A performance-guaranteed 5G load balancing method

Abstract

This invention discloses a performance-guaranteed 5G load balancing method. It addresses the problem of unbalanced load between cells and low system utilization in existing technologies. The method includes: identifying neighboring cells as target cells; calculating the physical resource block occupancy of each target cell and comparing it with the balancing benchmark value of physical resource blocks of all cells to select balancing target cells; screening overloaded users; performing target cell transfer based on performance-guaranteed target cell search; convergence decision; and dynamically updating the balancing threshold until the balancing conditions are met. This invention starts with overloaded cells, identifies target overloaded users, analyzes the resource occupancy and signal-to-noise ratio of the current user's cell, and finds target cells with environments no worse than the current cell environment to implement load transfer, ensuring the overall performance efficiency of the system throughout the load transfer process. Dynamically setting the load transfer threshold ensures efficient load balancing convergence, thereby guaranteeing the overall performance of the 5G system.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a 5G load balancing method with guaranteed performance. Background Technology

[0002] As the scale of 5G users continues to expand, congestion or peak resource utilization in hotspot cells will become increasingly common, severely impacting users' normal use and service experience. Meanwhile, some cells experience low system utilization. Therefore, balancing the load between cells and maximizing system efficiency is an imperative task. Patent number ZL201410324613.0, entitled "A Load Balancing Method Based on Neighbor Cell Sets (NSLB)," solves the problem of traffic imbalance by analyzing the load space that the least loaded cell can accommodate and then inferring the number of users to be transferred from the neighbor cell set. However, NSLB still has several problems: different users request different physical resource blocks, and for the same capacity space, the strategies for allocating it to different users are also different. Furthermore, the user environment is not considered when transferring from the source cell to the target cell, and cell performance sometimes actually degrades. Summary of the Invention

[0003] This invention primarily addresses the problems of unbalanced load between cells and low system utilization in existing technologies, providing a performance-assured 5G load balancing method. Starting with the target overloaded users, it fully considers the environmental factors of the target cell for load transfer, implementing load balancing without reducing or even improving overall system performance. This, in turn, improves overall resource utilization and ultimately ensures system performance.

[0004] The above-mentioned technical problems of the present invention are mainly solved by the following technical solution: a 5G load balancing method with guaranteed performance, comprising the following steps:

[0005] Step 1: Obtain neighboring cells as target cells;

[0006] Step 2: Calculate the physical resource block occupancy of each target cell, compare it with the equilibrium benchmark value of physical resource blocks of all cells, and select the equilibrium target cells;

[0007] Step 3: Calculate the number of overloaded cells and select the same number of users from the overloaded cells as overloaded users;

[0008] Step 4: Calculate the performance guarantee value of each overloaded user relative to each equalization target cell. If the performance guarantee value of the overloaded user is not less than the set equalization threshold, the equalization target cell corresponding to the performance guarantee value is taken as a candidate cell. The cell with the largest performance guarantee value is selected from the candidate cells as the transfer target cell, and the overloaded user is transferred from the overloaded cell to the transfer target cell.

[0009] Step 5: Calculate the equilibrium factor for the target cell, set the equilibrium conditions, and make a convergence decision. If the equilibrium conditions are not met, repeat steps 1 to 5 until the equilibrium conditions are met.

[0010] Step 6: If the convergence count is exceeded and the load balancing condition is not met, dynamically update the load balancing threshold and repeat steps 1 to 6 until the load balancing condition is met and a load-balanced cell with guaranteed performance is obtained.

[0011] This invention starts with overloaded cells, identifies target overloaded users, analyzes the resource occupancy and signal-to-noise ratio of the current user's cell, and finds target cells with environments no worse than the current cell to implement load transfer, ensuring the overall performance efficiency of the system throughout the load transfer process. Furthermore, it dynamically sets the load transfer threshold to ensure efficient load balancing convergence, thereby guaranteeing the overall performance of the 5G system.

[0012] As a preferred approach, step one, obtaining neighboring cells as target cells, specifically includes:

[0013] Obtain n mutually neighboring cells as the target cell NBS n ={gNB1, gNB2, ..., gNB n}, obtain the load of each cell in the target cell {Cld1, Cld2, ..., Cld n The total number of physical resource blocks {PrT1, PrT2, ..., PrT} n}, the minimum signal-to-noise ratio within the cell coverage area {SNRmin1, SNRmin2, ..., SNRmin n} and maximum signal-to-noise ratio {SNRmax1, SNRmax2, ..., SNRmax} n}

[0014] As a preferred option, step two specifically includes the following processes:

[0015] (2-1): For NBS n ={gNB1, gNB2, ..., gNB n Each cell gNB in ​​} i Calculate its physical resource block occupancy.

[0016] OcuP i =ceil(PrT i*Cld i )

[0017] Where ceil() represents the ceiling function, i = 1, 2, ..., n;

[0018] Calculate the equilibrium benchmark value for all physical resource blocks in all cells.

[0019]

[0020] (2-2): Regarding NBS n ={gNB1, gNB2, ..., gNB n For each cell in the array, calculate the available space in its physical resource blocks.

[0021] LftP i =Prb Av -OcuP i

[0022] (2-3): This will satisfy LftP i Cells with a value greater than 0 are added to the target cell list for equalization. The number of cells is m, and the corresponding physical resource block space reserves are respectively

[0023] As a preferred option, step three includes the following specific steps:

[0024] (3-1): Calculate the HSet of the overloaded cell = NBS n -LSet, whose quantity is Nm HSet =nm;

[0025] (3-2): Select Nm quantities from HSet in descending order of throughput. HSet Users who are overloaded are recorded as such. Determine the cell to which the overloaded user belongs.

[0026] Set heavy user throughput

[0027] Signal-to-noise ratio

[0028] Bit error rate

[0029] As a preferred embodiment, step four includes the following process:

[0030] (4-1): Calculate the decimal signal-to-noise ratio for overloaded users.

[0031]

[0032] Where SNR′t =power(10, SNR) t / 10), t=1,2,…,Nm HSet power() represents the exponential function;

[0033] Calculate the bit error rate conversion value for overloaded users.

[0034]

[0035] Among them BER′ t = -ln(5*BER) t ) / 1.5, ln() represents the natural logarithm function;

[0036] Calculate the median value.

[0037]

[0038] Where Mf t =log2(1+SNR) t ′ / BER t log2() represents the logarithmic function with base 2;

[0039] (4-2): Set the subcarrier bandwidth Wd Sc The number of subcarriers Nm occupied by a single physical resource block Prb Calculate the bandwidth of a single physical resource block.

[0040] Wd Prb =Wd Sc *Nm Prb / 1024

[0041] Calculate the amount of physical resource blocks occupied by each heavily loaded user.

[0042] UPrb t =ceil(Thr t / (Wd Prb *Mf t ));

[0043] (4-3): For a balanced target cell, calculate the minimum signal-to-noise ratio in decimal.

[0044] SNRmn k =power(10,SNRmin) k / 10 )

[0045] Calculate the maximum signal-to-noise ratio in decimal.

[0046] SNRmx k =power(10, SNRmax) k / 10 )

[0047] Calculate the median.

[0048]

[0049] Calculate the expected signal-to-noise ratio (SNR) of each cell in the target equalization cell.

[0050]

[0051] Where Log10 represents the logarithmic function with base 10, k = j, j+1, ..., j+m-1;

[0052] For each overloaded user Usr t Iterate through each gNB in ​​the target cell for equalization. k Calculate the performance guarantee value.

[0053]

[0054] Where exp() is an exponential function with the natural constant e as its base;

[0055] (4-4): Set the equalization threshold LB Thr , 0 < LB Thr <1, if Pfas t，k ≥LB Thr Then gNB k As candidate cells, they constitute the candidate set CaSt Lb ;

[0056] Determine the candidate set CaSt Lb Is it empty? If empty, cancel this user's username / data. t Load transfer; if not empty, select CaSt. Lb China has the largest Pfas t，k Value of the community gNB mxk and will User Usr t From the community gNB Hs+t -1 transferred to the community gNB mxk middle;

[0057] Update community gNB mxk Physical resource possession OcuP mxk =OcuP mxk +UPrb t Physical resource block space margin Lftp mxk =Lftp mxk -UPrb t ;

[0058] Update Usr tThe residential community Physical resource possession Physical resource block space remaining

[0059] (4-5): In each user's User t Complete the traversal and update the load Cld for all cells. i =OcuP i / PrT i .

[0060] As a preferred option, step five includes the following specific steps:

[0061] (5-1): Set an acceptable equilibrium state β0 and a convergence target value ε for the target cell NBS. n ={gNB1, gNB2, ..., gNB n} Calculate the load balancing factor

[0062]

[0063] (5-2): Determine whether the equilibrium condition is met.

[0064]

[0065] If the equilibrium condition is met, the step ends; if the equilibrium condition is not met, the convergence counter Δ = Δ + 1.

[0066] (5-3): Set the convergence level SL mx Determine whether Δ≤SL is satisfied. mx If yes, repeat steps one through five until the equilibrium condition is met; otherwise, proceed to step six.

[0067] As a preferred option, step six includes the following specific steps:

[0068] Set equilibrium threshold to shrink variables If Δ>SL mx If the equilibrium condition is not met, then the equilibrium threshold multiple is dynamically updated. Repeat steps one through six until the equilibrium condition is met.

[0069] Therefore, the advantages of this invention are: starting with heavily loaded cells, identifying target overloaded users, analyzing the resource occupancy and signal-to-noise ratio of the current user's cell, and finding target cells with environments no worse than the current cell to implement load transfer, thus ensuring the overall performance efficiency of the system throughout the load transfer process. Furthermore, dynamically setting the load transfer threshold ensures efficient load balancing convergence, thereby guaranteeing the overall performance of the 5G system. Attached Figure Description

[0070] Figure 1 This is a schematic diagram of a process of the present invention;

[0071] Figure 2 This is a curve showing the load balancing convergence rate in the simulation experiment of this invention;

[0072] Figure 3 This is the overload curve of the maximum load cell in the simulation experiment of this invention;

[0073] Figure 4 This is a load change curve of the minimum load base station in the simulation experiment of this invention;

[0074] Figure 5 This is a graph showing the cell throughput iteration curve in the simulation experiment of this invention. Detailed Implementation

[0075] The technical solution of the present invention will be further described in detail below through embodiments and in conjunction with the accompanying drawings.

[0076] Example:

[0077] This embodiment presents a performance-guaranteed 5G load balancing method, such as... Figure 1 As shown, it includes the following steps:

[0078] Step 1: Obtain neighboring cells as target cells; specifically including:

[0079] Obtain n mutually neighboring cells as the target cell NBS n ={gNB1, gNB2, ..., gNB n}, obtain the load of each cell in the target cell {Cld1, Cld2, ..., Cld n The total number of physical resource blocks {PrT1, PrT2, ..., PrT} n}, the minimum signal-to-noise ratio within the cell coverage area {SNRmin1, SNRmin2, ..., SNRmin n} and maximum signal-to-noise ratio {SNRmax1, SNRmax2, ..., SNRmax} n}

[0080] Step 2: Calculate the physical resource block occupancy of each target cell and compare it with the equilibrium benchmark value of physical resource blocks of all cells to select the equilibrium target cells; the specific process includes:

[0081] (2-1): For NBS n ={gNB1, gNB2, ..., gNB n Each cell gNB in ​​} i Calculate its physical resource block occupancy.

[0082] OcuP i =ceil(PrT i *Cld i )

[0083] Where ceil() represents the ceiling function, i = 1, 2, ..., n;

[0084] Calculate the equilibrium benchmark value for all physical resource blocks in all cells.

[0085]

[0086] (2-2): Regarding NBS n ={gNB1, gNB2, ..., gNB n For each cell in the array, calculate the available space in its physical resource blocks.

[0087] LftP i =Prb Av -OcuP i

[0088] (2-3): This will satisfy LftP i Cells with a value greater than 0 are added to the target cell list for equalization. The number of cells is m, and the corresponding physical resource block space reserves are respectively

[0089] Step 3: Calculate overloaded cells and select users from the overloaded cells whose numbers match the number of overloaded users; the specific process includes:

[0090] (3-1): Calculate the HSet of the overloaded cell = NBS n -LSet, whose quantity is Nm HSet =nm;

[0091] (3-2): Select Nm quantities from HSet in descending order of throughput. HSet Users who are overloaded are recorded as such. Determine the cell to which the overloaded user belongs.

[0092] Set heavy user throughput

[0093] Signal-to-noise ratio

[0094] Bit error rate

[0095] Step 4: Calculate the performance guarantee value of each overloaded user relative to each equalization target cell. If the performance guarantee value of an overloaded user is not less than the set equalization threshold, the equalization target cell corresponding to the performance guarantee value is selected as a candidate cell. From the candidate cells, the one with the largest performance guarantee value is selected as the transfer target cell, and the overloaded user is transferred from its overloaded cell to the transfer target cell. The process includes:

[0096] (4-1): Calculate the decimal signal-to-noise ratio for overloaded users.

[0097]

[0098] Where SNR′ t =power(10, SNR) t / 10), t=1,2,…,Nm HSet power() represents the exponential function;

[0099] Calculate the bit error rate conversion value for overloaded users.

[0100]

[0101] Among them BER′ t =-In(5*BER) t ) / 1.5, ln() represents the natural logarithm function;

[0102] Calculate the median value.

[0103]

[0104] Where Mf t =log2(1+SNR) t ′ / BER t log2() represents the logarithmic function with base 2;

[0105] (4-2): Set the subcarrier bandwidth Wd Sc The number of subcarriers Nm occupied by a single physical resource block Prb Calculate the bandwidth of a single physical resource block.

[0106] Wd Prb =Wd Sc *Nm Prb / 1024

[0107] Calculate the amount of physical resource blocks occupied by each heavily loaded user.

[0108] UPrb t =ceil(Thr t / (Wd Prb *Mf t ));

[0109] (4-3): For a balanced target cell, calculate the minimum signal-to-noise ratio in decimal.

[0110] SNRmn k =power(10,SNRmin) k / 10)

[0111] Calculate the maximum signal-to-noise ratio in decimal.

[0112] SNRmx k =power(10, SNRmax) k / 10)

[0113] Calculate the median.

[0114]

[0115] Calculate the expected signal-to-noise ratio (SNR) of each cell in the target equalization cell.

[0116]

[0117] Where Log10 represents the logarithmic function with base 10, k = j, j+1, ..., j+m-1;

[0118] For each overloaded user Usr t Iterate through each gNB in ​​the target cell for equalization. k Calculate the performance guarantee value.

[0119]

[0120] Where exp() is an exponential function with the natural constant e as its base;

[0121] (4-4): Set the equalization threshold LB Thr , 0 < LB Thr <1, if Pfas t，k ≥LB Thr Then gNB k As candidate cells, they constitute the candidate set CaSt Lb ;

[0122] Determine the candidate set CaSt Lb Is it empty? If empty, cancel this user's username / data. t Load transfer; if not empty, select CaSt. Lb China has the largest Pfas t，k Value of the community gNB mxk and will User Usr t From the neighborhood Transferred to the community gNB mxk middle;

[0123] Update community gNB mxk Physical resource possession OcuP mxk =OcuP mxk +UPrb t Physical resource block space margin Lftp mxk =Lftp mxk -UPrb t ;

[0124] Update Usr t The residential community Physical resource possession Physical resource block space remaining

[0125] (4-5): In each user's User t Complete the traversal and update the load Cld for all cells. i =OcuP i / PrT i .

[0126] Step 5: Calculate the equilibrium factor for the target cell, set equilibrium conditions, and perform convergence judgment. If the equilibrium conditions are not met, repeat steps one through five until the equilibrium conditions are met. The specific process includes:

[0127] (5-1): Set an acceptable equilibrium state β0 and a convergence target value ε for the target cell NBS. n ={gNB1, gNB2, ..., gNB n} Calculate the load balancing factor

[0128]

[0129] (5-2): Determine whether the equilibrium condition is met.

[0130]

[0131] If the equilibrium condition is met, the step ends; if the equilibrium condition is not met, the convergence counter Δ = Δ + 1.

[0132] (5-3): Set the convergence level SL mx Determine whether Δ≤SL is satisfied. mx If yes, repeat steps one through five until the equilibrium condition is met; otherwise, proceed to step six.

[0133] Step Six: If the convergence count is exceeded and the load balancing condition is not met, dynamically update the load balancing threshold and repeat steps one through six until the load balancing condition is met, thus obtaining a load-balanced cell with guaranteed performance. The specific process includes:

[0134] Set equilibrium threshold to shrink variables If Δ>SL mx If the equilibrium condition is not met, the equilibrium threshold is dynamically updated. Repeat steps one through six until the equilibrium condition is met.

[0135] The following example illustrates the method of this embodiment. Taking n=5 as an example, the cell load of the 5G system is shown in Table 1:

[0136] Table 1 Business Activities

[0137]

[0138] The random user information for the community is shown in Table 2:

[0139] Table 2

[0140] residential area Throughput (Mbps) SNR BER Belonging to the community Usrl 4 2 0.07 Community 1 Usr2 2 5 0.02 Community 5 Usr3 1 6 0.01 Community 5

[0141] The basic data is shown in Table 3:

[0142] Table 3 Basic Data

[0143]

[0144]

[0145] A performance-guaranteed 5G load balancing method includes the following steps:

[0146] Step 1: Obtain 5 neighboring cells as target cells NBS5 = {gNB1, gNB2, gNB3, gNB4, gNB5}; as shown in Table 1, obtain the load, total number of physical resource blocks, minimum signal-to-noise ratio and maximum signal-to-noise ratio of each cell in the target cells.

[0147] Step 2: Calculate the physical resource block occupancy of each target cell, compare it with the equilibrium benchmark value of physical resource blocks of all cells, and select the equilibrium target cells.

[0148] (2-1): For NBS n ={gNB1, gNB2, ..., gNB n Each cell gNB in ​​} i Calculate its physical resource block occupancy.

[0149] OcuP i =ceil(PrT i *Cld i ) = {132, 69, 96, 129, 145}

[0150] Calculate the equilibrium benchmark value for all physical resource blocks in all cells.

[0151]

[0152] (2-2): Regarding NBS n ={gNB1, gNB2, ..., gNB n For each cell in the array, calculate the available space in its physical resource blocks.

[0153] LftP i =Prb Av -OcuP i ={-17, 46, 19, -14, -30}

[0154] (2-3): This will satisfy LftP i Cells with a value greater than 0 are added to the target cell list for equalization. The number of cells is m=2, and the corresponding physical resource block space reserves are respectively

[0155]

[0156] Step 3: Calculate the number of overloaded cells and select the same number of users from the overloaded cells as overloaded users;

[0157] (3-1): Calculation of overloaded cells

[0158] HSet=NBS n -LSet = {gNB1, gNB4, gNB5}, with a quantity of Nm HSet =nm=3;

[0159] (3-2): Select Nm quantities from HSet in descending order of throughput. HSet Users with a value of 3, as shown in Table 2, are denoted as overloaded users {Usr1, Usr2, ..., Usr3}. The cells to which the overloaded users belong are determined to be {gNB1, gNB2, gNB5}.

[0160] Step 4: Calculate the performance guarantee value of each overloaded user relative to each equalization target cell. If the performance guarantee value of the overloaded user is not less than the set equalization threshold, the equalization target cell corresponding to the performance guarantee value is taken as a candidate cell. The cell with the largest performance guarantee value is selected from the candidate cells as the transfer target cell, and the overloaded user is transferred from the overloaded cell to the transfer target cell.

[0161] (4-1): Calculate the decimal signal-to-noise ratio for overloaded users.

[0162] SNR HS = {1.58, 3.16, 3.98}

[0163] Calculate the bit error rate conversion value for overloaded users.

[0164] BER HS = {0.72, 1.5, 1.77}

[0165] Calculate the median value.

[0166]

[0167] (4-2): Set the subcarrier bandwidth Wd Sc =30kHz, number of subcarriers Nm occupied by a single physical resource block Prb =12, calculate the bandwidth of a single physical resource block.

[0168]

[0169] Calculate the amount of physical resource blocks occupied by each heavily loaded user.

[0170] UPrb t =ceil(Thr t / (Wd Prb *Mf t ))={23,12,6};

[0171] (4-3): Calculate the expected signal-to-noise ratio of each cell in the equalization target cell.

[0172]

[0173] For each overloaded user Usr t Iterate through each gNB in ​​the target cell for equalization. k Calculate the performance guarantee value;

[0174] First, for Usr1, which belongs to gNB1, iterate through each cell gNB in ​​the equalization target cell list LSet = {gNB2, gNB3}. k Calculate the performance guarantee value.

[0175]

[0176] (4-4): Set the equalization threshold LB Thr =0.2, if Pfas t，k ≥LB Thr Then gNB k As candidate cells, they constitute the candidate set CaSt Lb The candidate set is CaSt. Lb ={gNB2}, gNB mxk If the value is gNB2, then user Usr1 will be transferred from cell gNB1 to cell gNB2.

[0177] Update the physical resource occupancy of gNB2 in the updated cell: OcuP2 = OcuP2 + UPrb1 = 92, and the physical resource block space reserve: Lftp2 = Lftp2 - UPrb1 = 23;

[0178] Update the physical resource occupancy of the cell gNB1 where Usr1 is located: OcuP1 = OcuP1 - UPrb1 = 109, and the physical resource block space reserve: Lftp1 = Lftp1 + UPrb1 = 6.

[0179] After completing Usr1, proceed to complete {Usr2, Usr3}, with a performance guarantee value Pfas. t，k These are Usr2: {-1.55, -3.89}, Usr3: {-2.55, -4.89}, and CaSt. Lb All are empty, and none satisfy Pfas t，k ≥0.2, no load transfer is implemented.

[0180] (4-5): Update the load of all cells.

[0181]

[0182] Step 5: Calculate the equilibrium factor for the target cell, set the equilibrium conditions, and make a convergence decision. If the equilibrium conditions are not met, repeat steps 1 to 5 until the equilibrium conditions are met.

[0183] (5-1): Set an acceptable equilibrium state β0 and a convergence target value ε for the target cell NBS. n ={gNB1, gNB2, ..., gNB n} Calculate the load balancing factor.

[0184]

[0185] (5-2): Determine whether the equilibrium condition is met.

[0186]

[0187] If the equilibrium condition is met, the step ends; if the equilibrium condition is not met, the convergence counter Δ = Δ + 1 = 1, and Δ is initially 0.

[0188] (5-3): Set the convergence level SL mx Determine whether Δ≤SL is satisfied. mx If yes, repeat steps one through five until the equilibrium condition is met; otherwise, proceed to step six.

[0189] Step Six: Set the equilibrium threshold and shrink the variable If Δ>SLmx If the equilibrium condition is not met, the equilibrium threshold is dynamically updated. Repeat steps one through six until the equilibrium condition is met.

[0190] Simulation experiment:

[0191] The PALB method of this invention was simulated on a MATLAB platform along with the previous NSLB method. Basic data information is shown in Tables 1-3 above. A certain number of users were randomly selected and random services were configured. The results are as follows: Figures 2 to 5 As shown.

[0192] like Figure 2 The load balancing convergence process of the two methods is shown. NSLB converges after about 73 cycles, while PALB uses a dynamic load transfer threshold to select the resource space that is closest to the transferable load, resulting in higher convergence efficiency and taking about 53 cycles.

[0193] like Figure 3 The maximum load cell overload curves for the two algorithms are shown. It is evident that PALB deloads faster than NSLB, completing deload in approximately the 12th iteration, while NSLB takes approximately 24 iterations.

[0194] like Figure 4 The diagram shows the minimum load base station load change for the two algorithms. As a load transfer cell, the speed of load transfer determines the efficiency of the load balancing algorithm. PALB's load balancing speed is almost twice that of NSLB;

[0195] like Figure 5 The graphs show the cell performance changes for the two algorithms. NSLB, because it doesn't consider cell performance during load transfer, exhibits severe performance fluctuations, while PALB shows only minor oscillations. Since users are generally only allowed to transfer to target cells with better environments, the overall cell performance even shows a steady improvement; however, because NSLB converges slower than PALB... Figure 5 The number of convergence steps of PALB was selected for comparison.

[0196] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A performance guaranteed 5G load balancing method, characterized by: The method comprises the following steps, Step one: obtaining cells as target cells which are adjacent to each other; Step two: calculating the physical resource block occupancy of each target cell, and the difference between the physical resource block occupancy and the balance reference value of all cells to obtain the physical resource block space margin, and screening out cells with a physical resource block space margin greater than 0 as balance target cells; Step three: calculating the overload cells, and selecting the same number of users as the overload cells from the overload cells as overload users; Step four: calculating the overload user decimal signal-to-noise ratio and the bit error rate conversion value, and calculating the intermediate value; Calculating the single physical resource block bandwidth, and calculating the physical resource amount occupied by each overload user; Calculating the signal-to-noise ratio expectation value of each cell of the balance target cell; Calculating the performance guarantee value of each overload user relative to each balance target cell = min(exp(target cell space margin-overload cell occupied physical resource amount), 1) * (target cell signal-to-noise ratio expectation value-overload user signal-to-noise ratio); If the performance guarantee value of the overload user is not less than the set balance threshold, the balance target cell corresponding to the performance guarantee value is taken as a candidate cell, and the overload user is transferred from the overload cell to the transfer target cell selected from the candidate cell with the maximum performance guarantee value; Step five: calculating the balance factor of the target cell, setting the balance condition, and making a convergence decision, if the balance condition is not met, repeating steps one to five until the balance condition is met; Step six: if the convergence number is exceeded and the balance condition is not met, dynamically updating the balance threshold, repeating steps one to six until the balance condition is met, and obtaining the performance guaranteed load balancing cell.

2. The performance guaranteed 5G load balancing method of claim 1, wherein The step one of obtaining cells as target cells which are adjacent to each other specifically comprises: Obtain n cells as target cells NBS n = {gNB1, gNB2, …, gNB n}, obtain the load of each cell in the target cell {Cld1, Cld2, …, Cld n}, the total number of physical resource blocks {PrT1, PrT2, …, PrT n}, the minimum signal-to-noise ratio in the cell coverage range {SNRmin1, SNRmin2, …, SNRmin n}, and the maximum signal-to-noise ratio {SNRmax1, SNRmax2, …, SNRmax n}.

3. The performance guaranteed 5G load balancing method of claim 2, wherein The specific process of step two comprises: (2-1): For NBS n = {gNB1, gNB2,..., gNB n} for each cell gNB i , compute its physical resource block occupancy, OcuP i = ceil(PrT i * Cld i ) Wherein ceil() represents the ceiling function, i = 1, 2, …, n; Calculating the balance reference value of the physical resource block of all cells, (2-2): To NBS n = {gNB1, gNB2,..., gNB n For each cell in the set of {gNB1, gNB2,..., gNB , compute its spatial margin of physical resource blocks, LftP i = Prb Av - OcuP i (2-3): A cell satisfying LftP i is added to the balancing target cell list whose cell number is m, and the corresponding physical resource block space margin is respectively 4. The performance guaranteed 5G load balancing method of claim 3, wherein The specific process of step three comprises: (3-1): Accounting for heavy load cell HSet = NBS n - LSet, the number of Nm HSet = n - m; (3-2): Selecting Nm users from the HSet according to the principle of descending throughput, and recording them as heavy users HSet Determining the cell to which the heavy user belongs ​ Setting heavy user throughput signal-to-noise ratio bit error rate 5. The performance guaranteed 5G load balancing method of claim 4, wherein The specific process of step four comprises: (4-1): calculating the overload user decimal signal-to-noise ratio, wherein power() denotes the exponential function; Calculating the bit error rate conversion value of the overload user, where BER' = -ln(5*BER) / 1.5, ln() denotes the natural logarithm function; t t where BER' = -ln(5*BER) / 1.5, ln() denotes the natural logarithm function;​ Calculating the intermediate value, wherein log2() denotes the logarithm function with base 2; (4-2) : Set subcarrier bandwidth Wd Sc , number of subcarriers occupied by a single physical resource block Nm Prb , calculate the bandwidth of a single physical resource block, Wd Prb = Wd Sc *Nm Prb / 1024 Calculating the physical resource block amount occupied by each overload user, UPrb t = ceil(Thr t / (Wd Prb *Mf t )); (4-3): for the balance target cell, calculating the decimal minimum signal-to-noise ratio, SNRmin k = power(10, SNRmin k / 10) Calculating the decimal maximum signal-to-noise ratio, SNRmx k = power(10, SNRmax k / 10) Calculating the median value, Calculating the signal-to-noise ratio expectation value of each cell of the balance target cell, where Log10 denotes the logarithm function with base 10, k = j1, j2,..., j m ; For each overloaded user Usr t Iterate through each gNB in ​​the target cell for equalization. k Calculate the performance guarantee value. Wherein exp() is the exponential function with the natural constant e as the base; (4-4) : Setting the balance threshold LB Thr , 0 < LB Thr < 1, if Pfas t,k ≥ LB Thr , then the gNB k forms a candidate set CaSt Lb as a candidate cell; Determine the candidate set CaSt Lb Is it empty? If empty, cancel this user's username / data. t Load transfer; if not empty, select CaSt. Lb China has the largest Pfas t,k Value of the community gNB mxk and will User Usr t From the neighborhood Transferred to the community gNB mxk middle; updating a physical resource occupancy, OcuP, of a cell gNB mxk mxk mxk t mxk mxk t ;​​​​​​ Update Usr t Cell where it is located Physical resource occupancy Physical resource block space margin (4-5): In each user Usr t After the completion of the traversal, the load Cld of all cells is updated i = OcuP i / PrT i .

6. The performance guaranteed 5G load balancing method of claim 5, wherein The specific process of step five comprises: (5-1): Set acceptable equilibrium state β0, convergence target value ε, for target cell NBS n = {gNB1, gNB2,..., gNB n}, calculate load balancing factor (5-2): judging whether the balance condition is met, If the balance condition is met, ending the step; if the balance condition is not met, the convergence counter Δ = Δ + 1; (5-3) : Set the convergence series SL mx , determine whether Δ ≤ SL mx is satisfied, if yes, repeat steps 1 to 5 until the balance condition is satisfied, if no, go to step 6.

7. The performance guaranteed 5G load balancing method of claim 1, wherein The specific process of step six comprises: Setting an equalization threshold shrink variable If Δ > SL mx And the equalization condition is not met, then dynamically updating the equalization threshold Repeating steps one through six until the equalization condition is met.

Images