A multiplexing method and system for improving the spectrum efficiency of 5G mobile communications

Abstract

The application discloses a kind of multiplexing method and system for improving 5G mobile communication spectrum efficiency, belong to dynamic spectrum multiplexing communication technical field.The method includes: collecting real-time information, predicting spectrum availability and inter-band interference using neural network model, generating dynamic spectrum multiplexing strategy, issuing strategy and adjusting parameters and continuous monitoring and optimization.The application adopts intelligent prediction model of fusion MLP and LSTM network, provides spectrum availability and inter-band interference prediction, and generates prediction confidence information;Adaptive multi-objective intelligent optimization algorithm is used, and the algorithm can dynamically adjust optimization weight according to real-time network;Through the intelligent reconstruction capability of 5G base station, local optimality is judged in combination with local real-time information;And using online learning mechanism realizes continuous monitoring and closed-loop optimization.The application can realize the dynamic intelligent management of 5G network spectrum resources, and significantly improve the spectrum efficiency of 5G mobile communication.

Description

Technical Field

[0001] The present invention relates to the field of dynamic spectrum multiplexing communication technology, and in particular to a multiplexing method and system for improving the spectrum efficiency of 5G mobile communications. Background Art

[0002] With the rapid development and large-scale deployment of 5G mobile communications technology, demand for high bandwidth, low latency, and massive connections is growing across all industries. As critical infrastructure supporting the development of the digital economy, the performance of 5G networks depends directly on the efficient use of limited spectrum resources. Therefore, maximizing spectrum efficiency to meet growing service demands has become a core issue in the mobile communications field.

[0003] Existing spectrum resource management methods currently face widespread adaptability and efficiency bottlenecks when addressing the complex and volatile 5G network environment. Traditional static or semi-static spectrum allocation mechanisms struggle to fully capture instantaneous channel variations and dynamic user demands, resulting in suboptimal spectrum resource utilization. Furthermore, high-density deployment and flexible resource reuse also introduce significant inter-band interference, impacting overall network performance and user experience. Summary of the Invention

[0004] To solve the above problems, the present invention provides a multiplexing method and system for improving the spectrum efficiency of 5G mobile communications. It adopts an intelligent prediction model that integrates MLP and LSTM networks to generate spectrum availability prediction, inter-band interference prediction and confidence information; uses an adaptive multi-objective intelligent optimization algorithm combined with confidence information to generate dynamic strategies and dynamically adjust the optimization weights according to the real-time network; intelligently reconstructs the strategies through 5G base stations combined with local information; and uses an online learning mechanism to achieve continuous monitoring and closed-loop optimization, which can realize refined, dynamic and intelligent management of 5G network spectrum resources, and significantly improve the spectrum efficiency of 5G mobile communications and user service quality.

[0005] The above objectives can be achieved through the following solutions:

[0006] A reuse method for improving the spectrum efficiency of 5G mobile communications, comprising collecting real-time information of multiple 5G base stations, wherein the real-time information includes time slot channel occupancy status, user service demand information, and a frequency resource priority mapping table preset by a network management system; based on the real-time information, using a neural network model to predict spectrum availability and inter-band interference to generate a prediction result; generating a dynamic spectrum reuse strategy based on the prediction result and the preset frequency resource priority mapping table; wherein the dynamic spectrum reuse strategy includes a joint scheduling scheme of time division multiplexing and frequency division multiplexing; sending the dynamic spectrum reuse strategy to the corresponding 5G base station to adjust the spectrum usage parameters; continuously monitoring the actual use effect of the adjusted spectrum and updating the neural network model in real time to optimize the spectrum reuse strategy.

[0007] Optionally, the real-time information collection of multiple 5G base stations includes: collecting the instantaneous signal-to-noise ratio, time slot occupancy historical data and cyclic prefix interference level of each 5G base station; obtaining the uplink and downlink service traffic, delay tolerance and multi-band parallel transmission capability reported by the user terminal; and the network management system receiving the frequency resource priority mapping table determined according to the user's contracted service level agreement and service type.

[0008] Optionally, the use of a neural network model to predict spectrum availability and inter-band interference includes: constructing a neural network model including an MLP network and an LSTM network; the MLP network in the neural network model extracts and fuses the static features and nonlinear relationships of real-time information to generate static feature data; the LSTM network in the neural network model captures the time series features of real-time information to generate dynamic change patterns; and generating prediction results by deeply fusing the static feature data and dynamic change patterns, wherein the prediction results are spectrum availability prediction data, inter-band interference prediction data and prediction confidence information, and the prediction confidence information is a quantitative assessment of the reliability or uncertainty of the spectrum availability prediction data and the inter-band interference prediction data.

[0009] Optionally, the generation of a dynamic spectrum reuse strategy includes: based on the spectrum availability prediction data and the inter-band interference prediction data, combined with the real-time time slot channel occupancy status, constructing a spectrum reuse decision space, and quantifying the reuse interference risk of each idle resource block or low-interference resource block; according to the preset frequency resource priority mapping table, user service demand information and prediction confidence information, using an adaptive intelligent optimization algorithm to determine the optimal spectrum resource block allocation scheme within the spectrum reuse decision space; based on the optimal spectrum resource block allocation scheme, collaboratively adjusting the beamforming parameters and resource block scheduling scheme of the 5G base station to generate a dynamic spectrum reuse strategy.

[0010] Optionally, sending the dynamic spectrum reuse strategy to the corresponding 5G base station includes: the network management system sends the dynamic spectrum reuse strategy to the corresponding 5G base station in real time through control signaling according to the real-time load status of the base station and the geographical distribution information of the users; the corresponding 5G base station intelligently reconstructs the resource block mapping and allocation strategy of the uplink and downlink according to the received dynamic spectrum reuse strategy; the corresponding 5G base station adaptively adjusts the spectrum usage parameters according to the reconstructed resource block mapping and allocation strategy in combination with the dynamic spectrum reuse strategy.

[0011] Optionally, the continuous monitoring of the actual use effect of the adjusted spectrum and real-time updating of the neural network model include: continuously collecting and analyzing the user experience quality indicators, spectrum utilization and actual interference level of each 5G base station after adjustment to generate performance feedback data; based on the deviation analysis between the performance feedback data and the prediction results, using an online learning mechanism to update the neural network model in real time and fine-tune the parameters; based on the updated neural network model, adaptively re-evaluate and iteratively optimize the dynamic spectrum reuse strategy to continuously improve spectrum efficiency and user service quality.

[0012] Optionally, determining the optimal spectrum resource block allocation scheme within the spectrum reuse decision space includes: the adaptive intelligent optimization algorithm uses minimizing overall network interference, maximizing system throughput and ensuring high-priority user service quality as the target optimization function to search for the best within the spectrum reuse decision space; the adaptive intelligent optimization algorithm adjusts the optimization weights in real time according to real-time network load changes, user mobility and channel environment dynamics to generate the optimal spectrum resource block allocation scheme.

[0013] Optionally, based on the static feature data and dynamic change rules, generating a prediction result includes: the neural network model deeply fuses the static feature data and the dynamic change rules through an attention mechanism or a gated recurrent unit to generate a deep fusion result; based on the deep fusion result, generating spectrum availability prediction data and inter-band interference prediction data; performing a prediction confidence evaluation on the spectrum availability prediction data and the inter-band interference prediction data to generate prediction confidence information.

[0014] Optionally, the intelligent reconstruction of the resource block mapping and allocation strategy for the uplink and downlink includes: the corresponding 5G base station performs a local optimality judgment based on the received dynamic spectrum reuse strategy, combined with the real-time channel quality, local user service queue status and neighboring cell interference information; based on the local optimality judgment, the corresponding 5G base station dynamically adjusts the resource block mapping and allocation strategy to generate a resource block mapping and allocation strategy that adapts to the current network environment.

[0015] Based on the same inventive concept, the application also provides a multiplexing system for improving the spectrum efficiency of 5G mobile communication, the system comprising:

[0016] a real-time information collection module for collecting real-time information of a plurality of 5G base stations;

[0017] a spectrum prediction module for predicting spectrum availability and inter-band interference based on the real-time information using a neural network model to generate a prediction result;

[0018] a dynamic strategy generation module for generating a dynamic spectrum multiplexing strategy according to the prediction result and a preset frequency resource priority mapping table, wherein the dynamic spectrum multiplexing strategy comprises a joint scheduling scheme of time division multiplexing and frequency division multiplexing;

[0019] a strategy deployment and parameter adjustment module for issuing the dynamic spectrum multiplexing strategy to the corresponding 5G base stations and adjusting spectrum usage parameters;

[0020] a performance monitoring and optimization module for continuously monitoring the actual usage effect of the adjusted spectrum and updating the neural network model in real time to optimize the spectrum multiplexing strategy.

[0021] Compared with the prior art, the application has the following advantages:

[0022] 1. The application innovatively combines MLP network and LSTM network and deeply integrates them by introducing attention mechanism or gated recurrent unit, which not only accurately predicts spectrum availability and inter-band interference, but also further provides prediction confidence evaluation. This significantly improves the accuracy and reliability of the prediction result, enabling the system to make decisions based on more comprehensive information when formulating the dynamic spectrum multiplexing strategy, effectively avoiding resource waste or interference intensification caused by prediction uncertainty, thereby improving decision quality and system robustness;

[0023] 2. Unlike traditional single or fixed weight optimization methods, the application uses an adaptive multi-objective intelligent optimization algorithm that can adjust the optimization weight in real time according to real-time network load changes, user mobility, and channel environment dynamics, achieving dynamic balance among multiple objectives such as minimizing overall network interference, maximizing system throughput, and ensuring high-priority user service quality. In addition, combined with the local intelligent reconstruction capability of the 5G base station, the algorithm uses real-time channel quality, local user traffic queue state, and neighboring cell interference information to make local optimality judgments, achieving fine and dynamic allocation of spectrum resources, and significantly improving overall spectrum utilization efficiency and user service quality.

[0024] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0026] Figure 1 This is a flow chart of a multiplexing method for improving 5G mobile communication spectrum efficiency according to an embodiment of the present invention.

[0027] Figure 2 This is an area diagram of resource block availability changing over time according to an embodiment of the present invention.

[0028] Figure 3 This is a convergence line diagram of performance prediction deviation under online learning according to an embodiment of the present invention.

[0029] Figure 4 3 is a schematic diagram of the Pareto front of multi-objective optimization according to an embodiment of the present invention.

[0030] Figure 5 This is a structural diagram of a multiplexing system for improving 5G mobile communication spectrum efficiency according to an embodiment of the present invention. DETAILED DESCRIPTION

[0031] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0032] Reference Figure 1The application provides a multiplexing method for improving the spectrum efficiency of 5G mobile communication, which adopts an intelligent prediction model combining MLP and LSTM network to generate spectrum availability prediction, inter-band interference prediction and confidence information; an adaptive multi-objective intelligent optimization algorithm combined with confidence information is used to generate a dynamic strategy and optimize the weight according to real-time network dynamics; the strategy is intelligently reconstructed by 5G base stations combined with local information; and an online learning mechanism is used to realize continuous monitoring and closed-loop optimization, which can realize fine and dynamic intelligent management of 5G network spectrum resources, significantly improve the spectrum efficiency of 5G mobile communication and user service quality.

[0033] The method of the embodiment specifically comprises:

[0034] Collecting real-time information of a plurality of 5G base stations, wherein the real-time information includes time slot channel occupation state, user service demand information and a frequency resource priority mapping table preset by a network management system;

[0035] Specifically, in a 5G mobile communication network, network devices with data collection capability, such as base stations, core network elements or special detection devices, are deployed to collect multi-dimensional data in the coverage area of each 5G base station in a periodic or event-triggered manner. The time slot channel occupation state can include channel quality indicators at the current time or in a certain period, such as reference signal received power, reference signal received quality, etc. The user service demand information can reflect the real-time appeal of users in terms of bandwidth, delay, reliability, etc., such as obtained through service traffic statistics or application layer feedback. The frequency resource priority mapping table is pre-configured by the network management system according to the user service level agreement or service type to distinguish the service importance of different users.

[0036] Based on the real-time information, a neural network model is used to predict spectrum availability and inter-band interference to generate a prediction result;

[0037] Specifically, the collected real-time information is input into a pre-trained neural network model. The neural network model can learn and analyze the complex patterns and correlations in the real-time information, and then predict the available state of each frequency band or resource block in the future period, such as which resources are idle, low utilization or can be multiplexed. At the same time, the model can also predict the degree of mutual interference that may be generated when some spectrum resources are multiplexed, and quantify the interference level between frequency bands. The prediction result is output in the form of structured data, including numerical estimates of spectrum availability and inter-band interference.

[0038] According to the prediction result and the preset frequency resource priority mapping table, a dynamic spectrum multiplexing strategy is generated;

[0039] Specifically, the spectrum prediction results and the frequency resource priority mapping table are used to make decisions through a strategy generation module. This module will consider how to maximize the reuse efficiency of spectrum resources and minimize interference under the premise of ensuring the service experience of high-priority users. The decision-making process will evaluate different spectrum allocation schemes, adjust the allocation of resource blocks, determine the direction or power allocation of beamforming, and form a set of specific operation instructions, i.e., dynamic spectrum reuse strategy. The dynamic spectrum reuse strategy aims to achieve the optimal balance of network performance and includes a joint scheduling scheme of time division multiplexing and frequency division multiplexing.

[0040] The dynamic spectrum reuse strategy is issued to the corresponding 5G base station to adjust the spectrum usage parameters.

[0041] Specifically, the generated dynamic spectrum reuse strategy is distributed to the corresponding 5G base station in the network that needs to adjust the spectrum usage parameters through the corresponding control interface or signaling mechanism by the network management system. After receiving the strategy, the base station will immediately or at a specific time point adjust its own spectrum usage parameters according to the strategy instructions, reconfigure the carrier frequency of certain spectrum bands, adjust the scheduling priority of resource blocks, or change the beamforming mode of the antenna, to achieve the spectrum reuse target planned in the strategy.

[0042] The actual use effect of the adjusted spectrum is continuously monitored and the neural network model is updated in real time to optimize the spectrum reuse strategy.

[0043] Specifically, after the strategy is issued and the parameters are adjusted, the system will continuously monitor the actual running state of the 5G base station and the network performance. This includes real-time collection of adjusted user actual experience indicators, network throughput, latency, and actual interference situation data. These monitoring data are fed back to the neural network model for evaluating the accuracy of the model prediction and the effect of the strategy execution. If the monitoring data shows deviation or there is room for further optimization, the neural network model will be updated in real time or adjusted in a small range based on the new data, thereby improving its prediction ability. Based on the updated model, the system will re-evaluate and iteratively optimize the current spectrum reuse strategy, forming a closed-loop control system to ensure continuous improvement of spectrum efficiency.

[0044] The comprehensive collection and processing of real-time information of 5G base stations, the prediction of spectrum availability and inter-band interference based on deep neural network model, the generation of dynamic spectrum reuse strategy based on prediction results and user priority, the issuance of the strategy to the base station for parameter adjustment, and the realization of closed-loop optimization through continuous monitoring and real-time model updating can realize fine and dynamic intelligent management and collaborative control of 5G network spectrum resources, significantly improving the spectrum efficiency, network performance, and user service quality of 5G mobile communication.

[0045] Optionally, collecting real-time information of multiple 5G base stations includes:

[0046] Collect the instantaneous signal-to-noise ratio, historical data of time slot occupancy, and cyclic prefix interference level of each 5G base station;

[0047] Specifically, at a 5G base station, real-time measurement and processing of received downlink signals allows the instantaneous signal-to-noise ratio (SNR) to be determined. This SNR is a key metric that compares signal strength with background noise and interference levels, directly reflecting channel quality. The base station also measures historical timeslot occupancy data and cyclic prefix interference levels to comprehensively assess channel status and interference.

[0048] For example, analysis of continuously received signals from a 5G base station deployed in an area densely populated with tall buildings revealed significant fluctuations in its instantaneous signal-to-noise ratio over specific time periods. Furthermore, historical data on the base station's time slot occupancy showed peaks, and the cyclic prefix interference level increased, indicating that wireless resources in the area were limited and potential interference existed.

[0049] Obtain uplink and downlink service traffic, delay tolerance, and multi-band parallel transmission capabilities reported by user terminals;

[0050] Specifically, user terminals periodically report their uplink and downlink traffic data to the connected 5G base station based on their data transmission activity. This directly reflects the user's instantaneous demand for network bandwidth. For different types of services, the user's tolerance for transmission latency is also obtained. For example, real-time voice or video services typically require extremely low latency. Furthermore, the user terminal's multi-band parallel transmission capabilities are collected, such as the number of concurrent frequency bands supported or its carrier aggregation capability, which reflects the user terminal's ability to simultaneously utilize multiple frequency bands for data transmission.

[0051] For example, a user terminal that is conducting an online video conference will continuously report a high downlink service traffic demand, indicate that its service has a low tolerance for delay, and report that it has the ability to transmit data in parallel on two frequency bands.

[0052] The network management system receives a frequency resource priority mapping table determined according to the service level agreement signed by the user and the service type.

[0053] Specifically, the network management system, as the network's central control entity, receives and maintains a frequency resource priority mapping table based on the terms of the service-level agreement (SLA) signed between the user and the operator, as well as the specific service type currently being used by the user. This mapping table assigns a corresponding priority identifier or value to each user. For example, an enterprise user with a mission-critical service agreement might have its service priority set to the highest level to ensure it receives priority in network resource competition. This priority information serves as a key basis for subsequent dynamic resource allocation algorithms, enabling differentiated services and ensuring service quality.

[0054] Optionally, the predicting spectrum availability and inter-band interference using a neural network model includes:

[0055] Build a neural network model including MLP network and LSTM network;

[0056] Specifically, the neural network model adopts a hybrid architecture that combines the advantages of processing static and time-series data. The MLP network is responsible for processing non-serialized, real-time information features, such as user priorities and the fixed geographic location of base stations. The LSTM network focuses on capturing time dependencies and dynamic patterns in the data, such as changes in channel status or traffic volume over time. The two networks can process their respective inputs in parallel or interact with each other through specific layers, forming a powerful prediction engine. The model's number of layers, nodes, activation functions, and connection methods are optimized based on the training data and prediction task requirements to ensure effective learning of complex patterns.

[0057] For example, a hybrid neural network model consisting of four MLP layers (each layer contains a different number of neurons) and two LSTM layers is designed. The MLP layer is used to process the base station deployment type and user static level, and the LSTM layer is used to process continuous signal-to-noise ratio and service traffic sequences.

[0058] The MLP network in the neural network model extracts and fuses the static features and nonlinear relationships of real-time information to generate static feature data;

[0059] Specifically, the MLP network receives static features extracted from real-time information, such as the user's service level agreement category, the base station deployment environment type (such as urban or suburban), and static priority tags preset by the network management system. The MLP processes these static features layer by layer and performs deep nonlinear mapping through multiple fully connected layers and nonlinear activation functions to discover complex relationships and patterns hidden in the data. This process converts the original static input into high-dimensional static feature data, which is used as supplementary information for subsequent prediction tasks. The computation process of an MLP layer can be expressed as:

[0060] ;

[0061] in, is the output vector of the lth layer of the MLP network, It is The input vector of the layer, It is The weight matrix of the layer, It is The bias vector of the layer, It is a commonly used rectified linear unit activation function.

[0062] Exemplarily, the MLP network learns the nonlinear correlation between the intensity of spectrum resource demand of high-priority users and their contract types from user data of different service level agreements, and encodes it into a set of discriminative static feature data.

[0063] The LSTM network in the neural network model captures the time series characteristics of real-time information and generates dynamic change patterns;

[0064] Specifically, the LSTM network is specifically designed to process real-time information sequences with time dependencies, such as continuous instantaneous signal-to-noise ratio measurements, dynamically changing sequences of user traffic, and periodic fluctuations in multipath propagation delay spread. LSTM effectively learns and memorizes long-term dependencies through its unique input gate, forget gate, and output gate mechanisms, avoiding the gradient vanishing or gradient exploding problems that may occur in traditional recurrent neural networks. It can identify dynamic patterns, trends, mutation points, and periodic changes in these time series data, thereby generating regular data that reflects the dynamic changes in the spectrum environment or user behavior. Cell state of LSTM unit and hidden state The update formula is as follows

[0065] ;

[0066] ;

[0067] ;

[0068] ;

[0069] ;

[0070] ;

[0071] in, is the input vector at the current moment, The hidden state vector at the previous moment, is the cell state vector at the previous moment, and are the weight matrix and bias vector, yes activation function, is the hyperbolic tangent activation function, Represents the Hadamard product (element-wise multiplication). , , They are forget gate, input gate and output gate respectively.

[0072] For example, by analyzing the business traffic change curve of a hotspot area within a day, the LSTM network identifies a regular pattern in which the business traffic in the area surges significantly and the signal-to-noise ratio decreases during the morning and evening peak hours on weekdays.

[0073] Based on the static feature data and dynamic change rules, a prediction result is generated, wherein the prediction result is spectrum availability prediction data, inter-band interference prediction data and prediction confidence information, and the prediction confidence information is a quantitative assessment of the reliability or uncertainty of the spectrum availability prediction data and the inter-band interference prediction data.

[0074] Specifically, the static feature data obtained from the MLP network and the dynamic change rules captured from the LSTM network are combined and deeply integrated through an advanced fusion layer. The fusion layer can adopt an attention mechanism to enable the model to dynamically focus on the importance of different time steps or different feature dimensions to the final prediction results; or adopt technologies such as gated recurrent units to further optimize the processing and fusion of time series information. The fused information is sent to the output layer of the neural network to generate spectrum availability prediction data for each frequency band or resource block in the future, as well as inter-band interference prediction data that may be generated under a specific reuse scheme. At the same time, in order to quantify the reliability of these predictions, the model also outputs prediction confidence information. Prediction confidence information is usually expressed as a probability range, variance or score of the predicted value, which reflects the degree of certainty of the model in its own prediction results. For example, for a predicted value , its confidence interval [ , ] can be obtained by Bayesian neural network or Monte Carlo Dropout method, satisfying

[0075] ;

[0076] in, is the true value, When training a model, its performance can be optimized by minimizing a composite loss function that includes prediction error and confidence evaluation.

[0077] like Figure 2As shown, the predicted availability distribution of different resource blocks in different time slots is displayed, and the dynamic distribution and usage of resources are reflected in the form of area perception.

[0078] For example, the neural network model predicts that a certain area of a sports venue will reach a high congestion state in a specific frequency band within the next 10 minutes during a large event, with a prediction confidence of 90%, indicating that the prediction result has high reliability.

[0079] Optionally, the generating a dynamic spectrum multiplexing strategy comprises:

[0080] Based on the spectrum availability prediction data and the inter-band interference prediction data, combined with the real-time time slot channel occupancy state, a spectrum multiplexing decision space is constructed, and the multiplexing interference risk of each idle resource block or low-interference resource block is quantified.

[0081] Specifically, combined with the spectrum availability prediction data and the inter-band interference prediction data provided by the spectrum prediction module, and by introducing the current time slot channel occupancy state obtained from the 5G base station in real time, a multi-dimensional spectrum multiplexing decision space is constructed. This decision space includes all idle resource blocks available for multiplexing or resource blocks currently in a low-interference state, and considers their potential performance under different multiplexing configurations. For each candidate resource block multiplexing scheme, the multiplexing interference risk it may cause is quantified. This is usually done by evaluating the interference effect of the multiplexed signal on the desired signal, such as calculating the impact of the interference on the expected signal quality degradation, to assess the risk level.

[0082] For example, in an indoor scenario, the system identifies a number of underutilized 5GHz frequency band resource blocks based on the prediction data. For inter-band multiplexing of these resource blocks, the system evaluates the expected interference that each multiplexing scheme may cause to neighboring active users, and marks the schemes with higher interference as high-risk.

[0083] According to the preset frequency resource priority mapping table, user service demand information, and prediction confidence information, an adaptive intelligent optimization algorithm is used to determine the optimal spectrum resource block allocation scheme within the spectrum multiplexing decision space.

[0084] Specifically, the frequency resource priority mapping table obtained from the network management system, the service demand information reported by the user terminal, and the prediction confidence information provided by the spectrum prediction module are used as inputs of the adaptive intelligent optimization algorithm. The algorithm will perform complex multi-objective optimization in the constructed spectrum multiplexing decision space. The optimization targets include but are not limited to minimizing the overall network interference, maximizing the system total throughput, and ensuring the quality of service and real-time performance of low-latency services of high-priority users. The algorithm has adaptive capability and can dynamically adjust the weights of different optimization targets or uncertain prediction results according to the prediction confidence, so as to reduce the decision risk while pursuing optimal performance.

[0085] For example, for a burst traffic event, the optimization algorithm prioritizes the allocation of resource blocks with high availability and low interference risk to emergency services considering the high priority of emergency communication users and the low prediction confidence of the related area, and dynamically adjusts the decision weight to balance the system throughput and latency requirements.

[0086] Based on the optimal spectrum resource block allocation scheme, the beamforming parameters of the 5G base station and the resource block scheduling scheme are adjusted to generate a dynamic spectrum multiplexing strategy.

[0087] Specifically, based on the optimal spectrum resource block allocation scheme determined by the adaptive intelligent optimization algorithm, the physical layer transmission parameters of the 5G base station are adjusted to implement the scheme. This includes precisely adjusting the beamforming parameters of the antenna, such as adjusting the phase and amplitude of each unit of the array antenna to form a more directional beam, thereby enhancing the target user signal while effectively suppressing the interference to the adjacent users. At the same time, the scheduling scheme of the resource block is carefully adjusted, such as determining which users use which resource blocks in which time slot and frequency band, and the resource allocation ratio of uplink and downlink transmission, to optimize the utilization of spectrum time-frequency two-dimensional resources. The collaborative configuration of these beamforming parameters and resource block scheduling schemes constitutes the final dynamic spectrum multiplexing strategy, aiming to maximize the spectrum efficiency and ensure network performance.

[0088] Optionally, the dynamic spectrum multiplexing strategy is sent to the corresponding 5G base station, including:

[0089] The network management system sends the dynamic spectrum multiplexing strategy to the corresponding 5G base station in real time through control signaling according to the real-time load state of the base station and the user geographic distribution information;

[0090] Specifically, the network management system continuously monitors the current load of each 5G base station under its jurisdiction, such as the number of active users, resource block utilization, or C-RAN (Centralized-Radio Access Network) central processing load. It also obtains geographic distribution information of user terminals, such as through GPS positioning or triangulation data, to understand the density of users within the coverage areas of different base stations. Based on this real-time load and geographic distribution information, the network management system distributes the generated dynamic spectrum reuse strategy in real time to the target 5G base stations that are most in need of adjustment or are related to the strategy through control signaling defined by the 5G network. This distribution mechanism ensures that the strategy can be delivered accurately and promptly to the base stations within the affected area, thereby guiding their spectrum usage behavior.

[0091] For example, during a major event, the network management system detects a surge in traffic load on base stations near a stadium, along with a high concentration of users. In this case, the system prioritizes a dynamic spectrum reuse strategy, including a high-density reuse solution, customized for that area and delivers it in real time to multiple 5G base stations around the stadium via the NG-RAN-AMF interface.

[0092] The corresponding 5G base station intelligently reconstructs the resource block mapping and allocation strategy of the uplink and downlink according to the received dynamic spectrum reuse strategy;

[0093] Specifically, upon receiving the dynamic spectrum reuse strategy from the network management system, the corresponding 5G base station does not simply passively execute it, but instead activates its internal intelligent reconstruction module. This module performs localized intelligent analysis based on the received macro-spectrum reuse strategy and combines its own current local real-time information, such as the instantaneous channel quality of each antenna port, the service queue status of currently connected users, and real-time interference measurement data from neighboring cells. Through this analysis, the base station dynamically adjusts and refines the uplink and downlink resource block mapping and allocation instructions to adapt to its current local wireless environment and user needs, forming a more optimized resource block mapping and allocation strategy that adapts to local conditions.

[0094] For example, after receiving a policy instruction, a base station discovers that a user in its service area is performing a high-priority, low-latency service on a good channel with minimal interference from neighboring cells. The base station intelligently adjusts resource block allocation, prioritizing contiguous resource blocks with minimal interference for the low-latency service, and may also fine-tune the mapping method to ensure latency and reliability.

[0095] The corresponding 5G base station adaptively adjusts spectrum usage parameters based on the reconstructed resource block mapping and allocation strategy combined with the dynamic spectrum reuse strategy.

[0096] Specifically, the corresponding 5G base station will map the refined resource block mapping and allocation strategy obtained by intelligent reconstruction as the core guidance, and combine other parameters in the dynamic spectrum multiplexing strategy received from the network management system to adaptively adjust its own spectrum usage parameters. This includes adjusting the beamforming weights of the antenna to accurately focus the energy to the target user direction, while forming a null to suppress the radiation to the interfering users; adjusting the transmit power to minimize the interference to other users while meeting the user link budget; adjusting the modulation and coding scheme to select the best modulation and coding level according to the actual channel quality and service demand to maximize the spectrum efficiency; and adjusting the subcarrier spacing to adapt to the requirements of different service types on latency and robustness. This adaptive adjustment ensures the flexible and efficient execution of the strategy at the base station end, maximizing the spectrum utilization efficiency.

[0097] For example, after receiving the reconstructed resource block mapping strategy, the base station will adaptively adjust the modulation and coding scheme of some resource blocks from 16QAM to 64QAM to support high-rate download services, while adjusting the beamforming weights to ensure that users can also obtain good signal quality under higher-order modulation.

[0098] Optionally, the continuous monitoring of the actual effect of the adjusted spectrum and the real-time updating of the neural network model comprise:

[0099] Continuous collection and analysis of user experience quality indicators, spectrum utilization rate and actual interference level of each 5G base station after adjustment to generate performance feedback data;

[0100] Specifically, after the dynamic spectrum multiplexing strategy is issued and the parameters are adjusted, the system will continuously collect multi-dimensional performance data from each 5G base station. These data include user experience quality indicators such as average throughput, download rate, upload rate, latency, jitter and packet loss rate, which directly reflect the quality of service experienced by users. At the same time, it also includes the actual spectrum utilization rate of each base station, i.e. the ratio of the number of resource blocks allocated to users to the total number of available resource blocks, and the real-time interference level obtained by actual measurement or inference, such as inter-cell interference, co-frequency interference, etc. These raw data are analyzed and summarized in real time to form structured performance feedback data for subsequent model updating and strategy optimization. The average throughput of a single user in a certain time period may be represented as:

[0101] ;

[0102] where is the number of resource blocks allocated to the user , and is the user​ In the resource block At the moment The instantaneous capacity of the base station.

[0103] ;

[0104] in It is a base station The actual number of resource blocks used, is the total number of resource blocks available to the base station, is the number of base stations.

[0105] For example, after a 5G base station adjusted its strategy, the system monitored in real time that the average download rate of users in its coverage area increased by 20%, and the actual interference level in the same frequency band decreased by 1.5dB. These data were recorded and used as performance feedback.

[0106] Based on the deviation analysis between the performance feedback data and the prediction results, an online learning mechanism is used to update the neural network model in real time and fine-tune the parameters;

[0107] Specifically, the collected performance feedback data will be compared in real time with the prediction results previously generated by the neural network model for deviation analysis. This deviation analysis can quantify the gap between the prediction and the actual situation. Based on this deviation, the system uses an online learning mechanism to update the neural network model in real time. Online learning means that the model can gradually adjust its internal parameters and weights in an incremental manner based on new feedback data without retraining the entire model. This method improves the adaptability of the model, enabling it to quickly adapt to changing network environments and unknown interference patterns. Parameter fine-tuning can include adjusting the weights, biases, or learning rates of the neural network layers. The mean square error (MSE) that measures the prediction error can be expressed as:

[0108] ;

[0109] in, is the actual observed value of the i-th sample, is the model prediction value of the i-th sample, is the total number of samples. In online learning, the model parameters The update can be performed using stochastic gradient descent, such as:

[0110] ;

[0111] in, is the model parameter at the current moment, is the learning rate, is the gradient of the loss function with respect to the parameters, is the input and target output at the current moment.

[0112] like Figure 3 As shown in the figure, during the operation of the system, as the neural network model is continuously updated through online learning, the deviation between the actual network performance and the model prediction performance gradually decreases and tends to be stable, thereby verifying the effectiveness of the closed-loop optimization mechanism of the present invention.

[0113] For example, a neural network model predicted that the interference level in a certain area would drop by 2dB after a policy adjustment, but actual monitoring showed a decrease of only 0.5dB. Based on this deviation, the model used an online learning algorithm to fine-tune the parameters related to the predicted interference to improve the accuracy of future predictions.

[0114] Based on the updated neural network model, the dynamic spectrum reuse strategy is adaptively re-evaluated and iteratively optimized to continuously improve spectrum efficiency and user service quality.

[0115] Specifically, the prediction ability of the neural network model after being updated by the online learning mechanism has been improved, and it can more accurately reflect the current and future network status. Based on this updated model, the policy generation module will adaptively re-evaluate its current dynamic spectrum reuse strategy. The re-evaluation process will re-examine the rationality and effectiveness of the existing strategy and compare it with the updated prediction results. If there is room for improvement, the system will start an iterative optimization process to generate a new and better spectrum reuse strategy. This iterative process can be a continuous feedback loop to ensure that the network always operates in an optimal or near-optimal spectrum utilization state, thereby continuously improving the overall spectrum efficiency and ensuring user service quality. Policy optimization can usually be modeled as a Markov decision process, in which the state ,action ,award and state transition probability Define the problem. Strategy It is a mapping from state to action. The optimization goal is to maximize the expected cumulative reward:

[0116] ;

[0117] in, It's a strategy performance, It is in strategy Lower expected value, is the discount factor, is the reward at time t. Through methods such as reinforcement learning, the model can learn the optimal strategy.

[0118] Exemplarily, after detecting that the user mobility pattern of a certain area has changed for a long time, the updated neural network model predicts that the spectrum availability pattern of the area has also changed. According to this new prediction, the strategy module re-evaluates and iteratively generates a new set of dynamic spectrum multiplexing strategies that are more suitable for the new mobility pattern, thereby avoiding future resource waste between frequency bands.

[0119] Optionally, the determining the optimal spectrum resource block allocation scheme in the spectrum multiplexing decision space comprises:

[0120] The adaptive intelligent optimization algorithm takes the minimization of overall network interference, the maximization of system throughput, and the guarantee of high-priority user service quality as the objective optimization function to search in the spectrum multiplexing decision space.

[0121] Specifically, the adaptive intelligent optimization algorithm performs a search process in the pre-constructed spectrum multiplexing decision space. The core of the algorithm is its multi-objective optimization function, which comprehensively considers multiple mutually influencing objectives, including: minimizing the total interference generated between different cells and different users in the entire 5G network due to spectrum multiplexing; maximizing the total data throughput of the network under a given spectrum resource; and ensuring that high-priority users can obtain the service quality guaranteed by their subscription service level agreement. The algorithm searches for various spectrum resource block allocation schemes in the decision space through iteration, evaluates the performance of each scheme on these objectives, and attempts to find a Pareto optimal solution set or a weighted optimal solution. A multi-objective optimization problem can be formalized as:

[0122] ;

[0123] wherein, is the decision variable to be optimized (spectrum resource block allocation scheme), is the th optimization objective. In order to convert multiple objectives into a single optimizable value, the weighted summation method is commonly used:

[0124] ;

[0125] wherein, is the overall network interference, is the negative value of the system total throughput, is the penalty term for high-priority user service quality guarantee. is the weight of each objective, and . The optimization algorithm searches the solution space of to find the optimal allocation scheme that satisfies the constraint conditions.

[0126] As Figure 4The figure shows the trade-off between total system throughput and total network interference level during the system optimization process. The black lines connecting the scattered points represent Pareto optimal solutions, indicating that one objective can be further improved without sacrificing the other, while the gray scattered points represent suboptimal allocation solutions.

[0127] For example, for network deployment during a large-scale sports event, the optimization algorithm seeks the best in the spectrum reuse decision space, with the goal of ensuring the high bandwidth requirements of live broadcasts while avoiding mutual interference caused by fans' mobile phone access, while ensuring extremely high priority services for event organizers' communications.

[0128] The adaptive intelligent optimization algorithm adjusts the optimization weights in real time according to real-time network load changes, user mobility and channel environment dynamics to generate the optimal spectrum resource block allocation solution.

[0129] Specifically, the adaptive intelligent optimization algorithm has the ability to dynamically adjust the weights of its optimization objectives to respond to real-time changes in the 5G network environment. For example, when a sharp increase in network load is detected in a certain area, the algorithm will increase the weight of the goal of maximizing system throughput in real time; when large-scale movement of user groups is identified, the algorithm will adjust the weights related to user mobility; and when the channel environment changes dynamically due to weather or obstacles, the algorithm will also adjust the weight of the goal of minimizing interference accordingly. This real-time adjustment ensures that the algorithm can make trade-offs based on the latest network conditions and forecast information, thereby generating the optimal spectrum resource block allocation plan that best adapts to the current complex and dynamic environment. Weight adjustments can be based on real-time feedback from network performance indicators and preset policy rules.

[0130] For example, at a transportation hub, during the morning rush hour, when a surge in users and traffic is detected, the optimization algorithm immediately increases the weighting for maximizing system throughput, prioritizing more resources to meet high concurrency demands. Meanwhile, during the low-load period at night, the algorithm might increase the implicit weighting for minimizing energy consumption or ensuring the utmost reliability for a small number of high-priority users.

[0131] Optionally, generating a prediction result based on the static feature data and the dynamic change rule includes:

[0132] The neural network model deeply fuses static feature data and dynamic change rules through an attention mechanism or a gated recurrent unit to generate a deep fusion result;

[0133] Specifically, after receiving static feature data from the MLP network and dynamic changes from the LSTM network, the neural network model deeply integrates them through a dedicated fusion layer. This fusion layer can use an attention mechanism, which allows the model to dynamically weight the importance of different features, thereby highlighting the static and dynamic information that has the greatest impact on the prediction results; or it can use efficient recurrent neural network structures such as gated recurrent units to more effectively process and fuse long-term dependencies and static context in time series information. Through this deep fusion, the model can capture more complex and deeper connections between static background information and time series dynamics, leveraging the complementary strengths of the two different types of data to produce a unified and information-rich deep fusion result.

[0134] For example, when predicting spectrum availability in a certain area, the attention mechanism enables the model to pay special attention to the impact of current user priorities and base station load on future availability while processing historical channel data, thereby generating more accurate fusion results.

[0135] generating spectrum availability prediction data and inter-band interference prediction data according to the deep fusion result;

[0136] Specifically, the result after processing by the deep fusion layer includes a comprehensive understanding of the real-time network status and the encoding of complex patterns. This deep fusion result is input into the output layer of the neural network model, which is designed to generate predictions for the specific goals of the spectrum reuse task. Specifically, it will output future spectrum availability prediction data, indicating whether each specific frequency band or resource block will be idle, busy, or available in the future. At the same time, it will also output inter-band interference prediction data, quantifying the intensity of mutual interference or the impact on service quality that may occur when these spectrum resources are reused by different users or cells. These prediction data are given in numerical form to provide basic input for subsequent strategy generation.

[0137] For example, the fusion results indicate that within the next 50 milliseconds, there is an 80% probability that two resource blocks in a specific high frequency band will be idle, and if they are allocated to users in two adjacent cells, the model predicts that the mutual interference level will be around -80dBm.

[0138] A prediction confidence evaluation is performed on the spectrum availability prediction data and the inter-band interference prediction data to generate prediction confidence information.

[0139] Specifically, after generating spectrum availability prediction data and inter-band interference prediction data, the neural network model also performs a prediction confidence assessment on these prediction results. The prediction confidence assessment aims to quantify the reliability or uncertainty of each prediction result. This can be achieved through a variety of methods. For example, the model can output a probability distribution associated with the predicted value, or provide a confidence score between 0 and 1, with a higher score indicating a more reliable prediction. This assessment takes into account the noise level of the input data, the historical performance of the model in similar situations, and the volatility of the predicted value itself, and ultimately generates prediction confidence information. This information is crucial for subsequent decision-making because it allows the system to adopt more conservative or more aggressive strategies when faced with uncertainty.

[0140] For example, for the prediction of high spectrum availability during a certain period of time, the model may also give a 95% confidence level. However, for the prediction of strong interference at a certain point in the future, if the model determines that the input data is incomplete or fluctuates greatly, it may give a 70% confidence level, indicating to the decision-maker that the prediction is highly uncertain.

[0141] Optionally, the intelligent reconstruction of uplink and downlink resource block mapping and allocation strategies includes:

[0142] The corresponding 5G base station performs local optimality judgment based on the received dynamic spectrum reuse strategy, combined with real-time channel quality, local user service queue status and neighboring cell interference information;

[0143] Specifically, when a 5G base station receives a dynamic spectrum reuse strategy from the network management system, it immediately analyzes it in conjunction with its current local real-time data to determine local optimality. This local real-time data includes, but is not limited to: real-time channel quality measured by the base station's own antenna port or receiver module, such as instantaneous signal-to-noise ratio and reference signal received power; the service queue status of each connected user, such as the amount of uplink data waiting to be sent, the length of the buffer for data waiting to be sent, and the average queuing delay of data packets; and real-time neighboring cell interference information obtained from reports from neighboring cell base stations or user terminals, such as co-channel interference intensity and neighboring cell signal strength. The base station will comprehensively evaluate this local information and the received policy instructions to determine the best implementation method for the current strategy in the local environment.

[0144] For example, a 5G base station receives a policy instruction to encourage high reuse rate, but at the same time its local real-time monitoring data shows that the channel quality in a specific direction suddenly deteriorates, and there is a neighboring cell that is performing high-power transmission nearby. The base station will immediately determine that performing high reuse in this specific direction may cause local performance degradation.

[0145] Based on the local optimality judgment, the corresponding 5G base station dynamically adjusts the resource block mapping and allocation strategy to generate a resource block mapping and allocation strategy that adapts to the current network environment.

[0146] Specifically, after completing the local optimality judgment, the 5G base station will dynamically adjust and refine the resource block mapping and allocation instructions in the received dynamic spectrum reuse strategy based on the judgment result. This adjustment is not a simple execution of the original strategy, but an optimization after considering factors such as local channel conditions, user needs, and interference conditions. For example, if a local judgment finds that a resource block is recommended for reuse by the strategy but the current interference is too high, the base station may choose to allocate it to a service that is not sensitive to interference, or adjust its power and beamforming mode to avoid interference, or even temporarily avoid high-priority service transmission on this resource block, in order to generate a resource block mapping and allocation strategy that is more suitable for the current local network environment, ensuring that resource utilization efficiency and user experience are optimal locally.

[0147] For example, based on its local judgment, the base station remaps the resource blocks allocated to the video streaming service in a certain area in the original strategy to another set of resource blocks with better current channel quality to avoid sudden interference detected locally, and at the same time selects a low-priority IoT service that is insensitive to interference for allocation to the original resource block, thereby achieving optimal resource utilization.

[0148] Based on the same inventive concept, Figure 5 As shown, the present invention also provides a multiplexing system for improving 5G mobile communication spectrum efficiency, the system comprising:

[0149] Real-time information collection module, used to collect real-time information of multiple 5G base stations;

[0150] a spectrum prediction module, configured to predict spectrum availability and inter-band interference using a neural network model based on the real-time information and generate a prediction result;

[0151] A dynamic strategy generation module is used to generate a dynamic spectrum reuse strategy based on the prediction result and a preset frequency resource priority mapping table; wherein the dynamic spectrum reuse strategy includes a joint scheduling scheme of time division multiplexing and frequency division multiplexing;

[0152] A policy deployment and parameter adjustment module is used to send the dynamic spectrum reuse policy to the corresponding 5G base station and adjust the spectrum usage parameters;

[0153] The performance monitoring and optimization module is used to continuously monitor the actual usage effect of the adjusted spectrum and update the neural network model in real time to optimize the spectrum reuse strategy.

[0154] It should be noted that the electrical connections between the above-mentioned units do not necessarily mean direct connections of lines. Indirect connections are applicable to the embodiments of the present invention as long as the purpose of the present invention is achieved. The above description is only an exemplary embodiment of the present invention and is not intended to limit the scope of the present invention.

[0155] That is, any equivalent changes and modifications made according to the teachings of the present invention are still within the scope of the present invention. Those skilled in the art will readily conceive of other embodiments of the present invention after considering the disclosure of the specification and practical truths. This application is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary technical means in the art not described herein.

Claims

1. A multiplexing method for improving 5G mobile communication spectrum efficiency, characterized in that: The method comprises: Collecting real-time information of multiple 5G base stations, wherein the real-time information includes time slot channel occupancy status, user service demand information, and a frequency resource priority mapping table preset by the network management system; Based on the real-time information, a neural network model is used to predict spectrum availability and inter-band interference to generate prediction results, including: constructing a neural network model including a gradient grower (MLP) network and a long short-term memory (LSTM) network; the MLP network in the neural network model extracts and fuses static features and nonlinear relationships of the real-time information to generate static feature data; the LSTM network in the neural network model captures time series features of the real-time information to generate dynamic change patterns; and generating prediction results by deeply fusing the static feature data and the dynamic change patterns, the prediction results comprising spectrum availability prediction data, inter-band interference prediction data, and prediction confidence information, wherein the prediction confidence information is a quantitative assessment of the reliability or uncertainty of the spectrum availability prediction data and the inter-band interference prediction data; A dynamic spectrum reuse strategy is generated based on the prediction results and the preset frequency resource priority mapping table; wherein the dynamic spectrum reuse strategy includes a joint scheduling scheme of time division multiplexing and frequency division multiplexing; including: based on the spectrum availability prediction data and the inter-band interference prediction data, combined with the real-time time slot channel occupancy status, a spectrum reuse decision space is constructed, and the reuse interference risk of each idle resource block or low-interference resource block is quantified; based on the preset frequency resource priority mapping table, user service demand information and prediction confidence information, an adaptive intelligent optimization algorithm is used to determine the optimal spectrum resource block allocation scheme within the spectrum reuse decision space; based on the optimal spectrum resource block allocation scheme, the beamforming parameter power allocation and resource block scheduling scheme of the 5G base station are coordinated to generate a dynamic spectrum reuse strategy; Send the dynamic spectrum reuse strategy to the corresponding 5G base station and adjust the spectrum usage parameters; Continuously monitor the actual use of the adjusted spectrum and update the neural network model in real time to optimize the spectrum reuse strategy; Among them, the determination of the optimal spectrum resource block allocation scheme within the spectrum reuse decision space includes: the adaptive intelligent optimization algorithm uses minimizing overall network interference, maximizing system throughput and ensuring high-priority user service quality as the target optimization function to search for the best within the spectrum reuse decision space; the adaptive intelligent optimization algorithm adjusts the optimization weights in real time according to real-time network load changes, user mobility and channel environment dynamics to generate the optimal spectrum resource block allocation scheme.

2. A multiplexing method for improving 5G mobile communication spectrum efficiency according to claim 1, characterized in that: The collecting of real-time information of multiple 5G base stations includes: Collect the instantaneous signal-to-noise ratio, historical data of time slot occupancy, and cyclic prefix interference level of each 5G base station; Obtain uplink and downlink service traffic, delay tolerance, and multi-band parallel transmission capabilities reported by user terminals; The network management system receives a frequency resource priority mapping table determined according to the service level agreement signed by the user and the service type.

3. A multiplexing method for improving 5G mobile communication spectrum efficiency according to claim 1, characterized in that: Sending the dynamic spectrum reuse strategy to the corresponding 5G base station includes: The network management system sends the dynamic spectrum reuse strategy to the corresponding 5G base station in real time through control signaling based on the real-time load status of the base station and the geographical distribution information of users; The corresponding 5G base station intelligently reconstructs the resource block mapping and allocation strategy of the uplink and downlink according to the received dynamic spectrum reuse strategy; The corresponding 5G base station adaptively adjusts spectrum usage parameters based on the reconstructed resource block mapping and allocation strategy combined with the dynamic spectrum reuse strategy.

4. A multiplexing method for improving 5G mobile communication spectrum efficiency according to claim 1, characterized in that: The continuously monitoring the actual use effect of the adjusted spectrum and updating the neural network model in real time includes: Continuously collect and analyze user experience quality indicators, spectrum utilization, and actual interference levels of each adjusted 5G base station to generate performance feedback data; Based on the deviation analysis between the performance feedback data and the prediction results, an online learning mechanism is used to update the neural network model in real time and fine-tune the parameters; Based on the updated neural network model, the dynamic spectrum reuse strategy is adaptively re-evaluated and iteratively optimized to continuously improve spectrum efficiency and user service quality.

5. The multiplexing method for improving 5G mobile communication spectrum efficiency according to claim 1, characterized in that: Generating prediction results based on the static feature data and dynamic change rules includes: The neural network model deeply fuses static feature data and dynamic change rules through an attention mechanism or a gated recurrent unit to generate a deep fusion result; generating spectrum availability prediction data and inter-band interference prediction data according to the deep fusion result; A prediction confidence evaluation is performed on the spectrum availability prediction data and the inter-band interference prediction data to generate prediction confidence information.

6. A multiplexing method for improving 5G mobile communication spectrum efficiency according to claim 3, characterized in that: The intelligent reconstruction of uplink and downlink resource block mapping and allocation strategy includes: The corresponding 5G base station performs local optimality judgment based on the received dynamic spectrum reuse strategy, combined with real-time channel quality, local user service queue status, and neighboring cell interference information; Based on the local optimality judgment, the corresponding 5G base station dynamically adjusts the resource block mapping and allocation strategy to generate a resource block mapping and allocation strategy that adapts to the current network environment.

7. A multiplexing system for improving 5G mobile communication spectrum efficiency is applied to a multiplexing method for improving 5G mobile communication spectrum efficiency as described in any one of claims 1 to 6, characterized in that: The system comprises: A real-time information collection module collects real-time information of multiple 5G base stations, wherein the real-time information includes time slot channel occupancy status, user service demand information, and a frequency resource priority mapping table preset by the network management system; A spectrum prediction module is configured to predict spectrum availability and inter-band interference based on the real-time information using a neural network model to generate prediction results, including: constructing a neural network model comprising a gradient grower (MLP) network and a long short-term memory (LSTM) network; the MLP network in the neural network model extracting and fusing static features and nonlinear relationships of the real-time information to generate static feature data; the LSTM network in the neural network model capturing time series features of the real-time information to generate dynamic change patterns; and generating prediction results by deeply fusing the static feature data and dynamic change patterns, the prediction results comprising spectrum availability prediction data, inter-band interference prediction data, and prediction confidence information, wherein the prediction confidence information is a quantitative assessment of the reliability or uncertainty of the spectrum availability prediction data and the inter-band interference prediction data; A dynamic strategy generation module is used to generate a dynamic spectrum reuse strategy based on the prediction results and a preset frequency resource priority mapping table; wherein the dynamic spectrum reuse strategy includes a joint scheduling scheme of time division multiplexing and frequency division multiplexing; including: based on the spectrum availability prediction data and the inter-band interference prediction data, combined with the real-time time slot channel occupancy status, constructing a spectrum reuse decision space, and quantifying the reuse interference risk of each idle resource block or low-interference resource block; according to the preset frequency resource priority mapping table, user service demand information and prediction confidence information, using an adaptive intelligent optimization algorithm to determine the optimal spectrum resource block allocation scheme within the spectrum reuse decision space; based on the optimal spectrum resource block allocation scheme, collaboratively adjusting the beamforming parameter power allocation and resource block scheduling scheme of the 5G base station to generate a dynamic spectrum reuse strategy; A policy deployment and parameter adjustment module is used to send the dynamic spectrum reuse policy to the corresponding 5G base station and adjust the spectrum usage parameters; The performance monitoring and optimization module is used to continuously monitor the actual use of the adjusted spectrum and update the neural network model in real time to optimize the spectrum reuse strategy; Among them, the determination of the optimal spectrum resource block allocation scheme within the spectrum reuse decision space includes: the adaptive intelligent optimization algorithm uses minimizing overall network interference, maximizing system throughput and ensuring high-priority user service quality as the target optimization function to search for the best within the spectrum reuse decision space; the adaptive intelligent optimization algorithm adjusts the optimization weights in real time according to real-time network load changes, user mobility and channel environment dynamics to generate the optimal spectrum resource block allocation scheme.

Images