the weak covering cluster has k vertices, and the coordinates are sorted in counterclockwise direction, namely A1(x1,y1), A2(x2,y2), A3(x3,y3)…Ak(xk,yk). Taking A1(x1,y1) as the reference, the line segments of all other non-adjacent vertices are used to divide the polygon (k edges) into k-2 triangles, and the area and centroid of each triangle are calculated.

[0104] Suppose the centroid of one of the triangles is G(cx, cy), and the vertex coordinates are A1(x1, y1), A2(x2, y2), and A3(x3, y3), then the centroid is:

[0105] cx=(x1+x2+x3) / 3

[0106] cy=(y1+y2+y3) / 3

[0107] The area is:

[0108] S=((x2-x1)*(y3-y1)-(x3-x1)*(y2-y1)) / 2

[0109] The center of mass of the cluster is Focus i (cx',cy') is:

[0110] cx'=(∑cx[i]*s[i]) / ∑s[i] / 3

[0111] cy'=(∑cy[i]*s[i]) / ∑s[i] / 3

[0112] Where (cx[i], cy[i]), s[i] are the centroid coordinates and area of ​​the i-th triangle divided respectively.

[0113] The area of ​​the cluster is i They are:

[0114] Area i =∑s[i]

[0115] The perimeter of the cluster is i for:

[0116] Perimeter i =∑l[i]

[0117] Where l[i] is the length of the i-th edge of the cluster, for example, l[1] is the length from vertex A1 to vertex A2, l[2] is the length from vertex A2 to vertex A3, and so on.

[0118] In this embodiment, Figure 4 As shown, the planning station initial plan output module 103 includes the following specific steps:

[0119] Step 10301, Focus on the centroid of the weakly covered cluster i (cx',cy') is the location of the planned station. i , to plan the station location i The distribution of weak coverage grids in the cluster is calculated, and the first K directions where weak coverage grids are concentrated are selected as the direction angle Azimuth of the planned station cell. i-n , and ensure that the angle between the direction angles of the planned station cells is greater than the minimum threshold angle Angle min .

[0120] Based on the planning station location i As the center of the circle, with due north as 0 degrees, calculate its positive and negative Angle bandwidth The number of weakly covered grids within the angle range of degrees is Grid0, and the positive and negative Angles in the 1 degree direction are calculated clockwise. bandwidth The number of weakly covered grids in the angle range of degrees is Grid1, and so on, and finally the positive and negative Angle in the 359 degree direction is calculated. bandwidth Number of weakly covered grids within the angle range Grid 359. Sort by the number of weak coverage grids, select the one with the largest number of weak coverage grids as the preferred antenna direction angle Bearing1, then select the direction angle with the second largest number of weak coverage grids, compare it with the preferred antenna direction angle Bearing1, and determine whether the angle between the two is greater than the minimum threshold angle Angle min If it is greater than, the second direction angle of the weak coverage grid number is used as the second determined antenna direction angle Bearing2, and so on, to calculate the nth determined antenna direction angle Bearing n Finally, the direction angles are numbered in ascending order to determine the first, second, third, etc. cell direction angles of the planned station.

[0121] Step 10302, planning station location Location i Take the center of the circle as the center, obtain the surrounding adjacent base stations, and extract the location of the corresponding base stations whose coverage direction is not facing the planned station location. i Combined with the engineering parameter information of the surrounding cells, the average station height and average downtilt angle of the surrounding adjacent cells are used as the station height of the planned station. i Downtilt i-n , thus obtaining the initial plan of the planning station, the calculation formula is as follows:

[0122]

[0123]

[0124] Among them Hight t and Downtilt t Identify the site height and downtilt angle of the tth adjacent cell respectively.

[0125] In this embodiment, Figure 5 As shown, the planning station initial plan output module 104 includes the following specific steps:

[0126] Step 10401, based on the MR of the adjacent cell obtained in step 10302, obtain the effective MR of the adjacent cell through preprocessing, and the basic process is as follows.

[0127] For each cell, calculate the cumulative proportion of its MR in each TADV F(X). If the cumulative proportion of TADV F(X) is greater than the threshold CDF THR , then the corresponding TADV value is used as the coverage radius of the existing network cell It represents the coverage radius of the jth adjacent cell corresponding to the i-th planned station, and the cumulative proportion F(X) is expressed as follows:

[0128]

[0129] Where f(x k ) is the MR proportion when TADV=k, the value range of k is an integer greater than or equal to zero, and its expression is:

[0130]

[0131] Taking the MR of cell 86101-50 as an example, the following table shows the proportion of MRs corresponding to each TADV value f(x k ) and the cumulative proportion F(X), set the cumulative proportion threshold CDF THR If the coverage radius of the cell is 99%, is 7.

[0132]

[0133] By radius The MR in the range is taken as the effective MR and divided into two parts. The first part is the MR with longitude and latitude without AOA parameters, and the second part is the MR without longitude and latitude without AOA parameters. The MR direction angle prediction model is obtained by training and modeling the MR with longitude and latitude without AOA parameters in the first part, and then the MR without longitude and latitude in the second part is input into the MR direction angle prediction model to obtain the predicted direction angle of each MR in the second part. The modeling process includes the following key parts:

[0134] a) Calculate the direction angle of the first part of the MR position relative to the cell to which it belongs, and the calculation function is:

[0135] dir=getDegree(lon cell ,lat cell ,lon mr ,lat mr )

[0136] where lon cell ,lat cell ,lon mr ,lat mr They are respectively the longitude of the cell, the latitude of the cell, the longitude of the MR, and the latitude of the MR.

[0137] b) Feature vector screening and processing, the screened features are used as independent variables for model training and establishment, where the features include two major features, the first part is the original feature vector, and the second part is the feature vector after cleaning and deformation processing based on the original feature vector, which can better count the distribution of MR in various combinations, which is conducive to improving the effect of machine learning modeling. The first part of the feature vectors mainly includes serving cell reference signal power (LTESCRSRP), neighboring cell reference signal power (LTENCRSRP), serving cell reference signal quality (LTESCRSRQ), neighboring cell reference signal quality (LTENCRSRQ), serving cell frequency number (LTESCEARFCN), serving cell PCI (LTESCPCI), neighboring cell frequency number (LTENCEARFCN), neighboring cell PCI (LTENCPCI), timing advance (LTESCTADV), UE transmit power margin (LTESCPHR), uplink signal-to-noise ratio (LTESCSINRUL), etc. The characteristic vectors of the second part are the number of MRs (LTESCRSRP_Q_TA_CNT) with the same time advance, serving cell reference signal power, and serving cell reference signal quality, the number of MRs (LTESCRSRP_TA_CNT) with the same time advance and serving cell reference signal power, the number of MRs (LTESCRSRP_TA_CNT) with the same time advance and serving cell reference signal quality (LTESCRSRQ_TA_CNT), the number of MRs with the same serving cell reference signal power (LTESCRSRP_CNT), the number of MRs with the same serving cell reference signal quality (LTESCRSRQ_CNT), etc.

[0138] Feature selection is based on feature engineering and selects the top N most important features for machine learning modeling according to the importance of features in model analysis. In this example, the random forest algorithm RandomForestRegressor is used as an example for modeling training. The example is as follows: rf = RandomForestRegressor (n_estimators = 100, ax_depth = None)

[0139] Then, the pipeline model Pipeline is used to analyze and calculate the feature importance scores. The features are sorted in descending order and the top N important features are selected. The calculation method is as follows, where importance is a list of scores for each feature importance. The corresponding intuitive presentation of feature importance is as follows: Figure 8 As shown:

[0140] rf_pipe=Pipeline([('imputer',

[0141] SimpleImputer(strategy='median')),('standardize',

[0142] StandardScaler()),('rf',rf)])

[0143] rf_pipe.fit(X_train,y_train)

[0144] importance=rf.feature_importances

[0145] c) After normalizing the feature vector, the optimal random forest model is selected based on the grid search algorithm to establish the MR direction angle prediction model. Fig. 9 It can be seen that the proportion of predicted angle errors between plus or minus 5 is more than 70%.

[0146] After the features are digitized, their values ​​are different, so the distance of sample points in the feature space will be dominated by individual feature values, and will be less affected by other features, which will affect the accuracy of modeling prediction. Therefore, the features need to be normalized before modeling to reduce the impact of individual features on modeling. In this example, the mean and standard deviation of the feature data are used for data processing. The processed data conforms to the normal distribution, that is, μ = 0, σ = 1, and the conversion function is:

[0147]

[0148] Where x is the value of each feature, μ is the feature mean, σ is the feature standard deviation, and x * is the normalized value of the feature.

[0149] Step 10402: Input the second part of MR without AOA parameters and without longitude and latitude into the prediction model to obtain the direction angle Direction of the second part of MR j , whose value range is [0,360). The model training uses grid search GridSearchCV to calculate and analyze the parameters of the random forest model, and selects the optimal combination parameters as the parameters of the random forest model. In this example, its expression is as follows:

[0150] clf=GridSearchCV(RandomForestRegressor(random_state=720),params,cv=5,scoring=score)

[0151] Where params is the following 96 parameter combinations:

[0152] params={

[0153] 'n_estimators':range(500,2000,500),

[0154] 'max_depth':[None,100,200,300],

[0155] 'min_samples_split':[2,4,6,8,10,12,14,16]

[0156] }

[0157] In this example, the optimal combination parameters obtained through the above analysis are:

[0158] {'max_depth':None,

[0159] 'min_samples_split':8,

[0160] 'n_estimators':500

[0161] }

[0162] After the optimal combination parameters are determined, the MR direction angle prediction model can be established, and then the second part of MR is input into the prediction model to obtain the predicted direction angle Direction of the second part of MR j .

[0163] Step 10403, based on the result outputted in step 10402, combined with the corrected TADV value, finally obtain the MR longitude and latitude (Lon j , Lat j ), thus achieving the coverage presentation of all MRs in the existing network cells. The calculation function is:

[0164] (Lon j , Lat j )=getLocation(lon cell ,lat cell ,Direction j ,TADV j )

[0165] where lon cell ,lat cell ,Direction j ,TADV j They are respectively the longitude of the cell, the latitude of the cell, the predicted direction angle of the j-th MR without AOA parameters and excluding longitude and latitude, and the correction value of TADV reported by the j-th MR without AOA parameters and excluding longitude and latitude.

[0166] Step 10404, based on the full MR of the surrounding adjacent cells, the actual wireless propagation model of the scene environment around the planning station is obtained through the support vector machine algorithm and the cross-validation algorithm, so as to avoid the problem that the traditional propagation model is not suitable for the specific environment. The effect is as follows: Fig.10 As shown, the function expression is as follows:

[0167] Best_model=getmodel(gpid,df)

[0168] Among them, gpid is the label of the cluster, and df is the set of all MRs of the surrounding adjacent cells.

[0169] Step 10405: Based on the actual wireless propagation model obtained in step 10404, combined with the antenna horizontal radiation pattern and the antenna vertical radiation pattern, the planned station coverage radius Distance is calculated. i 、Hight i 、Azimuth i-n Downtilt i-n , predict the grid RSRP coverage within the coverage radius of the planned station, the example effect is as follows Fig.11 As shown. The RSRP calculation function expression is:

[0170] RSRP_value=cal(gpid,Best_model,center_lon,center_lat,azmuths,tilts,lon_lst,lat_lst,height)

[0171] Wherein gpid is the label of the cluster, Best_model is the real wireless propagation model of the surrounding environment of the planning station obtained in step 10404, center_lon is the longitude of the planning station, center_lat is the latitude of the planning station, azmuths is the direction angle set of the planning station cell, tilts is the downtilt angle set of the planning station cell, lon_lst is the grid longitude within the coverage range of the planning station, lat_lst is the grid latitude within the coverage range of the planning station, and height is the height of the planning station.

[0172] Step 10406: Based on the initial plan of the planned station, determine whether the coverage prediction of the planned station can achieve the expected effect. If it does not meet the expectations, adjust the station height. i Downtilt i-n .

[0173] Step 10407, based on the coverage prediction result of the planned station obtained in step 10406, the full amount of MR of the surrounding adjacent existing LTE cells is merged, and the cell corresponding to the maximum RSRP value in the same grid is taken as the main coverage cell, so as to finally obtain the coverage prediction result after the planned station is opened, such as Fig.12 shown.

[0174] This example not only realizes the accurate planning and filling of blind spots in the planned sites during the mature stage of LTE network construction, but also realizes the evaluation of the surrounding wireless environment after the planned sites are built. Compared with the "Multi-dimensional Assessment and Accurate Construction Investment Evaluation Method Based on Coverage and Value Association" and the "Site Planning Method, Device, Electronic Equipment and Storage Medium Based on Clustering", the planning scheme is more complete and has a wireless environment evaluation after the planned sites are built, filling the gap in the industry.

[0175] 1. Technical points of the invention that the inventor believes should be protected

[0176] The technical points of the present invention include but are not limited to the following points:

[0177] Technical point 1: Search for algorithms that progressively mine potential coverage issues;

[0178] Technical point 2: Algorithm for MR positioning based on MR without AOA parameters;

[0179] Technical point three: the analysis and calculation algorithm of all engineering parameter information of the planning station;

[0180] Technical point 4: Planning station coverage prediction algorithm;

[0181] In summary, the present invention provides a coverage prediction method for planning stations based on no AOA, which can realize intelligent planning and coverage prediction of planning stations in the mature stage of network construction, can comprehensively obtain the site and azimuth angle, station height, and downtilt angle of the planning station, can accurately and reliably predict the wireless coverage status of the planning station, and the prediction process can effectively adapt to various environmental scenarios. At the same time, it is a long-term development LTE network site planning method to ensure the efficiency of network resource investment.

[0182] It should be understood that the above description of the specific implementation methods of this patent is only an exemplary description listed to facilitate the understanding of the patent solution by ordinary technicians in this field, and does not imply that the protection scope of this patent is limited to these individual examples. Ordinary technicians in this field can fully understand the technical solution of this patent and, without any creative work, obtain more specific implementation methods by combining technical features, replacing some technical features, adding more technical features, etc. to the examples listed in this patent. All these specific implementation methods are within the scope of the claims of this patent. Therefore, these new specific implementation methods should also be within the scope of protection of this patent.

Claims

1. A planning station coverage prediction method based on no AOA, characterized in that: The following steps are involved: Step 1: Extract the MR information of all cells in the city, and divide the MR information into two parts: the first part is the MR information with longitude and latitude, and the second part is the MR information without longitude and latitude; rasterize the first part of the MR information, obtain the RSRP coverage value of each grid, cluster the weak coverage grids based on the search progressive method to generate clusters, obtain the range of the weak coverage area, and calculate the area of ​​the cluster. i , Perimeter i And the center of gravity Focus i (cx',cy'): Area i =∑s[i] Perimeter i =∑l[i] Focus i (cx',cy')=((∑cx[i]*s[i]) / ∑s[i] / 3,(∑cy[i]*s[i]) / ∑s[i] / 3) Where s[i] is the area of ​​the i-th triangle that divides the cluster, l[i] is the length of the i-th side of the cluster, and (cx[i], cy[i]) is the centroid coordinates of the i-th triangle that divides the cluster; Focus i (cx',cy') is the location of the planned station. i , while focusing on the center of gravity i (cx',cy') is the center of the circle, with the north direction as 0 degrees, calculate its positive and negative angle bandwidth The number of weakly covered grids Grid0 within the angle range of degrees, and calculate the positive and negative Angle in the clockwise direction of 1 degree bandwidth The number of weakly covered grids in the angle range of degrees is Grid1, and so on. Finally, the positive and negative Angles in the clockwise direction of 359 degrees are calculated. bandwidth Number of weakly covered grids within the angle range Grid 359 ; Sort by the number of weakly covered grids, and select the first K grids with the largest number of weakly covered grids n The corresponding angle Bearing n Azimuth as the direction angle of the planned station area i-n , the angle between the direction angles of the planned station cells is greater than the minimum threshold angle Angle min ; Planning station location Location i Take the center of the circle as the center, obtain the surrounding adjacent base stations, and extract the location of the corresponding base stations whose coverage direction is not facing the planned station location. i The cell and its MR information; obtain the engineering parameter information of the surrounding adjacent cells, and use the average station height and average downtilt angle of the surrounding adjacent cells as the station height of the planned station i Downtilt i-n , thus obtaining the initial plan for the planning station; among them: In the formula, Hight t and Downtilt t Identify the site height and downtilt angle of the tth adjacent cell respectively; Step 2: Based on the MR information of the neighboring cells around the planning station, the first part of the MR information is cleaned and transformed to obtain two parts of feature vectors. The first part is the original MR feature vector, and the second part is the feature vector after cleaning and deformation based on the original feature vector. The importance of the feature vector in the model analysis is analyzed by using the pipeline model Pipeline. According to the score of the feature vector importance, the top N most important feature vectors are selected as the basis for modeling. The selected feature vectors are normalized to reduce the influence of individual feature vectors on modeling. Then, the model is trained based on the random forest method. The grid search method GridSearchCV is combined to select the optimal combination of the three important parameters n_estimators, max_depth, and min_samples_split in the random forest method, so as to finally generate the MR direction angle prediction model. The second part of MR information is input into the MR direction angle prediction model to obtain the direction angle Direction of the second part of MR information j , combined with the corrected TADV time advance in the MR information, the MR longitude and latitude (Lon j ,Lat j ), thereby achieving full MR coverage of the existing network cells; MR longitude and latitude (Lon j ,Lat j )The calculation function is as follows: (Lon j ,Lat j )=getLocation(lon cell ,lat cell ,Direction j ,TADV j ) Among them, lon cell ,lat cell ,Direction j ,TADV j They are respectively the longitude of the cell, the latitude of the cell, the predicted direction angle of the jth MR without AOA parameters and excluding longitude and latitude, and the correction value of TADV reported by the jth MR without AOA parameters and excluding longitude and latitude; Step 3: Based on the full MR of the surrounding adjacent cells, the actual wireless propagation model of the scene environment around the planning station is obtained through support vector machine and cross-validation method: Best_model=getmodel(gpid,df) Among them, gpid is the label of the cluster, df is the set of all MRs of the surrounding adjacent cells; Combine the antenna horizontal pattern and antenna vertical pattern, and calculate the coverage radius Distance of the planned station. i 、Hight i 、Azimuth i-n , Downtilt i-n , predict the grid RSRP coverage within the coverage radius of the planned station; RSRP calculation function: RSRP_value = cal(gpid, Best_model, center_lon, center_lat, azmuths, tilts, lon_lst, lat_lst, height) Among them, gpid is the label of the cluster, Best_model is the real wireless propagation model of the surrounding environment of the planning station, center_lon is the longitude of the planning station, center_lat is the latitude of the planning station, azmuths is the direction angle set of the planning station cell, tilts is the downtilt angle set of the planning station cell, lon_lst is the grid longitude within the coverage range of the planning station, lat_lst is the grid latitude within the coverage range of the planning station, and height is the height of the planning station; Adjust the height i Downtilt i-n , and then merge it with the full MR of the surrounding adjacent existing LTE cells, take the cell corresponding to the maximum RSRP value in the same grid as the main coverage cell, and obtain the coverage prediction result after the planned station is opened.