A mobile communication network coverage optimization method based on breathing histogram algorithm

Optimizing the antenna parameters of 4G/5G mobile communication networks through the breathing histogram algorithm solves the problem of lack of global and high-dimensional optimization capabilities in the existing technology, and realizes more efficient network coverage optimization, which is suitable for large-scale networks and reduces the Optimization error.

CN114169244BActive Publication Date: 2024-11-15SOUTH CHINA AGRICULTURAL UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202111514472.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-12-13
Publication Date
2024-11-15
Estimated Expiration
2041-12-13

AI Technical Summary

Technical Problem

The existing 4G/5G mobile communication network lacks global and high-dimensional optimization capabilities in outdoor coverage optimization, making it difficult to achieve the overall optimization effect of the network without affecting the surrounding area.

Method used

Using the mobile communication network coverage optimization method based on the respiratory histogram algorithm, the signal strength in the base station cell is calculated by constructing a dual-polar antenna signal propagation model, and the decision variables are iteratively optimized to adjust the transmission power of the antenna , azimuth and downtilt angles to achieve the balance of the optimization goal.

Benefits of technology

It achieves a balance between high-dimensional optimization capabilities and global optimization effects, improves optimization accuracy and diversity, avoids quantization errors, is suitable for large-scale networks, and is more in line with the signal propagation model in engineering practice.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114169244B_ABST
    Figure CN114169244B_ABST
Patent Text Reader

Abstract

The invention discloses a mobile communication network coverage optimization method based on a breathing histogram algorithm, the specific steps of which include constructing a dual-polarized antenna signal propagation model for each antenna of all base station cells in all base stations that need to be optimized, extracting parameters and optimization targets in the base station cells, encoding decision variables according to the determined optimization targets of all base station cells to construct a population, iterating the population composed of decision variables by the breathing histogram algorithm to obtain the optimal solution of the optimization target, and outputting the optimization parameters corresponding to the correction of each antenna in the optimization target of the optimal solution. The method of the invention has high-dimensional optimization capability, faster convergence speed during operation, and is more in line with the engineering practice of base station network optimization.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of mobile communication networks, and in particular relates to a mobile communication network coverage optimization method based on a breathing histogram algorithm. Background Art

[0002] 5G and 4G mobile communication networks have technical continuity, and their wireless networks have universal coverage optimization methods. At present, there are a large number of 4G / 5G mobile communication network base stations, and there are mutual constraints between high-dimensional decision variables such as outdoor transmission power, antenna azimuth, and downtilt angle between adjacent base stations. In addition, since electromagnetic wave propagation is nonlinear, the relationship between the optimization target and the decision variable is also nonlinear and strongly coupled. Therefore, the outdoor coverage optimization problem has the characteristics of multi-objective, high-dimensional, nonlinear, strongly coupled, and mixed variables.

[0003] At present, the actual outdoor coverage optimization work of 4G / 5G mobile communication networks often relies on the personal experience of engineers, and the uncertainty of optimization effect is difficult to overcome. For network optimization methods using automatic methods, some are optimized based on the results of big data or machine learning, such as the automatic optimization algorithm of 5G problem areas based on pattern optimization, the automatic optimization algorithm of problem areas based on cell contribution analysis based on big data analysis, and the automatic mathematical method of traversal exhaustive method based on MR multidimensional data analysis. These methods have good optimization effects on local areas, but due to the strong coupling of 4G / 5G network parameters, the change of problem area parameters will also have a certain impact on the surrounding areas. This type of optimization method lacks globality and it is difficult to ensure that the overall network achieves the optimal effect. Some are based on biological evolution algorithms, such as genetic algorithms, particle swarm algorithms, and differential evolution algorithms. These methods have good practicality due to the discretization characteristics used in engineering facilities, but are limited by the insufficient high-dimensional optimization capabilities of biological evolution algorithms. These methods are suitable for the optimization of small grid networks and do not yet have the ability to optimize the entire network.

[0004] It can be seen that the existing technology lacks an optimization method that optimizes the global network without causing negative impact on the neighborhood, can take into account the mutual influence between optimization objectives, adapts to large-scale networks and has high-dimensional optimization capabilities. Summary of the invention

[0005] In order to overcome the defects and shortcomings of the prior art, the present invention provides a mobile communication network coverage optimization method based on a breathing histogram algorithm, which aims to achieve a balance between high-dimensional optimization capability and global optimization effect for 4G / 5G outdoor optimization scenarios.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A method for optimizing mobile communication network coverage based on a breathing histogram algorithm comprises the following steps:

[0008] A dual-polarization antenna signal propagation model is constructed for each antenna of all base station cells in each base station, and the signal strength of a sampling point in each base station cell is calculated from the dual-polarization antenna signal propagation model;

[0009] The three parameters of RSRP, SINR and OCR related to the signal strength of each base station cell and the sampling point are taken as optimization targets, and the transmission power P of the antenna of the base station cell in each base station is taken as i , azimuth angle A, and downtilt angle D are encoded as decision variables for network optimization to form a population;

[0010] For the population encoded by the decision variables, the breathing histogram algorithm is used to iterate the population, and the population is continuously updated until the difference between the aggregate fitness values ​​of the two iterative populations is less than the set threshold or the number of iterations reaches the set maximum number of iterations;

[0011] The optimized reference signal receiving level RSRP, signal-to-noise ratio SINR, and overlap coverage ratio OCR are obtained from the last updated population, and the decision variable encoding decoding operation is performed on the population to obtain the corresponding optimized antenna transmission power P in each base station cell. i , azimuth angle A, downtilt angle D parameters.

[0012] Preferably, the steps of constructing a dual-polarized antenna signal propagation model to calculate the signal strength of a sampling point in each base station cell are:

[0013] According to the antenna gain G, horizontal 3dB power angle θ1, vertical 3dB power angle θ2 of the dual-polarized directional antenna, the measured reference signal receiving level RSRP, the measured signal-to-noise ratio SINR, the bottom center coordinates O(X0, Y0) of the base station, the antenna height H, azimuth angle A, downtilt angle D, and transmission power P i , calculate the signal strength of any sampling point S(X1,Y1) within the coverage range of the dual-polarized directional antenna electromagnetic wave signal.

[0014] Furthermore, the steps for calculating the signal strength of any sampling point S(X1, Y1) within the coverage range of the dual-polarized directional antenna electromagnetic wave signal are as follows:

[0015] According to the following formula, the central lobe point M(X M ,Y M )coordinate:

[0016]

[0017] From the above formula, we can get M(X M ,Y M ), the central lobe point M(X M ,Y M ) to the dual-polarized directional antenna is as follows:

[0018]

[0019] From the above formula, we can get the central lobe point M(X M ,Y M ) to the distance d from the dual-polarized directional antenna, and then the central lobe point M(X M ,Y M ) M The following formula is:

[0020]

[0021] Taking the logarithm of 10 on both sides of the above equation, we can get the central lobe point M(X M ,Y M ) M The logarithm of is expressed as follows:

[0022]

[0023] After transformation, P M The unit is decibel milliwatt; P T is the transmitting power of the antenna, in watts, and n is the attenuation coefficient of the electromagnetic wave emitted by the antenna;

[0024] Then, by P M Obtain the signal strength P of the electromagnetic wave emitted by the dual-polarization directional antenna at the horizontal 3dB power point N N The following formula is:

[0025]

[0026] From the above formula, the distance d1 from any sampling point S(X1, Y1) to the antenna within the coverage range of the dual-polarized directional antenna electromagnetic wave signal is solved as follows:

[0027]

[0028] Where d0 is the sampling point S(X S ,Y S ) to the horizontal distance of the base station coordinates O(X0,Y0);

[0029] The vertical deflection angle between the sampling point S (X1, Y1) and the propagation direction of the central lobe of the antenna electromagnetic wave signal is solved from the following formula:

[0030]

[0031] Where d2 is the distance from the sampling point S(X1,Y1) to the center lobe point M(X M ,Y M ) horizontal distance;

[0032] Then, calculate the signal strength P at the horizontal offset projection point P P The following formula is:

[0033]

[0034] in, The horizontal offset projection point P is the intersection of the arc with O as the center and OM as the radius on the horizontal plane and the extension line of the line segment OS;

[0035] Calculate the signal strength P at the vertical offset projection point Q Q The following formula is:

[0036]

[0037] The vertical offset point Q is the intersection of the arc in the vertical plane with O' as the center and d1 as the radius and the straight line O'P;

[0038] Calculate the power P at the vertical 3dB power point R R The following formula is:

[0039] P R =P Q -3

[0040] Finally, the signal strength P of the sampling point S(X1, Y1) is calculated as follows S for:

[0041]

[0042] in,

[0043] Signal sampling is performed multiple times on each dual-polarization directional antenna, and the value of its signal attenuation coefficient n is adjusted according to the sampling results.

[0044] Furthermore, the steps for encoding the decision variables are:

[0045] The number of base station cells is set to K, and the azimuth angle A, downtilt angle D, and transmission power P of each base station cell antenna are i Each is encoded as an individual; the azimuth angle A and the downtilt angle D are set as integer discrete variables and encoded using integer indexes, and the transmission power P iSet as continuous variables and use float type real number encoding; azimuth angle A, downtilt angle D, transmission power P i All are randomly sampled for encoding; azimuth angle A, downtilt angle D, antenna transmission power P i After N random samplings are repeated, N individuals are formed, and N individuals constitute a population.

[0046] Furthermore, the steps of iterating using the breathing histogram algorithm are:

[0047] Initialize the boundaries of the statistical histogram and use 3K histograms to represent the probability distribution of all decision variables;

[0048] The average reference signal receiving level RSRP, average signal-to-noise ratio SINR, and overlap coverage ratio OCR of each individual in population X are calculated and normalized, and then the corresponding aggregate fitness value scalarfit of each individual is calculated according to the following formula:

[0049] scalarfit=[w1,w2,...,w k ]·[y1,y2,...,y k ] T

[0050] Among them, [w1,w2,...,w k ] is the weight vector, w k represents the aggregate weight of the kth optimization objective, [y1,y2,...,y k ] is the individual fitness matrix, y k Represents the fitness value of the kth optimization objective of the individual;

[0051] The minimum aggregate fitness value within the boundary of each histogram is counted. If the minimum aggregate fitness value in a column of the histogram is smaller than the minimum fitness value of its left and right neighbor columns, the column is the sensitive area of ​​the histogram, otherwise it is the insensitive area.

[0052] Assign histogram probabilities, assign subpopulation probabilities to non-sensitive areas and sensitive areas respectively, and perform contraction operation and then expansion operation on sensitive areas;

[0053] Generate a new population X'={X'1, X'2, ..., X' N}, then the number of individuals in each subpopulation is calculated by the probability and probability density function of each subpopulation in the non-sensitive area and the sensitive area, and a predetermined number of subpopulations are generated by sampling in the non-sensitive area and the sensitive area respectively. All subpopulations constitute the ideal probability distribution population R = {R1, R2, ..., R N}, each individual of the original population X is updated according to the following formula to obtain a new population;

[0054]

[0055] Among them, B j is the neighbor vector of the jth individual in the original population X, k1 and k2 are the serial numbers of the two random elements of the neighbor vector, and n1 and n2 are the numbers of the jth individual X in the original population. j two neighbors, rand is a random number in the range [0,1], Δd is the random variation distance, round is the rounding function, and n is X j The mutated sequence numbers of the two neighbors of X' j is the jth individual of the new population X';

[0056] According to the three optimization objectives, calculate the fitness values ​​of all individuals in the new population X' where fit i =[RSRP i ,SINR i ,OCR i ], then normalize it, and then calculate its aggregate fitness value scalarfit_new;

[0057] Perform mutation operation on the new population X'; calculate the crowding distance of all individuals in the new population X' according to the following formula:

[0058]

[0059] Then, the individuals are sorted from small to large according to the crowding distance, and the data of one dimension is randomly selected from the top 40% of the individuals and subjected to random mutation according to the following formula;

[0060] X' w = rand·(X max (w)-X min (w))+X min (w)

[0061] Among them, Cd i represents the crowding distance of the ith individual, m represents the dimension of the individual fitness value, min(fit(:,j)) and max(fit(:,j)) represent the minimum and maximum values ​​of the jth optimization target of all individuals, fit(i,j) represents the jth optimization target value of the ith individual, fit(i-1,j) and fit(i+1,j) represent the two optimization target values ​​before and after individual i when all individuals are sorted from small to large according to the jth optimization target value. w is a random positive integer within the range of the decision variable dimension, X′ w is a dimension of X′, X max (w), X min (w) are the maximum and minimum values ​​of the decision variables of the wth dimension, and rand is a random number in the range of [0,1];

[0062] Perform selection operations on the original population X and the new population X';

[0063] Iteration condition judgment: let the updated population be X", first calculate the fitness value of the selected new population X", then calculate the aggregate fitness value of X", if the difference between the aggregate fitness values ​​of the two iterative populations is less than the set threshold or the number of iterations reaches the set maximum number of iterations, then the iteration ends, otherwise continue to execute one iteration process;

[0064] Output the fitness values ​​of the population corresponding to the three optimization objectives and decode the decision variables.

[0065] Furthermore, the calculation formulas of the three optimization objectives are as follows:

[0066]

[0067]

[0068]

[0069] Among them, N s is the number of sampling points in the network, Pc i is the serving cell signal strength at the i-th sampling point, P i,j is the signal strength of the jth cell at the ith sampling point, ζ i is the ambient noise of the i-th sampling point, N oc is the number of overlapping sampling points.

[0070] Furthermore, the subpopulation probability is allocated to the non-sensitive area according to the following formula:

[0071]

[0072] The probability density function in the non-sensitive area is as follows:

[0073]

[0074] The probability of allocating subpopulations to sensitive areas is as follows:

[0075]

[0076] The probability density function in the sensitive area is as follows:

[0077]

[0078] Among them, Sr is the ratio of the number of individuals in the sensitive area to the total number of individuals in the population, V i is the probability value of the histogram of the i-th non-sensitive area after sorting by the local optimal solution, K and K sare the number of histograms and the number of sensitive areas, σ1 is the probability difference of each histogram in the non-sensitive area, and bl i 、bu i are the lower and upper boundaries of the i-th histogram in the non-sensitive area, V i ' is the probability value of the histogram of the i-th sensitive area after sorting by the local optimal solution, σ2 is the probability difference of each histogram in the sensitive area, σ is the shape parameter of the normal distribution, x i,best is the local optimal solution position of the i-th histogram in the sensitive area.

[0079] Furthermore, the specific process of performing the contraction operation and the expansion operation on the sensitive area is as follows:

[0080] The sensitive area is shrunk by the following formula:

[0081]

[0082] Among them, bl i 、bu i are the lower and upper boundaries of the ith histogram, respectively, and X i A new subpopulation generated by Gaussian distribution within the boundary of the i-th histogram;

[0083] The sensitive area is expanded to expand the left and right boundaries of the sensitive area according to the following formula;

[0084]

[0085] Among them, bl i,e 、bu i,e are the lower and upper boundaries of the ith histogram in the e-th iteration, respectively; α is the breathing factor, ranging from (0,1); Δl1 and Δl2 are the last contraction distances of the lower and upper boundaries of the ith histogram, respectively, calculated as follows:

[0086]

[0087] Furthermore, the steps of performing the selection operation on the original population X and the new population X' are:

[0088] For each individual in the new population X', perform the following operations in sequence:

[0089] For any individual X' in the new population X' i , and the original population X corresponds to individual X i Compare all neighbors of X' i With X i Neighbor X n If one of the following three conditions is met, the first condition of the following formula is executed and X' is used i Replace X in the original population Xn :

[0090]

[0091] After performing the above operations, the updated population X is denoted as X".

[0092] Further, according to the obtained transmission power P of the optimized antenna corresponding to each base station cell, i After adjusting the antenna parameters such as azimuth angle A and downtilt angle D, the optimization effect is verified through road tests, and the breathing histogram algorithm is adjusted according to the road test results to iterate the population.

[0093] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0094] (1) Generate new individuals according to the probability distribution of the population while basically preserving the combination information of individual decision variables, thus having good high-dimensional optimization capabilities;

[0095] (2) There is no crossover operation in the iterative process. When optimizing discrete variables, no points that are not on the quantized value will be generated. Therefore, no quantization operation is required, and there is no quantization error. Therefore, it has higher optimization accuracy in optimization problems with discrete variables.

[0096] (3) It uses a mutation strategy based on crowding distance and a joint selection strategy based on aggregated fitness value to maintain the diversity of the population. Compared with other evolutionary algorithms, its multi-objective optimization results are more diverse and can provide decision makers with a richer decision space.

[0097] (4) Index coding is used for discrete variables, which achieves higher coding efficiency for non-integer and non-uniform discrete variables. At the same time, index coding can compress the discrete search domain, making the breathing histogram algorithm have a faster convergence speed than the existing evolutionary algorithm in mixed variable optimization;

[0098] (5) According to the actual structure of the dual-polarized antenna, the signal strength calculation model at any point is derived in combination with the electromagnetic wave free space attenuation model. The signal attenuation parameters collected in the actual environment are also corrected. Compared with other signal propagation models used in the prior art, this model is more in line with the actual situation of engineering practice. BRIEF DESCRIPTION OF THE DRAWINGS

[0099] Figure 1 Schematic diagram of the flow of the mobile communication network coverage optimization method based on the breathing histogram algorithm;

[0100] Figure 2 for Figure 1 Schematic diagram of the signal propagation model of dual-polarization directional antenna;

[0101] Figure 3 for Figure 1 Schematic diagram of the decision variable encoding structure of the respiratory histogram algorithm;

[0102] Figure 4 for Figure 1 Schematic diagram of the flow of the breathing histogram algorithm. DETAILED DESCRIPTION

[0103] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0104] Example

[0105] like Figure 1 As shown, the specific steps of the mobile communication network coverage optimization method based on the breathing histogram algorithm are as follows:

[0106] S1. Determine all base stations that need to be optimized, and extract parameters and optimization targets within the base station cell for each base station. The specific steps are as follows:

[0107] S11. Construct the following for each dual-polarized directional antenna: Figure 2 The dual-polarization directional antenna signal propagation model shown:

[0108] According to the free space propagation theory of electromagnetic waves, the antenna gain G of the dual-polarized directional antenna, the horizontal 3dB power angle θ1, the vertical 3dB power angle θ2, the measured reference signal receiving level RSRP, the measured signal-to-noise ratio SINR, the bottom center coordinates of the base station are O(X0, Y0), according to the base station coordinates O(X0, Y0), the antenna height is H, the antenna azimuth angle A, the antenna downtilt angle D, the antenna transmission power P i , calculate the signal strength of any sampling point S(X1,Y1) within the coverage range of the dual-polarized directional antenna electromagnetic wave signal;

[0109] According to formula (1), the coordinates of the central lobe point M of the electromagnetic wave signal emitted by the dual-polarized directional antenna are solved. The coordinates of M are M(X M ,Y M ):

[0110]

[0111] From formula (1), we can get M(X M ,Y M ), the distance d from the center lobe point M to the dual-polarization directional antenna is calculated by equation (2):

[0112]

[0113] After obtaining the distance d from the center lobe point M to the dual-polarization directional antenna from equation (2), the signal power P at the center lobe point M is calculated by equation (3): M for:

[0114]

[0115] Taking the logarithm transformation of both sides of equation (3) to 10, we can get the signal strength P at the central lobe point M: M The logarithmic expression (4):

[0116]

[0117] After transformation, P M The unit is decibel milliwatt (unit symbol dBm); P T is the transmit power of the antenna in watts (unit symbol W), n is the attenuation coefficient of the electromagnetic wave emitted by an antenna, and the value range of n is [2,4]. The specific value is adjusted according to the electromagnetic environment around the base station reflected by the measured reference signal reception level RSRP;

[0118] Next, the signal strength P of the electromagnetic wave emitted by the dual-polarized directional antenna at the horizontal 3dB power point N is obtained from equation (3) or equation (4): N for:

[0119]

[0120] From equation (6), the distance d1 from any sampling point S(X1, Y1) to the antenna within the coverage range of the dual-polarized directional antenna electromagnetic wave signal is solved:

[0121]

[0122] Where d0 is the sampling point S(X S ,Y S ) to the horizontal distance of the base station coordinates O(X0,Y0);

[0123] From equation (7), we can solve the vertical deflection angle between the sampling point S(X1, Y1) and the propagation direction of the central lobe of the antenna electromagnetic wave signal:

[0124]

[0125] Where d2 is the distance from the sampling point S(X1,Y1) to the center lobe point M(X M ,Y M ) horizontal distance;

[0126] Then, calculate the signal strength P at the horizontal offset projection point P P for:

[0127]

[0128] in, The horizontal offset projection point P is the intersection of the arc with O as the center and OM as the radius on the horizontal plane and the extension line of the line segment OS;

[0129] Calculate the signal strength P at the vertical offset projection point Q Q for:

[0130]

[0131] The vertical offset point Q is the intersection of the arc in the vertical plane with O' as the center and d1 as the radius and the straight line O'P;

[0132] Calculate the power P at the vertical 3dB power point R R for:

[0133] P R =P Q -3 (10)

[0134] Finally, we get formula (11) to calculate the signal strength P of the sampling point S (X1, Y1) S for:

[0135]

[0136] in,

[0137] At this point, the dual-polarization directional antenna signal propagation model is constructed. In all base stations that need to be optimized, multiple sampling is performed on each dual-polarization directional antenna to adjust the value of its signal attenuation coefficient n, and the corresponding corrected dual-polarization directional antenna signal propagation model is obtained to calculate the signal strength of any point within the coverage range of the dual-polarization directional antenna electromagnetic wave signal;

[0138] S12. Based on the dual-polarization directional antenna signal propagation model, extract parameters and optimization targets within the base station cell for each base station. The specific steps are as follows:

[0139] S121, extracting the longitude and latitude of the base station and the height H of the base station antenna, taking the longitude and latitude of the base station as the origin of the rectangular coordinate system, and transforming the coordinates of other points within the coverage range of the antenna electromagnetic wave signal in the rectangular coordinate system accordingly with the longitude and latitude of the base station as the reference point;

[0140] S122, extract the transmission power P of each antenna i , azimuth angle A, downtilt angle D;

[0141] S123, setting sampling points and extracting the latitude and longitude of the sampling points, the sampling points are evenly set on the road test line;

[0142] S124, using three types of indicators, namely, reference signal reception level RSRP, signal-to-noise ratio SINR, and overlap coverage ratio OCR, as optimization targets for each base station cell;

[0143] Overlapping coverage means that the RSRP of at least three sampling points in the neighboring area of ​​a cell within a base station is within 6dBm. The overlapping coverage rate (OCR) refers to the ratio of the number of sampling points in overlapping coverage to the total number of sampling points. A base station includes three antennas, and each antenna corresponds to a base station cell.

[0144] S2, decision variable encoding, combined Figure 3 The coding structure shown is based on the transmission power P of all base station cell antennas. i , azimuth angle A, and downtilt angle D are encoded as decision variables for network optimization, as follows:

[0145] The total number of base station cells that need to be optimized is set to K, and the azimuth angle A, downtilt angle D, and transmit power P of each base station cell antenna are set to i They are respectively regarded as decision variables of one dimension. All decision variables are encoded together as an individual. Then each individual includes 3K decision variables of one dimension. In the actual measurement process, the azimuth angle A and the downtilt angle D are set as integer discrete variables. The value range of the azimuth angle A is an arithmetic sequence with a tolerance of 10 {0,10,20,30,…,350}, and the value range of the downtilt angle D is an arithmetic sequence with a tolerance of 1 {0,1,2,3,…,20}. The discrete variables are encoded using integer indexes. The azimuth angle A is encoded as an element in the integer sequence {1,2,3,…,36}. The transmission power P i It is a continuous variable with a value range of [0,50] and is encoded using float real numbers. Azimuth angle A, downtilt angle D, and transmit power P i The azimuth angle A, downtilt angle D, and antenna transmission power P are randomly sampled and encoded within their respective value ranges. i After repeating random sampling N times, according to Figure 3 The coding structure of is composed of N individuals, and these N individuals constitute the population of the breathing histogram algorithm;

[0146] S3, use the breathing histogram algorithm to iterate the population, combined with Figure 4 As shown, the specific steps are as follows:

[0147] S31, initializing and setting the breathing factor, mutation probability, population size, neighbor size, maximum number of iterations, and signal attenuation coefficient n of the population;

[0148] S32, initialize and set the population X = (X1, X2, ..., X N), for all individuals in population X, the azimuth angle A, downtilt angle D, and transmission power P i Initialize by taking random values ​​within their respective value ranges;

[0149] S33, initializing the boundaries of the statistical histogram, using 3K histograms to represent the probability distribution corresponding to all dimensional decision variables in all individuals;

[0150] S34, calculate the average reference signal receiving level RSRP, average signal-to-noise ratio SINR, and overlap coverage rate OCR of each individual in population X and normalize them. These three optimization target values ​​are also called individual fitness values. The calculation formulas of the three optimization targets are as follows:

[0151]

[0152]

[0153]

[0154] Among them, N s is the number of sampling points in the network, Pc i is the serving cell signal strength at the i-th sampling point, Pc i The signal propagation model of dual-polarized antenna established by S11 is used to calculate P i,j is the signal strength of the jth cell at the ith sampling point, P i,j ζ is calculated from the dual-polarization antenna signal propagation model established by S11. i is the ambient noise of the i-th sampling point, N oc is the number of overlapping sampling points. The overlapping sampling points are defined as the sampling points with three neighboring areas whose RSRP difference is within 6 dBm. Then, the corresponding aggregate fitness value scalarfit of each individual is calculated according to formula (15):

[0155] scalarfit=[w1,w2,...,w k ]·[y1,y2,...,y k ] T (15)

[0156] Among them, [w1,w2,...,w k ] is the weight vector, w k represents the aggregate weight of the kth optimization objective,

[0157] [y1,y2,...,y k ] is the individual fitness matrix, y k represents the fitness value of the kth optimization target of the individual. In this embodiment, since three optimization targets are selected, the maximum value of k is 3;

[0158] S35, counting the minimum aggregate fitness values ​​within the boundaries of each histogram, if the minimum aggregate fitness value in a column of the histogram is smaller than the minimum fitness value of its left and right neighbor columns (if any), then the column is called the sensitive area of ​​the histogram, that is, the potential area of ​​the global optimal solution, otherwise it is a non-sensitive area;

[0159] S36, assign histogram probability, assign subpopulation probability to non-sensitive areas according to formula (16),

[0160]

[0161] The probability density function in the non-sensitive area is uniformly distributed as shown in formula (17).

[0162]

[0163] The probability of allocating subpopulations to sensitive areas is calculated according to formula (18):

[0164]

[0165] The probability density function in the sensitive area is a Gaussian distribution as shown in formula (19).

[0166]

[0167] Among them, Sr is the sensitivity, that is, the ratio of the number of individuals in the sensitive area to the total number of individuals in the population, V i is the probability value of the histogram of the i-th non-sensitive area after sorting by the local optimal solution, K and K s are the number of histograms and the number of sensitive areas, respectively. σ1 is the probability difference of each histogram in the non-sensitive area. The more histograms there are, the smaller σ1 is. i 、bu i are the lower and upper boundaries of the i-th histogram in the non-sensitive area, V i ' is the probability value of the histogram of the i-th sensitive area after sorting by the local optimal solution, σ2 is the probability difference of each histogram of the sensitive area, the more histograms there are, the smaller σ2 is, σ is the shape parameter of the normal distribution, reflecting the degree of dispersion of the normal distribution, the larger σ is, the more dispersed the data distribution is, the smaller σ is, the more concentrated the data distribution is, in this embodiment, preferably σ = 0.2, x i,best is the local optimal solution position of the i-th histogram in the sensitive area;

[0168] S37, performing a shrinking operation on the sensitive area, and shrinking the left and right boundaries of the sensitive area according to formula (20);

[0169]

[0170] Among them, bl i 、bu iare the lower and upper boundaries of the ith histogram, respectively, and X i A new subpopulation generated by Gaussian distribution within the boundary of the i-th histogram;

[0171] S38, the sensitive area is expanded. To avoid missing the global optimal solution due to excessive shrinkage in step S37, the left and right boundaries of the sensitive area are expanded again according to formula (21);

[0172]

[0173] Among them, bl i,e 、bu i,e are the lower and upper boundaries of the ith histogram in the e-th iteration respectively; α is the breathing factor, and its value range is (0,1). The larger α is, the better the global search capability is; the smaller α is, the faster the convergence speed is. In this embodiment, the value can be 0.8; Δl1 and Δl2 are the last shrinkage distances of the lower and upper boundaries of the ith histogram respectively, and are calculated as shown in formula (22):

[0174]

[0175] S39, generate a new population X'={X'1, X'2, ..., X' N}, the number of individuals in the new population is the same as that in the original population, both of which are N. The number of individuals in each subpopulation is calculated according to equations (16) and (17) in step S36, and then a predetermined number of subpopulations are generated by sampling in the non-sensitive area and the sensitive area using equations (17) and (19), respectively. All subpopulations constitute an ideal probability distribution population R = {R1, R2, ..., R N}, each individual of the original population X is updated according to formula (23) to obtain a new population;

[0176]

[0177] Among them, B j is the neighbor vector of the jth individual in the original population X, k1 and k2 are the serial numbers of the two random elements of the neighbor vector, and n1 and n2 are the numbers of the jth individual X in the original population. j two neighbors, rand is a random number in the range [0,1], Δd is the random variation distance, round is the rounding function, and n is X j The mutated sequence numbers of the two neighbors of X' j is the jth individual of the new population X';

[0178] S310, calculate the fitness values ​​of all individuals in the new population X' according to formulas (12), (13), and (14) where fit i =[RSRP i ,SINR i,OCR i ], normalized according to the above, and then calculated its aggregated fitness value scalarfit_new according to formula (15);

[0179] S311, perform mutation operation on the new population X', calculate the crowding distance Cd of all individuals in the new population X' according to formula (24), Cd represents the diversity of individual fitness value, the larger the value, the better the diversity of individuals, and the smaller the value, the worse the diversity of individuals; then sort them from small to large according to the crowding distance, randomly select one dimension of data from the top 40% of the individuals and perform random mutation according to formula (25);

[0180]

[0181] X' w = rand·(X max (w)-X min (w))+X min (w) (25)

[0182] Among them, Cd i represents the crowding distance of each individual, m represents the dimension of individual fitness value, min(fit(:,j)) and max(fit(:,j)) represent the minimum and maximum values ​​of the jth optimization target of all individuals, fit(i,j) represents the jth optimization target value of the i-th individual, fit(i-1,j) and fit(i+1,j) represent the two optimization target values ​​before and after individual i when all individuals are sorted from small to large according to the jth optimization target value. w is a random positive integer within the range of decision variable dimension, X′ w is a dimension of X′, X max (w), X min (w) are the maximum and minimum values ​​of the decision variables of the wth dimension, and rand is a random number in the range of [0,1];

[0183] S312, perform selection operations on the original population X and the new population X', and perform the following operations on each individual of the new population X' in sequence: for any individual X' of the new population X' i , and the original population X corresponds to individual X i Compare all neighbors of X' i With X i Neighbor X n If one of the last three conditions of formula (26) is met, the first condition of formula (26) is executed and X' is used. i Replace X in the original population X n :

[0184]

[0185] After performing the above operations, the updated population X is represented as X”;

[0186] S313, iterative condition judgment, first calculate the fitness value of the selected new population X", then calculate the aggregate fitness value of X", if the difference between the aggregate fitness values ​​of the two iterative populations is less than the set threshold or the number of iterations reaches the set maximum number of iterations, then the iteration ends, otherwise return to step S34;

[0187] S314, output the fitness values ​​of the population corresponding to the three optimization targets, which is an N×3 matrix, where each row represents the target value of an individual on the three optimization targets, and then perform denormalization processing to obtain the average reference signal reception level RSRP, the average signal-to-noise ratio SINR, and the overlap coverage ratio OCR indicators;

[0188] S4. Extract the parameters of the decision variables of each dimension according to the coding structure of the decision variable coding, that is, the parameters of the optimized optimization target; output the parameters of the optimization target, and calculate the azimuth angle A, downtilt angle D, and transmission power P of all antennas in all base stations of the required network optimization. i The adjustment is implemented, and after the adjustment, road tests are carried out to verify the optimization effect, and the signal attenuation coefficient n and the breathing factor α of the population iteration process are adjusted based on the feedback of the road test results.

[0189] The mobile communication network coverage optimization method based on the breathing histogram algorithm has the following beneficial effects:

[0190] The breathing histogram is used as the probability model of the population, and its convergence and global search capabilities are balanced through expansion and contraction operations. The probability traction method is used to generate new individuals according to the probability distribution model of the population while basically preserving the information of individual decision variables, so as to have high-dimensional optimization capabilities. The mutation strategy based on crowding distance and the joint selection strategy of aggregated fitness value are used to improve the diversity of the population. Compared with other evolutionary algorithms in the prior art, the breathing histogram algorithm achieves better high-dimensional optimization capabilities. Since there is no crossover operation, the breathing histogram algorithm does not have non-quantization errors when optimizing discrete variables, thus achieving higher optimization accuracy.

[0191] The hybrid variable coding treats the antenna downtilt angle and azimuth angle in the decision variables as discrete variables, and treats the antenna transmit power as a continuous variable, so that the decision variables have both discrete variables and continuous variables. It achieves higher coding efficiency for non-integer and non-uniform discrete variables. At the same time, index coding can compress the discrete search domain, thereby accelerating the convergence speed of the breathing histogram algorithm.

[0192] According to the actual structure of the dual-polarized antenna and the free space attenuation model of electromagnetic waves, the calculation formula for the signal strength at any point is derived. The signal attenuation parameters collected in the actual environment are also corrected. Compared with other signal propagation models in the prior art, this model is more in line with the actual situation of engineering practice.

[0193] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A mobile communication network coverage optimization method based on a breathing histogram algorithm, characterized in that: The following steps are involved: A dual-polarization antenna signal propagation model is constructed for each antenna of all base station cells in each base station, and the signal strength of the sampling point in each base station cell is calculated from the dual-polarization antenna signal propagation model, specifically including: According to the antenna gain G, horizontal 3dB power angle θ1, vertical 3dB power angle θ2 of the dual-polarized directional antenna, the measured reference signal receiving level RSRP, the measured signal-to-noise ratio SINR, the bottom center coordinates O(X0, Y0) of the base station, the antenna height H, azimuth angle A, downtilt angle D, and transmission power P i , calculate the signal strength of any sampling point S(X1,Y1) within the coverage range of the dual-polarized directional antenna electromagnetic wave signal; The three parameters of RSRP, SINR and OCR related to the signal strength of each base station cell and the sampling point are taken as optimization targets, and the transmission power P of the antenna of the base station cell in each base station is taken as i , azimuth angle A, and downtilt angle D are encoded as decision variables for network optimization to form a population, including: The number of base station cells is set to K, and the azimuth angle A, downtilt angle D, and transmission power P of each base station cell antenna are i Each is encoded as an individual; the azimuth angle A and the downtilt angle D are set as integer discrete variables and encoded using integer indexes, and the transmission power P i Set as continuous variables and use float type real number encoding; azimuth angle A, downtilt angle D, transmission power P i All are randomly sampled for encoding; azimuth angle A, downtilt angle D, antenna transmission power P i After N random samplings are repeated, N individuals are formed, and N individuals constitute a population; For the population encoded by the decision variables, the breathing histogram algorithm is used to iterate the population, and the population is continuously updated until the difference between the aggregate fitness values ​​of the two iterative populations is less than the set threshold or the number of iterations reaches the set maximum number of iterations; The optimized reference signal receiving level RSRP, signal-to-noise ratio SINR, and overlap coverage ratio OCR are obtained from the last updated population, and the decision variable encoding decoding operation is performed on the population to obtain the corresponding optimized antenna transmission power P in each base station cell. i , azimuth angle A, downtilt angle D parameters.