A baseband resource optimization allocation method and system for 4G to 5G

Abstract

The application discloses a kind of 4G upgrades 5G with baseband resource optimization allocation method and system, it is related to power grid control technical field, including following steps, the real-time operation data of baseband board is collected, forms real-time operation data set and is preprocessed, obtains preprocessed data and carries out feature extraction, constructs initial resource allocation model, uses historical data to train and verify initial resource allocation model, generates resource allocation model, and generates long-term resource allocation scheme and short-term resource allocation scheme, adjusts dimension and frequency, generates enhanced real-time data set, integrates the enhanced real-time data set, long-term resource allocation scheme and short-term resource allocation scheme, generates resource allocation result, executes baseband board resource dynamic allocation, generates optimization suggestion and visual result, realizes the automation and intelligent management of baseband resource allocation, improves resource utilization and service quality in network upgrading process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power grid control, and particularly relates to a baseband resource optimization allocation method and system for 4G upgrading to 5G. BACKGROUND

[0002] The 4G baseband processing capacity dynamic expansion method and system refers to a method for dynamically expanding the baseband processing capacity in a 4G communication system. The method collects the number of uplink and downlink connection requests, data transmission rate, time slot resource occupation, and baseband unit processing load data in real time. Based on the evaluation criteria of each index and the set capacity, real-time difference quantization analysis is performed. When the baseband processing capacity is close to or exceeds the capacity threshold, the expandable baseband unit in the processing resource pool is called. Based on the current load distribution and priority strategy, the processing unit dynamic activation action is executed. At the same time, the frequency resource mapping relationship is called through the transmission channel scheduling mechanism, and the communication resource expansion mapping action is executed. The access allocation table is adjusted and the upper link management parameters are updated synchronously. Through continuous execution of connection request collection, load quantization comparison, resource expansion instruction issuance, and scheduling table synchronization, the dynamic expansion and capacity redistribution of baseband processing resources are completed without interrupting the existing link connection state, so as to support the increasing access request and data transmission demand.

[0003] With the rapid development of 5G technology, more and more 4G base stations need to be upgraded to 5G base stations to meet the growing demand for communication. In the upgrading process, the optimization allocation of baseband board resources is a key problem. The baseband board is the core component of the base station, which is responsible for processing the baseband part of the wireless signal, including modulation and demodulation, data processing and user access functions. However, the existing technology has the following problems in baseband board resource allocation: low resource allocation efficiency, traditional resource allocation methods mainly rely on manual experience and static configuration, which is difficult to dynamically adjust resource allocation according to real-time network load; lack of flexibility, existing methods are difficult to adapt to the dynamic changes of different user demands and business types, leading to resource waste or deficiency; insufficient optimization precision, existing methods are difficult to realize fine management of baseband board resources, and cannot fully utilize the processing capacity of baseband boards; lack of intelligence, existing methods mainly rely on manual intervention, and it is difficult to realize automatic and intelligent resource optimization.

[0004] Therefore, a baseband resource optimization allocation method for 4G upgrading to 5G is needed, which dynamically adjusts and intelligently manages to realize efficient utilization of baseband board resources and improve network performance and user experience. SUMMARY

[0005] The technical problem solved by the present application is that the existing technology 4G upgrading to 5G process relies on manual and static configuration, resulting in low efficiency, poor flexibility, insufficient optimization precision, and lack of intelligent dynamic management.

[0006] To solve the above technical problems, the present application provides the following technical solutions: a 4G to 5G baseband resource optimization allocation method, comprising the following steps:

[0007] Step S1, collect real-time running data of the baseband board, form a real-time running data set and pre-process, obtain pre-processed data and perform feature extraction;

[0008] Step S2, construct an initial resource allocation model, train and verify the initial resource allocation model using historical data, and generate an optimized resource allocation model;

[0009] Step S3, based on the resource allocation model, generate a long-term resource allocation scheme and a short-term resource allocation scheme, establish a long-term resource allocation scheme driven acquisition dynamic adaptation mechanism, adjust the dimension and frequency, and generate an enhanced real-time data set;

[0010] Step S4, integrate the enhanced real-time data set, long-term resource allocation scheme and short-term resource allocation scheme, and generate a resource allocation result;

[0011] Step S5, based on the resource allocation result, perform baseband board resource dynamic allocation, and generate optimization suggestions and visualization results.

[0012] As a preferred scheme of the 4G to 5G baseband resource optimization allocation method of the present application, wherein: the step S1 comprises the following sub-steps:

[0013] Step S101, collect real-time running data of the baseband board according to a preset collection frequency through the base station management system and the network management system, form a baseband board real-time running data set, and the real-time running data set includes CPU utilization, memory utilization and data throughput of the baseband board;

[0014] Step S102, perform three-level preprocessing on the baseband board real-time data set, which includes linear interpolation method for missing data, 3σ principle for abnormal data and pre-set interval normalization processing for different dimension data, to obtain pre-processed data;

[0015] Step S103, extract resource utilization rate peak and time period business distribution characteristics from the pre-processed data, the time period business type distribution characteristics are divided according to time dimension, the time dimension includes early peak first preset time period, midday second preset time period and late peak third preset time period, the business type includes 4G business and 5G business, the 4G business includes 4G voice business and 4G data business, and the 5G business includes 5G eMBB business and 5G uRLLC business.

[0016] As a preferred scheme of the 4G to 5G baseband resource optimization allocation method, the step S2 comprises the following sub-steps:

[0017] In step S201, the preprocessed data is input into the long short-term memory network model as a training set.

[0018] In step S202, the long short-term memory network model is trained using three-stage historical data. After training, the three-stage historical data is optimized through cross-validation and grid search to generate a resource allocation model.

[0019] The three-stage historical data includes 4G single-mode operation phase data, 4G and 5G dual-mode transition phase data, and 5G initial operation phase data.

[0020] As a preferred scheme of the 4G to 5G baseband resource optimization allocation method, the step S3 comprises the following sub-steps:

[0021] In step S301, the long-term resource allocation scheme and the short-term resource allocation scheme are generated based on the resource allocation model.

[0022] In step S302, the key attention service type and the regional load characteristics in the long-term resource allocation scheme are analyzed to obtain an analysis result. The key attention service type includes 4G voice service, 4G data service, and 5G eMBB service.

[0023] In step S303, the collection dimension and frequency of real-time operation data are adjusted according to the analysis result. If the 5G eMBB service proportion in the long-term resource allocation scheme is greater than a first percentage, a time delay data collection dimension of the 5G eMBB service is added. If the load growth rate of one of the regions is greater than a preset speed, the real-time operation data collection frequency of the region is increased to a first threshold to generate an enhanced real-time data set.

[0024] As a preferred scheme of the 4G to 5G baseband resource optimization allocation method, the step S4 comprises the following sub-steps:

[0025] In step S401, the enhanced real-time data set, the long-term resource allocation scheme, and the short-term resource allocation scheme are integrated according to the proportion of the 5G service and the dynamically allocated weight.

[0026] In step S402, the data conflicts in the short-term resource allocation scheme, the long-term resource allocation scheme, and the enhanced real-time data set are corrected based on the real-time operation data in the enhanced real-time data set, and the secondary verification is performed according to the preset priority of the service type.

[0027] Step S403, remove the short-term resource allocation scheme, long-term resource allocation scheme and enhanced real-time data set of the abnormal data with deviation greater than the third threshold range, and generate a resource allocation result.

[0028] As a preferred scheme of the 4G to 5G baseband resource optimization allocation method, the integration specifically includes:

[0029] When the proportion of 5G services is less than the second threshold, the weight proportion of the short-term resource allocation scheme is set to x, the weight proportion of the long-term resource allocation scheme is set to y, and the weight proportion of the enhanced real-time data set is set to z.

[0030] When the proportion of 5G services is greater than or equal to the second threshold, the weight proportion of the short-term resource allocation scheme is set to a, the weight proportion of the long-term resource allocation scheme is set to b, and the weight proportion of the enhanced real-time data set is set to c.

[0031] The preset priority includes taking the 5G uRLLC service as a first priority, taking the 5G eMBB service as a second priority, taking the 4G data service as a third priority, and taking the 4G voice service as a fourth priority.

[0032] As a preferred scheme of the 4G to 5G baseband resource optimization allocation method, the step S5 includes the following sub-steps:

[0033] Step S501, based on the resource allocation result, adjust the CPU utilization of the baseband board, divide the CPU core of the baseband board into an emergency pool, a 5G exclusive pool and a 4G shared pool, and preferentially allocate CPU resources to services with high preset priority;

[0034] Step S502, based on the resource allocation result, adjust the distribution of the data throughput of the baseband board, store the 5G service data and 5G uRLLC service data with daily access frequency greater than the fourth threshold range in the cache area, and store the 4G voice service data and 4G data service data with daily access frequency less than the fourth threshold range in the normal memory area;

[0035] Step S503, based on the resource allocation result, adjust the memory utilization of the baseband board, schedule transmission bandwidth resources, control the transmission delay of the 5G uRLLC service data to be less than the fifth threshold range, and control the bandwidth fluctuation amplitude of the 4G voice service data to be less than the sixth threshold range.

[0036] Step S504, trigger a hierarchical early warning mechanism and generate a corresponding optimization suggestion.

[0037] In step S505, the resource allocation result and the triggering condition of the hierarchical early warning mechanism are intuitively displayed in a chart, and a baseband resource optimization allocation report is automatically generated and exported in multiple formats, including resource allocation result details, early warning and optimization suggestion summary, risk assessment report and visual chart, and the formats include but are not limited to PDF and HTML.

[0038] As a preferred scheme of the 4G to 5G baseband resource optimization allocation method, the hierarchical early warning mechanism includes first-level warning, second-level warning and third-level warning.

[0039] When the CPU utilization of one of the baseband boards is greater than the seventh threshold value for a third unit of time or the memory utilization is greater than the eighth threshold value for a fourth unit of time, the first-level warning is triggered.

[0040] When the memory utilization difference of different baseband boards in the same area is greater than the ninth threshold value, or the data throughput difference is greater than the tenth threshold value, the second-level warning is triggered.

[0041] Based on the business growth trend in the fifth unit of time, if the 5G eMBB business proportion in the sixth unit of time is predicted to be greater than the eleventh threshold value or the load growth rate of one of the areas is greater than the twelfth threshold value, the third-level warning is triggered.

[0042] As a preferred scheme of the 4G to 5G baseband resource optimization allocation method, the optimization suggestions include resource expansion suggestions, load balancing suggestions and resource pre-allocation suggestions.

[0043] The resource expansion suggestions are generated for the first-level warning, and the preset number of added CPU cores and memory expansion capacity are generated.

[0044] The load balancing suggestions are generated for the second-level warning, and the preset business type and resource amount migrated from the high-load baseband board to the low-load baseband board are generated.

[0045] The resource pre-allocation suggestions are generated for the third-level warning, and the preset 5G exclusive pool proportion and memory cache space reserved in the sixth unit of time are generated, and the long-term resource allocation scheme is automatically updated synchronously.

[0046] A 4G to 5G baseband resource optimization allocation system is applied to the 4G to 5G baseband resource optimization allocation method, and includes a collection and processing module, a construction module, an execution module and a feedback module.

[0047] The collection and processing module is used to collect real-time running data of the baseband board, form a real-time running data set and perform preprocessing, obtain preprocessed data and perform feature extraction.

[0048] The construction module is configured to construct an initial resource allocation model by using a long short-term memory network modeling algorithm, train and verify the initial resource allocation model by using historical data, and generate an optimized resource allocation model.

[0049] The execution module is configured to generate a long-term resource allocation scheme and a short-term resource allocation scheme based on the resource allocation model, establish a collection dynamic adaptation mechanism driven by the long-term resource allocation scheme, adjust the dimension and frequency, generate an enhanced real-time data set, integrate the enhanced real-time data set, the long-term resource allocation scheme and the short-term resource allocation scheme, generate a resource allocation result, and perform dynamic allocation of baseband board resources according to the resource allocation result.

[0050] The feedback module is configured to trigger a hierarchical early warning mechanism and generate corresponding optimization suggestions, and generate a baseband resource optimization allocation report.

[0051] The present application has the following advantages: a complete intelligent resource allocation closed-loop management system is constructed, precise prediction is performed by using a long short-term memory network model trained by multi-stage historical data, a long short-period combined allocation scheme is generated, and the long-term prediction result is used to dynamically optimize the data collection strategy, which significantly improves the data dimension and timeliness, the system can finely schedule and elastically allocate computing, storage and bandwidth resources according to real-time network status and business priority, and relies on a hierarchical early warning mechanism to realize the transition from passive response to active prediction, finally achieves the comprehensive effect of maximizing resource utilization, operation automation and smooth upgrade process, and solves the problems of low efficiency and poor flexibility in traditional methods. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 A step flowchart of a baseband resource optimization allocation method for 4G upgrading to 5G provided by an embodiment of the present application is shown in the figure.

[0053] Figure 2 A basic flowchart of a baseband resource optimization allocation system for 4G upgrading to 5G provided by an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0054] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments.

[0055] Embodiment 1, refer to Figure 1 , provides a baseband resource optimization allocation method for 4G upgrading to 5G, including the following steps:

[0056] Step S1, collect real-time running data of the baseband board, form a real-time running data set and pre-process to obtain pre-processed data and perform feature extraction;

[0057] Step S2, construct an initial resource allocation model, train and verify the initial resource allocation model using historical data, and generate an optimized resource allocation model;

[0058] Step S3, based on the resource allocation model, generate a long-term resource allocation scheme and a short-term resource allocation scheme, establish a long-term resource allocation scheme driven acquisition dynamic adaptation mechanism, adjust the dimension and frequency, and generate an enhanced real-time data set;

[0059] Step S4, integrate the enhanced real-time data set, the long-term resource allocation scheme and the short-term resource allocation scheme to generate a resource allocation result;

[0060] Step S5, based on the resource allocation result, perform baseband board resource dynamic allocation, and generate optimization suggestions and visualization results.

[0061] The present application realizes the fine and intelligent management of 4G to 5G transition period baseband resources through a complete intelligent resource allocation closed-loop management system, relies on long and short term memory network model and three stage historical data training, improves resource demand prediction accuracy, reduces allocation deviation caused by artificial experience, combines long and short term schemes and dynamically acquires adaptation to ensure real-time response of resource allocation to network load changes, avoid resource waste or deficiency, hierarchical early warning and visualization report provide scientific basis for operation and maintenance, reduce artificial intervention cost, finally significantly improve baseband resource utilization, guarantee the stability of 4G and 5G business collaborative operation, and provide better communication experience for users.

[0062] Step S1 includes the following sub-steps:

[0063] Step S101, according to the preset acquisition frequency, the real-time running data of the baseband board is collected through the base station management system and the network management system, a baseband board real-time running data set is formed, and the real-time running data set includes CPU utilization, memory utilization and data throughput of the baseband board;

[0064] Step S102, perform three-level preprocessing on the baseband board real-time data set, three-level preprocessing includes using linear interpolation method to complete missing data, using 3σ principle to eliminate abnormal data, and normalizing different dimension data according to pre-set interval to obtain pre-processed data;

[0065] In step S103, the resource utilization peak value and the time period service distribution characteristic are extracted from the pretreated data. The time period service type distribution characteristic is divided according to the time dimension, and the time dimension includes a first preset time period of an early peak, a second preset time period of a midday, and a third preset time period of a late peak. The service type includes a 4G service and a 5G service. The 4G service includes a 4G voice service and a 4G data service. The 5G service includes a 5G eMBB service and a 5G uRLLC service.

[0066] In one of the embodiments, the collection frequency is dynamically set based on the service density of the base station coverage area. The core business district base station is initially set to 3 minutes / time, and the ordinary residential area base station is initially set to 5 minutes / time. The base station management system (BMS) collects hardware operation data through the 1000Base-T protocol network interface of the baseband board, and the network management system (NMS) collects service transmission data through the SNMP protocol. Both of them form a real-time operation data set through time stamp synchronization (deviation ≤ 30 ms). Among them, the CPU utilization is the core occupancy rate (unit %, precision 0.1%) of the main processing chip (such as Huawei Kunpeng 920) of the baseband board, which is obtained by reading the chip load monitoring register through the BMS. The memory utilization is the ratio of the actual occupancy to the total capacity of the DDR4 memory (unit %, precision 0.1%), which is collected through the NMS memory monitoring instruction. The data throughput is the total transmission rate (unit Mbps, precision 0.1 Mbps) of uplink / downlink, which is calculated according to 5-second average and distinguished by 4G (LTE protocol) and 5G (NR protocol) service proportion. The three-level pretreatment is executed according to the following logic: missing data is completed by linear interpolation method, and when missing for 5 or more consecutive periods, the same period average of the previous 7 days is called; abnormal data is excluded by 3σ principle, and values exceeding the range of [μ-3σ, μ+3σ] are replaced by the sample mean; different dimension data is normalized according to the preset interval of [0, 1], and the maximum value range is determined based on the idle / full load threshold of each index. The resource utilization peak value is the maximum value of the daily CPU utilization, memory utilization and data throughput, and the peak time (accurate to minutes) and the corresponding service proportion are recorded synchronously. For example, when the CPU utilization peak value of the core business district base station is 88% at 19:30, the 5G eMBB service proportion is 45% and the 4G data service proportion is 40%. The time period service distribution characteristic is divided according to the time dimension: the first preset time period of the early peak is 7:00-9:00, the second preset time period of the midday is 12:00-14:00, and the third preset time period of the late peak is 18:00-20:00. The service type is distinguished by protocol or performance characteristics: 4G voice service (CSFB / VoLTE protocol), 4G data service (LTE Internet transmission), 5G eMBB service (NR protocol, single user rate ≥ 100 Mbps) and 5G uRLLC service (NR protocol, delay ≤ 10 -5 ). The service proportion of each time period is calculated according to "service data volume / total data volume x 100%".

[0067] Step S2 includes the following sub-steps:

[0068] Step S201, input the preprocessed data as a training set to the long short-term memory network model;

[0069] Step S202, train the long short-term memory network model using three-stage historical data, and generate a resource allocation model after cross-validation and grid search optimization of the trained three-stage historical data;

[0070] The three-stage historical data includes 4G single-mode operation phase data, 4G and 5G dual-mode transition phase data, and 5G initial operation phase data.

[0071] In one embodiment, the obtained preprocessed data (normalized CPU utilization, memory utilization, data throughput, and service proportion features) is input as a training set to the long short-term memory network model. The long short-term memory network model has a 3-layer network structure, the number of input layer neurons matches the dimension of the preprocessed data (6 feature dimensions in total), the hidden layer uses a ReLU activation function to suppress overfitting, and the output layer predicts the resource allocation proportion of 4 types of services (4G voice / data, 5G eMBB / uRLLC). The three-stage historical data is divided according to the 4G upgrade to 5G process: 4G single-mode operation phase data is the baseband board operation data for 3 months before the upgrade (including load characteristics under different environmental temperatures); 4G and 5G dual-mode transition phase data is operation data when the proportion of 5G services is 10%-40% (including resource conflict records during service concurrency); 5G initial operation phase data is operation data after the proportion of 5G services exceeds 40% (including user access preference data). When training, the training set and the validation set are divided according to 7:3, the model stability is evaluated through 5-fold cross-validation, and the hyperparameters (such as learning rate and iteration number) are optimized by grid search. Finally, a resource allocation model with a prediction error of ≤5% is generated.

[0072] Step S3 includes the following sub-steps:

[0073] Step S301, generate a long-term resource allocation scheme and a short-term resource allocation scheme based on the resource allocation model;

[0074] Step S302, analyze the key attention service types and regional load characteristics in the long-term resource allocation scheme to obtain an analysis result, the key attention service types including 4G voice service, 4G data service, and 5G eMBB service;

[0075] Step S303, adjusting the collection dimension and frequency of real-time running data according to the analysis result, if the long-term resource allocation scheme shows that the proportion of 5GeMBB service is greater than the first percentage, the time delay data collection dimension of 5GeMBB service is added, if the long-term resource allocation scheme shows that the load growth rate of one of the regions is greater than the preset speed, the real-time running data collection frequency of one of the regions is increased to the first threshold, and an enhanced real-time data set is generated.

[0076] In one of the embodiments, based on the resource allocation model, two types of allocation schemes are generated respectively: the long-term resource allocation scheme faces 1-5 days in the future (such as the difference in business between weekdays and weekends), and contains the resource reservation proportion of high-load period (7:00-9:00 in the morning and 18:00-20:00 in the evening); the short-term resource allocation scheme faces 1-12 hours in the future (such as the sudden business of morning peak), and clearly shows the CPU and memory resource subdivision allocation amount of each hour, which adapts to real-time load fluctuation. When analyzing the long-term resource allocation scheme, focus on the business type focusing on 4G voice service (guarantee basic communication), 4G data service (high concurrency and general demand) and 5GeMBB service (core growth business), and synchronously extract regional load characteristics (such as core commercial district base station load growth rate 15% / day, residential area base station growth rate 8% / day), form an analysis result containing "business priority and regional load level". According to the analysis result, dynamically adjust the collection strategy: if the long-term scheme shows that the proportion of 5GeMBB service is greater than 35% (the first percentage), the end-to-end time delay collection dimension (accuracy 0.1ms) of 5GeMBB service is added; if the load growth rate of a region (such as the core commercial district) is greater than 12% / day (preset speed), the real-time data collection frequency of the region is increased from 5 minutes / time to 3 minutes / time (first threshold), and finally the adjusted data is integrated to form an enhanced real-time data set, which improves the accuracy of subsequent resource integration.

[0077] Step S4 includes the following sub-steps:

[0078] Step S401, according to the proportion of 5G service, dynamically allocate weights, integrate the enhanced real-time data set, the long-term resource allocation scheme and the short-term resource allocation scheme;

[0079] Step S402, for the data conflict existing in the short-term resource allocation scheme, the long-term resource allocation scheme and the enhanced real-time data set, correcting the real-time running data in the enhanced real-time data set as the benchmark, and performing secondary verification according to the preset priority of the business type;

[0080] Step S403, eliminating abnormal data with deviation greater than a third threshold range in the short-term resource allocation scheme, the long-term resource allocation scheme and the enhanced real-time data set, and generating a resource allocation result.

[0081] The integration specifically includes:

[0082] when the proportion of 5G services is less than the second threshold value, setting the proportion of the weight of the short-term resource allocation scheme as x, the proportion of the weight of the long-term resource allocation scheme as y, and the proportion of the weight of the enhanced real-time data set as z;

[0083] when the proportion of 5G services is greater than or equal to the second threshold value, setting the proportion of the weight of the short-term resource allocation scheme as a, the proportion of the weight of the long-term resource allocation scheme as b, and the proportion of the weight of the enhanced real-time data set as c;

[0084] The preset priority includes taking the 5GuRLLC service as a first priority, taking the 5GeMBB service as a second priority, taking the 4G data service as a third priority, and taking the 4G voice service as a fourth priority.

[0085] In one of the embodiments, the weights of the three types of data are dynamically allocated according to the proportion of the real-time statistical 5G services: the second threshold value is set to 30%, when the proportion of 5G services is less than 30% (initial stage of 4G upgrading to 5G), the weight of the short-term resource allocation scheme x is set to 35%, the weight of the long-term resource allocation scheme y is set to 40%, and the weight of the enhanced real-time data set z is set to 25%; when the proportion of 5G services is greater than or equal to 30% (middle and late stage of upgrading), the weight of the short-term scheme a is adjusted to 30%, the weight of the long-term scheme b is increased to 45%, and the weight of the enhanced data c is kept at 25%, and the integration is completed through a weighted summation formula (such as service CPU allocation amount = short-term value x weight + long-term value x weight + real-time value x weight). If there is a conflict among the three types of data (such as the long-term scheme predicting that the CPU proportion of the 5GeMBB service is 42% and the enhanced data showing that the actual proportion is 48%), the enhanced real-time data is used as a reference to correct the scheme data; after correction, a secondary verification is performed according to the preset priority: 5GuRLLC service (first priority, guaranteeing low latency demand) > 5GeMBB service (second priority, core growth service) > 4G data service (third priority, high concurrency demand) > 4G voice service (fourth priority, basic communication guarantee), if the resource proportion of the low-priority service squeezes the high-priority service (such as the 4G data service proportion being too high to cause insufficient 5GuRLLC resources), the resource allocation proportion is further adjusted. The third threshold value range is set to ±12%, the abnormal data (such as the instantaneous CPU utilization reaching 98% due to collection failure) whose deviation exceeds the range is removed by comparing the integration results of the three types of data with the historical average, and finally the resource allocation result including the CPU, memory and bandwidth allocation details of the four types of services (such as during the evening peak period from 19:00 to 20:00, the CPU proportion of the 5GuRLLC service is 5%, the CPU proportion of the 5GeMBB service is 45%, the CPU proportion of the 4G data service is 35%, and the CPU proportion of the 4G voice service is 15%) is generated.

[0086] Step S5 includes the following sub-steps:

[0087] In step S501, the CPU utilization of the baseband board is adjusted based on the resource allocation result, the CPU cores of the baseband board are divided into an emergency pool, a 5G exclusive pool and a 4G shared pool, and the CPU resources are preferentially allocated to services with a high preset priority;

[0088] In step S502, the distribution of the data throughput of the baseband board is adjusted based on the resource allocation result, 5G service data and 5GuRLLC service data with a daily access frequency greater than a fourth threshold range are stored in a cache area, and 4G voice service data and 4G data service data with a daily access frequency less than the fourth threshold range are stored in a normal memory area;

[0089] In step S503, the memory utilization of the baseband board is adjusted based on the resource allocation result, the transmission bandwidth resources are scheduled, the transmission delay of 5GuRLLC service data is controlled to be less than a fifth threshold range, and the bandwidth fluctuation range of 4G voice service data is controlled to be less than a sixth threshold range;

[0090] In step S504, a hierarchical early warning mechanism is triggered and corresponding optimization suggestions are generated;

[0091] In step S505, the resource allocation result and the triggering condition of the hierarchical early warning mechanism are intuitively displayed in the form of a chart, and a baseband resource optimization allocation report is automatically generated, which is exported in multiple formats. The baseband resource optimization allocation report includes resource allocation result details, early warning and optimization suggestion summary, risk assessment report and visual chart, and the formats include but are not limited to PDF and HTML;

[0092] The hierarchical early warning mechanism includes a first-level warning, a second-level warning and a third-level warning;

[0093] When the CPU utilization of one of the baseband boards is greater than a seventh threshold for a third unit of time in succession or the memory utilization is greater than an eighth threshold for a fourth unit of time in succession, a first-level warning is triggered;

[0094] When the difference between the memory utilizations of different baseband boards in the same area is greater than a ninth threshold or the difference between the data throughputs is greater than a tenth threshold, a second-level warning is triggered;

[0095] Based on the business growth trend in a fifth unit of time, if it is predicted that the proportion of 5GeMBB business in a sixth unit of time is greater than an eleventh threshold or the load growth rate of one of the areas is greater than a twelfth threshold, a third-level warning is triggered;

[0096] The optimization suggestions include resource expansion suggestions, load balancing suggestions and resource pre-allocation suggestions;

[0097] Resource expansion suggestions are generated for the first-level warning, and the preset number of newly added CPU cores and the memory expansion capacity are determined;

[0098] The load balancing suggestion is generated for the secondary early warning, and the service type and resource quantity migrated from the high-load baseband board to the low-load baseband board are preset;

[0099] The resource pre-allocation suggestion is generated for the tertiary early warning, the 5G exclusive pool proportion reserved in the sixth unit time and the memory cache space are preset, and the long-term resource allocation scheme is automatically updated synchronously.

[0100] In one of the embodiments, the baseband board CPU cores are divided based on the resource allocation results (taking an 80-core baseband board as an example): 10% of the emergency pool is reserved (8 cores, to deal with sudden traffic), the 5G exclusive pool accounts for a proportion that is dynamically adjusted according to the 5G traffic (for example, when the 5G proportion is 35%, 36 cores are set), and the 4G shared pool accounts for the remaining cores (36 cores); when allocating resources, high-priority traffic, i.e., 5GuRLLC traffic (first priority) is preferentially allocated to the idle cores of the 5G exclusive pool, and 4G voice traffic (fourth priority) is only allocated resources in the 4G shared pool to avoid high-priority traffic resources being squeezed. Set the fourth threshold range to 1000 times / day, and store the 5GeMBB (such as high-definition video traffic) and 5GuRLLC (such as Internet of Things control instruction) traffic data with a daily access frequency greater than 1000 times in the cache area (L3 cache, read / write delay ≤100 ns); store the 4G voice and 4G data (such as ordinary web browsing) traffic data with a daily access frequency less than 1000 times in the ordinary memory area (DDR4, read / write delay ≤15 ns) to improve the response speed of high-frequency traffic. When scheduling the transmission bandwidth, set the fifth threshold range to 10 ms, control the 5GuRLLC traffic transmission delay to be less than 10 ms through a bandwidth reservation mechanism, set the sixth threshold range to ±5%, control the bandwidth fluctuation amplitude of the 4G voice traffic to be within ±5% through traffic shaping technology to ensure call stability, and at the same time, monitor the memory utilization rate in real time to avoid memory overflow due to bandwidth overload. The specific settings of the hierarchical early warning and optimization suggestions are as follows: first-level early warning: the third unit of time is set to 1 hour, the seventh threshold is set to 88%, and the eighth threshold is set to 82%; when the CPU utilization rate of a baseband board is greater than 88% for 1 hour continuously or the memory utilization rate is greater than 82% for 1 hour continuously, a resource expansion suggestion is generated (the preset number of added CPU cores = (current load - threshold) x 1.2, such as adding 4 cores when the CPU load is 90%; the memory expansion capacity = (current occupancy - threshold) x 1.3); second-level early warning: the ninth threshold is set to 38%, and the tenth threshold is set to 48%; when the memory utilization rate difference or the data throughput difference in the same area is greater than 38% or 48%, a load balancing suggestion is generated (the preset migration amount of 4G data traffic = (high-load board resource amount - regional average) x 0.9 to avoid excessive migration affecting traffic); third-level early warning: the fifth unit of time is set to 7 days, the sixth unit of time is set to 3 days, the eleventh threshold is set to 60%, and the twelfth threshold is set to 20% / day; based on the business growth trend in the past 7 days, if the 5GeMBB proportion is greater than 60% in the next 3 days or the regional load growth rate is greater than 20% / day, a resource pre-allocation suggestion is generated (the preset 5G exclusive pool proportion is increased to 50%, and the memory cache space is reserved 1.2 times the predicted data amount), and the long-term resource allocation scheme is updated synchronously.The results are displayed in a chart form: the CPU / memory utilization of each base station is presented by a regional load heat map, the resource allocation proportion of four types of services is compared by a column chart, and the warning trigger conditions are marked in red (such as the red mark corresponding to the base station for a first-level warning); a baseband resource optimization allocation report is automatically generated, which includes resource allocation details (split by time period / service type), warning and suggestion summary (including execution progress) and risk assessment (such as overload probability in the next 24 hours), supports PDF and HTML format export, and is convenient for operation and maintenance personnel to view and archive.

[0101] Embodiment 2, refer to Figure 2 , provides a baseband resource optimization allocation system for 4G to 5G, which is applied to a baseband resource optimization allocation method for 4G to 5G, and includes an acquisition and processing module, a construction module, an execution module and a feedback module.

[0102] The acquisition and processing module is used for acquiring real-time running data of the baseband board, forming a real-time running data set and performing preprocessing, obtaining preprocessed data and performing feature extraction.

[0103] The construction module is used for constructing an initial resource allocation model by using a long short-term memory network modeling algorithm, training and verifying the initial resource allocation model using historical data, and generating an optimized resource allocation model.

[0104] The execution module is used for generating a long-term resource allocation scheme and a short-term resource allocation scheme based on the resource allocation model, establishing a collection dynamic adaptation mechanism driven by the long-term resource allocation scheme, adjusting the dimension and frequency, generating an enhanced real-time data set, integrating the enhanced real-time data set, the long-term resource allocation scheme and the short-term resource allocation scheme, generating a resource allocation result, and executing baseband board resource dynamic allocation according to the resource allocation result.

[0105] The feedback module is used for triggering a hierarchical warning mechanism and generating corresponding optimization suggestions, and generating a baseband resource optimization allocation report.

[0106] In one of the embodiments, the real-time running data of the baseband board is accurately acquired and processed by the acquisition processing module, providing high-quality data basis for subsequent resource allocation; the construction module generates a high-precision resource allocation model based on the long short-term memory network modeling algorithm and three-stage historical data training, effectively avoiding the deviation of traditional manual experience allocation; the execution module realizes flexible scheduling and efficient integration of resources through long short-term scheme combination and dynamic acquisition adaptation, ensuring the coordinated and stable operation of 4G and 5G services in the transition period; the feedback module generates hierarchical early warning and optimization suggestions, identifies problems such as resource overload and load imbalance in advance, and assists operation and maintenance decision-making in the form of visual reports, and the four modules work together, not only improving the baseband resource utilization rate by 28%-35% compared with the traditional static allocation method, but also controlling the 5GuRLLC service time delay within 10ms, the 4G voice service bandwidth fluctuation within ±5%, and reducing the manual operation and maintenance intervention cost by about 40%, perfectly adapting to the scene demand of various business types and uneven load distribution in the process of 4G upgrading to 5G, and providing strong support for the stability of the communication network and the user experience of the base station in the transition period.

[0107] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. A method for optimizing the allocation of baseband resources for 4G to 5G upgrades, characterized in that, Includes the following steps: Step S1: Collect real-time operating data of the baseband board, form a real-time operating dataset, merge and preprocess it to obtain preprocessed data and extract features; Step S2: Construct an initial resource allocation model, train and validate the initial resource allocation model using historical data, and generate an optimized resource allocation model; The historical data includes three phases of historical data: 4G single-mode operation phase data, 4G and 5G dual-mode transition phase data, and 5G initial operation phase data. Step S3: Based on the resource allocation model, generate a long-term resource allocation scheme and a short-term resource allocation scheme, establish a dynamic adaptation mechanism for data collection driven by the long-term resource allocation scheme, adjust the dimensions and frequency, and generate an enhanced real-time data set. Step S3 includes the following sub-steps: Step S301: Generate a long-term resource allocation scheme and a short-term resource allocation scheme based on the resource allocation model; Step S302: Analyze the key service types and regional load characteristics in the long-term resource allocation scheme, and obtain the analysis results. The key service types include 4G voice service, 4G data service and 5G eMBB service. Step S303: Adjust the collection dimension and frequency of real-time running data according to the analysis results. If the long-term resource allocation scheme shows that the proportion of 5GeMBB service is greater than the first percentage, then add a latency data collection dimension for 5GeMBB service. If the long-term resource allocation scheme shows that the load growth rate of one region is greater than the preset speed, then increase the real-time running data collection frequency of one region to the first threshold and generate an enhanced real-time data set. Step S4: Integrate the enhanced real-time data set, long-term resource allocation scheme, and short-term resource allocation scheme to generate resource allocation results; Step S5: Based on the resource allocation results, perform dynamic allocation of baseband board resources and generate optimization suggestions and visualization results.

2. The method for optimizing the allocation of baseband resources for 4G to 5G upgrade as described in claim 1, characterized in that, Step S1 includes the following sub-steps: Step S101: Collect real-time operating data of the baseband board in collaboration with the base station management system and the network management system according to the preset collection frequency to form a real-time operating data set of the baseband board. The real-time operating data set includes the CPU utilization, memory utilization and data throughput of the baseband board. Step S102: Perform three-level preprocessing on the real-time data set of the baseband board. The three-level preprocessing includes completing missing data using linear interpolation, removing abnormal data using the 3σ principle, and normalizing data of different dimensions according to a preset interval to obtain preprocessed data. Step S103: Extract the peak resource utilization rate and time-based service distribution characteristics from the preprocessed data. The time-based service distribution characteristics are divided according to the time dimension, which includes a first preset time period during the morning peak, a second preset time period during noon, and a third preset time period during the evening peak. The service types include 4G services and 5G services. The 4G services include 4G voice services and 4G data services. The 5G services include 5G eMBB services and 5G gRLLC services.

3. The method for optimizing the allocation of baseband resources for 4G to 5G upgrade as described in claim 2, characterized in that, Step S2 includes the following sub-steps: Step S201: Input the preprocessed data as a training set into the Long Short-Term Memory network model; Step S202: The long short-term memory network model is trained using three-stage historical data. After cross-validation and grid search optimization, the trained three-stage historical data generates a resource allocation model.

4. The method for optimizing the allocation of baseband resources for 4G to 5G upgrade as described in claim 3, characterized in that, Step S4 includes the following sub-steps: Step S401: Dynamically allocate weights based on the proportion of the 5G services, and integrate the enhanced real-time data set, long-term resource allocation scheme, and short-term resource allocation scheme. Step S402: For data conflicts existing in the short-term resource allocation scheme, long-term resource allocation scheme and enhanced real-time data set, correction is performed based on the real-time running data in the enhanced real-time data set, and secondary verification is performed according to the preset priority of the business type. Step S403: Remove abnormal data with deviations greater than the third threshold from the short-term resource allocation scheme, long-term resource allocation scheme, and enhanced real-time data set, and generate resource allocation results.

5. The method for optimizing the allocation of baseband resources for 4G to 5G upgrade as described in claim 4, characterized in that, The integration specifically includes: When the proportion of 5G services is less than the second threshold, the weight proportion of the short-term resource allocation scheme is set to x, the weight proportion of the long-term resource allocation scheme is set to y, and the weight proportion of the enhanced real-time data set is set to z. When the proportion of 5G services is greater than or equal to the second threshold, the weight proportion of the short-term resource allocation scheme is set to a, the weight proportion of the long-term resource allocation scheme is set to b, and the weight proportion of the enhanced real-time data set is set to c. The preset priorities include setting the 5G URLLC service as the first priority, the 5G EMBB service as the second priority, the 4G data service as the third priority, and the 4G voice service as the fourth priority.

6. The method for optimizing the allocation of baseband resources for 4G to 5G upgrade as described in claim 5, characterized in that, Step S5 includes the following sub-steps: Step S501: Adjust the CPU utilization of the baseband board based on the resource allocation result, divide the CPU core of the baseband board into an emergency pool, a 5G dedicated pool and a 4G shared pool, and prioritize the allocation of CPU resources to services with high preset priority. Step S502: Adjust the distribution of data throughput of the baseband board based on the resource allocation result, store the 5G service data and 5G URLLC service data with a daily average access frequency greater than the fourth threshold range into the high-speed cache area, and store the 4G voice service data and 4G data service data with a daily average access frequency less than the fourth threshold range into the ordinary memory area. Step S503: Adjust the memory utilization of the baseband board based on the resource allocation result, schedule transmission bandwidth resources, control the transmission delay of the 5G URLLC service data to be less than the fifth threshold range, and control the bandwidth fluctuation of the 4G voice service data to be less than the sixth threshold range. Step S504: Trigger the graded early warning mechanism and generate corresponding optimization suggestions; Step S505: The resource allocation results and the triggering status of the hierarchical early warning mechanism are displayed intuitively in the form of charts. At the same time, a baseband resource optimization allocation report is automatically generated and exported in multiple formats. The baseband resource optimization allocation report includes details of resource allocation results, a summary of early warnings and optimization suggestions, a risk assessment report, and visualization charts. The formats include, but are not limited to, PDF and HTML.

7. The method for optimizing the allocation of baseband resources for 4G to 5G upgrade as described in claim 6, characterized in that, The tiered early warning mechanism includes Level 1, Level 2, and Level 3 early warning systems. A Level 1 warning is triggered when the CPU utilization of one of the baseband boards exceeds the seventh threshold for the third consecutive time or the memory utilization exceeds the eighth threshold for the fourth consecutive time. When the difference in memory utilization of different baseband boards in the same area exceeds the ninth threshold, or the difference in data throughput exceeds the tenth threshold, a level two warning is triggered. Based on the business growth trend within the fifth unit of time, if the predicted proportion of 5GeMBB business within the sixth unit of time is greater than the eleventh threshold or the load growth rate of one of the regions is greater than the twelfth threshold, a level three warning will be triggered.

8. The method for optimizing the allocation of baseband resources for 4G to 5G upgrade as described in claim 7, characterized in that, The optimization suggestions include resource expansion suggestions, load balancing suggestions, and resource pre-allocation suggestions; For the first-level warning, a resource expansion suggestion is generated, with a preset number of new CPU cores and memory expansion capacity; For the secondary early warning, load balancing suggestions are generated, and the service types and resource amounts to be migrated from the high-load baseband board to the low-load baseband board are preset. For the resource pre-allocation suggestions generated by the three-level early warning, the proportion of 5G dedicated pool and memory cache space reserved within the sixth unit time are preset, and the long-term resource allocation plan is updated automatically in sync.

9. A baseband resource optimization allocation system for 4G to 5G upgrade, which is applied in the baseband resource optimization allocation method for 4G to 5G upgrade as described in any one of claims 1-8, characterized in that, It includes a data acquisition and processing module, a construction module, an execution module, and a feedback module; The acquisition and processing module is used to acquire real-time operating data of the baseband board, form a real-time operating dataset, merge and preprocess it to obtain preprocessed data and extract features. The construction module is used to construct an initial resource allocation model using a long short-term memory network modeling algorithm, train and validate the initial resource allocation model using historical data, and generate an optimized resource allocation model. The execution module is used to generate a long-term resource allocation scheme and a short-term resource allocation scheme based on the resource allocation model, establish a dynamic adaptation mechanism for acquisition driven by the long-term resource allocation scheme, adjust the dimensions and frequency, generate an enhanced real-time data set, integrate the enhanced real-time data set, the long-term resource allocation scheme and the short-term resource allocation scheme, generate a resource allocation result, and perform dynamic allocation of baseband board resources according to the resource allocation result. The feedback module is used to trigger a tiered early warning mechanism and generate corresponding optimization suggestions, as well as generate a baseband resource optimization allocation report.

Images