5G base station energy efficiency optimization method and system based on dynamic spectrum sharing

Abstract

The invention discloses a 5G base station energy efficiency optimization method and system based on dynamic spectrum sharing. The method comprises the following steps: S1, sensing user load distribution, service types and channel state information in a base station coverage area in real time; s2, dynamically allocating a shared spectrum resource pool composed of an authorized frequency band and an unauthorized frequency band according to the user load distribution and the service type; and S3, generating a spectrum allocation proportion and transmitting power joint optimization instruction based on a reinforcement learning algorithm. By dynamically fusing authorized and unauthorized spectrum resources and combining real-time service load awareness and reinforcement learning decision mechanisms, joint improvement of spectrum efficiency and energy consumption efficiency is realized on the premise of ensuring service quality. The prediction model embedded in the method can pre-judge network flow fluctuation in advance, symbol-level fine-grained energy-saving control is matched, the conflict between communication performance and equipment power consumption is effectively balanced, and the method is particularly suitable for heterogeneous network scenes with uneven user distribution and complex service types.

Description

Technical Field

[0001] The present invention relates to the technical field of 5G base stations, and particularly to a method and system for optimizing the energy efficiency of 5G base stations based on dynamic spectrum sharing. Background Art

[0002] With the large-scale deployment of 5G networks, the energy consumption of base stations has exceeded 60% of the total energy consumption of the mobile communication network, and the continuous high power consumption of large-scale antenna arrays and wideband radio frequency units has become a prominent problem.

[0003] Existing energy-saving technologies mainly reduce the operating power consumption of equipment through local optimization means such as symbol shutdown and carrier sleep. However, in actual applications, static energy-saving strategies are difficult to match dynamic changes in service loads. For example, in scenarios such as stadiums and transportation hubs, fixed sleep period settings often result in insufficient resource supply during peak service periods; The lack of cooperative control between base stations makes it difficult to globally schedule spectrum resources, and the phenomenon of adjacent base stations repeatedly activating the same frequency band is widespread; In addition, traditional methods usually isolate the processing of spectrum efficiency and energy consumption indicators, lacking a joint optimization mechanism for the overall energy efficiency ratio of the network, which is prone to the contradiction of coexistence of local hot spot overload and overall resource idleness in ultra-dense networking environments.

[0004] In view of the above problems, we have developed a method and system for optimizing the energy efficiency of 5G base stations based on dynamic spectrum sharing. Summary of the Invention

[0005] The present invention discloses a method and system for optimizing the energy efficiency of 5G base stations based on dynamic spectrum sharing, aiming to solve the technical problems in the background art.

[0006] To achieve the above object, the present invention adopts the following technical solutions: A method for optimizing the energy efficiency of 5G base stations based on dynamic spectrum sharing, comprising the following steps: S1. Real-time sense the user load distribution, service type, and channel state information within the coverage area of the base station; S2. Dynamically allocate a shared spectrum resource pool composed of authorized frequency bands and unlicensed frequency bands according to the user load distribution and service type; S3. Generate a joint optimization instruction for spectrum allocation ratio and transmit power based on a reinforcement learning algorithm, where the optimization objective is to maximize the ratio of spectrum efficiency to energy consumption; S4. Execute the power consumption mode switching operation of the radio frequency unit and the baseband processor according to the optimization instruction.

[0007] In a preferred solution, the step S1 is refined into the following steps: S11. Monitor the interference level of adjacent base stations through spectrum sensors deployed at the edge of the base station; S12. Classify and statistically analyze the latency and reliability requirements of eMBB, URLLC, and mMTC services.

[0008] In a preferred solution, the dynamic allocation of spectrum resources in step S2 includes: S21. Allocate subcarriers between the 3.5 GHz licensed band and the 5 GHz unlicensed band according to service priorities; S22. When detecting congestion in the unlicensed band, forcefully switch URLLC services to the licensed band.

[0009] In a preferred solution, the reinforcement learning algorithm in step S3 uses a Deep Q-Network (DQN), where: The state space includes the real-time spectrum utilization rate, the SINR value of the user equipment, and the base station heat dissipation power consumption; The action space is defined as the combination of 10 discretized levels of the transmit power and the spectrum allocation ratio; The reward function is calculated as: The number of latency violations; where is the total power consumption of the base station, is the penalty coefficient.

[0010] In a preferred solution, step S4 includes symbol-level energy-saving operations: S41. Identify idle subcarriers without data transmission within the OFDM symbol period; S42. Turn off the power supply of the RF amplifier corresponding to the idle subcarriers while maintaining the transmission of pilot signals.

[0011] A 5G base station energy efficiency optimization system based on dynamic spectrum sharing includes: A load awareness module for collecting the number of users, service types, and channel quality parameters; A dynamic spectrum sharing module configured to manage the pooled allocation of resources in licensed and unlicensed bands; An energy efficiency decision engine that outputs spectrum and power adjustment strategies through a reinforcement learning model; An energy-saving execution unit that controls the power consumption status of the RF front-end and baseband hardware according to the strategy.

[0012] In a preferred solution, the energy-saving execution unit implements a multi-level sleep strategy: First-level sleep: Turn off the baseband digital signal processing unit and maintain the RF channel synchronization signal; Second-level sleep: Disconnect the power supply link of the power amplifier and let neighboring base stations take over edge users.

[0013] In a preferred solution, it further includes a base station cluster cooperation controller for: Centralize the control plane signaling to the central node through the C / U separation architecture; Migrate user data to the designated base station during the off-peak traffic period, and the remaining base stations enter the secondary sleep state.

[0014] In a preferred solution, the energy efficiency decision engine integrates an LSTM prediction model, and the model: Inputs the traffic volume, weather events, and base station location data in the past 24 hours; Outputs the predicted value of the spectrum demand within the next 15 minutes, with an error rate lower than 8%.

[0015] In a preferred solution, the radio frequency front end supports dynamic antenna silencing: Turn off the outer circular antenna array according to the user location information; Adaptive adjust the beamforming codebook to maintain the RSSI not lower than -110 dBm.

[0016] A 5G base station energy efficiency optimization method and system based on dynamic spectrum sharing provided by the present invention have the following advantages: 1. By dynamically integrating licensed and unlicensed spectrum resources, combining real-time traffic load awareness and reinforcement learning decision-making mechanism, this method realizes the joint improvement of spectrum utilization rate and energy consumption efficiency on the premise of ensuring service quality. The prediction model embedded in the method can predict network traffic fluctuations in advance, and cooperate with symbol-level fine-grained energy-saving control, effectively balancing the conflict between communication performance and device power consumption, especially suitable for heterogeneous network scenarios with uneven user distribution and complex service types.

[0017] 2. This system adopts a modular hardware architecture and a distributed cooperative control design. Through the closed-loop linkage of load awareness, resource scheduling, decision-making engine, and execution unit, the algorithm strategy is transformed into device-level operations that can be implemented. The system supports multi-level sleep modes and dynamic antenna adjustment functions, significantly reducing the operating energy consumption while maintaining the base station coverage ability. Its federated learning mechanism and edge cooperation scheme further improve the global optimality of cross-base station resource scheduling, and have the engineering feasibility for large-scale commercial deployment. Description of the Drawings

[0018] Figure 1 It is a schematic diagram of the overall process of a 5G base station energy efficiency optimization method based on dynamic spectrum sharing proposed by the present invention.

[0019] Figure 2 It is a schematic diagram of the detailed steps of step S1 of a 5G base station energy efficiency optimization method based on dynamic spectrum sharing proposed by the present invention.

[0020] Figure 3 Schematic diagram of the detailed steps of step S2 of a 5G base station energy efficiency optimization method based on dynamic spectrum sharing proposed by the present invention.

[0021] Figure 4 Schematic diagram of the detailed steps of step S4 of a 5G base station energy efficiency optimization method based on dynamic spectrum sharing proposed by the present invention.

[0022] Figure 5 Schematic diagram of the structure of a 5G base station energy efficiency optimization system based on dynamic spectrum sharing proposed by the present invention. Specific implementation manners

[0023] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Usually, the components of the embodiments of the present application described and marked in the accompanying drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application to be protected, but only represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts fall within the scope of protection of the present application.

[0024] A 5G base station energy efficiency optimization method and system based on dynamic spectrum sharing disclosed by the present invention.

[0025] Referring to Figure 1 、 Figure 2 、 Figure 3 and Figure 4 As shown, a 5G base station energy efficiency optimization method based on dynamic spectrum sharing includes the following steps: S1. Real-time sense the user load distribution, service type, and channel state information within the base station coverage area; The step S1 is refined into the following steps: S11. Monitor the interference level of adjacent base stations through spectrum sensors deployed at the edge of the base station; S12. Classify and statistically analyze the latency and reliability requirements of eMBB, URLLC, and mMTC services.

[0026] S2. Dynamically allocate a shared spectrum resource pool composed of authorized frequency bands and unlicensed frequency bands according to the user load distribution and service type; The dynamic allocation of spectrum resources in the step S2 includes: S21. Allocate subcarriers between the 3.5 GHz authorized frequency band and the 5 GHz unlicensed frequency band according to the service priority; S22. When unauthorized frequency band congestion is detected, force the handover of URLLC services to the authorized frequency band.

[0027] S3. Generate a joint optimization instruction for spectrum allocation ratio and transmit power based on a reinforcement learning algorithm, where the optimization objective is to maximize the ratio of spectral efficiency to energy consumption. The reinforcement learning algorithm in step S3 adopts a deep Q-network (DQN), where: The state space includes real-time spectrum utilization rate, user equipment SINR value, and base station heat dissipation power consumption. The action space is defined as the combination of 10 discretized levels of transmit power and spectrum allocation ratio. The reward function is calculated as: The number of times of delay default; where is the total power consumption of the base station, is the penalty coefficient.

[0028] S4. Perform the power consumption mode switching operation of the radio frequency unit and the baseband processor according to the optimization instruction. Step S4 includes symbol-level energy-saving operations: S41. Identify idle subcarriers without data transmission within the OFDM symbol period. S42. Turn off the power supply of the radio frequency amplifier corresponding to the idle subcarrier while maintaining the transmission of pilot signals.

[0029] In this embodiment, when deploying the 5G base station energy efficiency optimization method based on dynamic spectrum sharing, first, through the spectrum sensors and user terminal reporting modules built in the base station, the service types and channel states of user equipment in the coverage area are collected in real time. For example, in a high-density scenario in a stadium, the sensors monitor the occupancy rates of the 3.5 GHz authorized frequency band and the 5 GHz unauthorized frequency band at a cycle of 1 second, and at the same time, classify and count the proportions of the currently connected eMBB video streams, URLLC remote control instructions, and mMTC sensor data. When the interference level of the unauthorized frequency band (5.15 - 5.85 GHz) exceeds -85 dBm, the dynamic resource scheduler forces the handover of URLLC services to the authorized frequency band to ensure its delay requirement of less than 1 ms, while the mMTC services continue to be transmitted in the unauthorized frequency band.

[0030] To generate a joint optimization strategy for spectrum and power, the system uses a Deep Q-Network (DQN) model for training. During specific implementation, the state input includes the spectrum utilization rate at the current moment, the average user SINR value, and the temperature of the base station radiator. The action space is defined as a combination of discrete transmit power levels from 10 dBm to 46 dBm and the allocation ratio of licensed and unlicensed frequency bands (such as 70%:30%). In the calculation of the reward function, the spectral efficiency is achieved through the logarithmic summation formula ∑log2(1 + SINR_i). The total power consumption P_total includes the real-time energy consumption data of the baseband processor, radio frequency unit, and cooling system. The penalty term α is set to 0.5 to constrain the delay violation rate of URLLC services to be lower than 3%. After training is completed, the model updates the policy table every 5 minutes, and the online inference delay is controlled within 50 ms.

[0031] In symbol-level energy-saving operations, the baseband processor analyzes the OFDM symbol scheduling information to identify idle subcarriers without data transmission. For example, within a 1 ms transmission time interval (TTI), if a subcarrier group modulated by 64QAM has not been assigned user data for two consecutive symbol periods, a turn-off instruction for the power supply circuit of the radio frequency front end is triggered, and only the subcarriers corresponding to the pilot signals are powered. This operation can reduce the power consumption of the radio frequency unit by about 18% while maintaining the channel estimation accuracy through the continuous transmission of pilot signals.

[0032] Furthermore, to cope with traffic fluctuations, the system integrates an LSTM prediction model. The input data includes the base station traffic within the past 24 hours, weather events (such as signal attenuation caused by heavy rain), and the load status of surrounding base stations. The model outputs the predicted value of the spectrum demand within the next 15 minutes, and the error rate is verified to be lower than 8% through the root mean square error (RMSE). For example, when it is predicted that there will be an instantaneous traffic peak when a concert ends, the system activates a 100 MHz spectrum resource block in the dormant state 10 minutes in advance and preheats the power amplifier to the standby state to avoid a decline in QoS caused by sudden loads.

[0033] In the scenario of base station cooperation, multiple base stations exchange load status through an edge server. When a certain area enters the off-peak business period (such as 1:00 - 5:00 in the morning), the central controller migrates users to the preset main service base station, and the remaining base stations execute a secondary sleep strategy: first, turn off the baseband digital signal processing unit and maintain the radio frequency channel synchronization signal required for the control plane signaling; if there is no user access for 30 consecutive minutes, completely cut off the power supply of the power amplifier, and the adjacent base stations take over the edge users by expanding the coverage radius. The spectrum resources released during the sleep period are dynamically incorporated into the shared resource pool for active base stations to call as needed.

[0034] A 5G base station energy efficiency optimization system based on dynamic spectrum sharing, including: A load awareness module, configured to collect the number of users, service types, and channel quality parameters; A dynamic spectrum sharing module, configured to manage the pooled allocation of resources in licensed and unlicensed frequency bands; An energy efficiency decision engine, configured to output spectrum and power adjustment strategies through a reinforcement learning model; An energy saving execution unit, configured to control the power consumption status of the radio frequency front end and baseband hardware according to the strategies.

[0035] The energy saving execution unit implements a multi-level sleep strategy: Level 1 sleep: Turn off the baseband digital signal processing unit and maintain the radio frequency channel synchronization signal; Level 2 sleep: Disconnect the power supply link of the power amplifier, and let neighboring base stations take over edge users.

[0036] It further includes a base station cluster cooperation controller, configured to: Centralize control plane signaling to a central node through a C / U separation architecture; Migrate user data to a specified base station during low-traffic periods, and let the remaining base stations enter the Level 2 sleep.

[0037] The energy efficiency decision engine integrates an LSTM prediction model, which: Inputs traffic volume, weather events, and base station location data for the past 24 hours; Outputs a predicted value of spectrum demand within the next 15 minutes, with an error rate lower than 8%.

[0038] The radio frequency front end supports dynamic antenna silencing: Turn off the outer circular antenna array according to user location information; Adaptively adjust the beamforming codebook to maintain the RSSI not lower than -110 dBm.

[0039] When deploying a 5G base station energy efficiency optimization system based on dynamic spectrum sharing, base station equipment is equipped with a multi-band radio frequency front end and a programmable baseband chipset. The load awareness module, through a spectrum scanner integrated in the AAU (active antenna unit), scans the occupancy status of licensed spectrum in the 3.5 GHz band and unlicensed spectrum in the 5 GHz band every 200 ms, and at the same time receives CSI (channel state information) reports reported by user terminals. For example, when a base station in an industrial park detects co-channel interference from Wi-Fi 6 devices in the 5 GHz band, the dynamic spectrum sharing module starts a federated learning agent node, exchanges historical spectrum occupancy data with 3 neighboring stations within a radius of 500 meters, predicts the available time window for this band within the next 2 minutes, and preferentially schedules URLLC services to clean subcarriers with interference lower than -82 dBm.

[0040] The energy efficiency decision-making engine is deployed within the CU (Central Unit) of the base station and adopts a DQN inference model accelerated by TensorRT. The input layer of the model receives real-time data streams from the load perception module, including the traffic distribution of 32 currently connected eMBB users, 5 URLLC devices, and 120 mMTC sensors, as well as the power consumption data reported by the baseband processor (such as the instantaneous power consumption of the baseband chipset being 28W). The decision-making engine outputs action instructions to the energy-saving execution unit. For example, it reduces the transmit power of the AAU from 40 dBm to 34 dBm, adjusts the authorized frequency band allocation ratio from 80% to 65%, and simultaneously triggers the activation of a 20 MHz bandwidth in the unlicensed frequency band. The execution unit adjusts the bias voltage of the radio frequency channel through the PMIC (Power Management Integrated Circuit), increasing the efficiency of the power amplifier from 38% to 45%, and saving approximately 1.2 kW·h of energy per hour for a single base station.

[0041] In the symbol-level shutdown scenario, the hardware energy-saving execution unit obtains the OFDM symbol mapping table through the baseband scheduler. When it detects that 16 outer antenna elements in a certain 64T64R antenna array have not served any users for three consecutive time slots, the execution unit sends an antenna silence instruction to the AAU, turns off the power supply of the corresponding elements, and switches to the 8T8R operating mode. Meanwhile, the beamforming controller recalculates the precoding matrix based on the positions of the remaining antennas to ensure that the RSRP (Reference Signal Received Power) at the coverage edge remains above -105 dBm. For example, during the low-load period after the mall closes, this system can reduce the overall power consumption of the antenna array by 62% and avoid coverage holes by dynamically adjusting the beam width.

[0042] The base station cluster cooperation controller collects the load status of 6 base stations in the area in real time through the X2 interface proxy module deployed on the edge MEC server. When the number of users of a base station in an office area drops below 10 between 23:00 at night and 6:00 the next day, the cooperation controller initiates the user migration process: First, it switches the VoNR voice service to an adjacent base station, and then issues a first-level sleep instruction for the target base station to the energy-saving execution unit, turning off the FFT / IP core of the baseband processor while retaining the clock synchronization circuit. If no new users are connected for 1 hour, it triggers a second-level sleep, completely disconnecting the 48V DC power supply link of the AAU. At this time, the power consumption of the base station drops from the normal 850W to 75W, and the released 100 MHz spectrum resources are dynamically allocated by the cooperation controller to the base stations still in service.

[0043] As described above, it is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. The substitution can be the substitution of part of the structure, device, method steps, or a complete technical solution. Any equivalent substitution or change made according to the technical solution of the present invention and its inventive concept should be covered within the protection scope of the present invention.

Claims

1. A 5G base station energy efficiency optimization method based on dynamic spectrum sharing, characterized in that: The following steps are involved: S1, real-time perception of user load distribution, service type and channel status information within the base station coverage area; S2. Dynamically allocating a shared spectrum resource pool consisting of licensed frequency bands and unlicensed frequency bands according to the user load distribution and service type; S3, generating a spectrum allocation ratio and a transmission power joint optimization instruction based on a reinforcement learning algorithm, wherein the optimization goal is to maximize the ratio of spectrum efficiency to energy consumption; S4. Execute a power consumption mode switching operation of the radio frequency unit and the baseband processor according to the optimization instruction.

2. The 5G base station energy efficiency optimization method based on dynamic spectrum sharing according to claim 1 is characterized in that: The step S1 is refined into the following steps: S11. Monitor the interference level of adjacent base stations through spectrum sensors deployed at the edge of the base station; S12. Classify and count the latency and reliability requirements of eMBB, URLLC, and mMTC services.

3. The 5G base station energy efficiency optimization method based on dynamic spectrum sharing according to claim 1 is characterized in that: The dynamic allocation of spectrum resources in step S2 includes: S21. Allocate subcarriers between the 3.5 GHz licensed band and the 5 GHz unlicensed band according to service priority; S22: When congestion in the unlicensed frequency band is detected, the URLLC service is forcibly switched to the licensed frequency band.

4. The 5G base station energy efficiency optimization method based on dynamic spectrum sharing according to claim 1 is characterized in that: The reinforcement learning algorithm in step S3 adopts a deep Q network (DQN), where: The state space includes real-time spectrum utilization, user equipment SINR value, and base station heat dissipation power consumption; The action space is defined as a combination of 10 discretized levels of transmit power and spectrum allocation ratios; The reward function is calculated as: Number of time delay breaches; in is the total power consumption of the base station, is the penalty coefficient.

5. The 5G base station energy efficiency optimization method based on dynamic spectrum sharing according to claim 1 is characterized in that: The step S4 includes symbol-level energy-saving operations: S41, identifying idle subcarriers with no data transmission within an OFDM symbol period; S42: Turn off the power supply of the radio frequency amplifier corresponding to the idle subcarrier, while maintaining the transmission of the pilot signal.

6. A 5G base station energy efficiency optimization system based on dynamic spectrum sharing, characterized in that: include: Load sensing module, used to collect the number of users, service types and channel quality parameters; A dynamic spectrum sharing module configured to manage resource pooling allocations of licensed and unlicensed frequency bands; Energy efficiency decision engine, which outputs spectrum and power adjustment strategies through reinforcement learning models; The energy-saving execution unit controls the power consumption state of the radio frequency front end and the baseband hardware according to the strategy.

7. The 5G base station energy efficiency optimization system based on dynamic spectrum sharing according to claim 6 is characterized in that: The energy-saving execution unit implements a multi-level sleep strategy: Level 1 sleep: turn off the baseband digital signal processing unit and maintain the RF channel synchronization signal; Secondary sleep: disconnect the power amplifier power supply link, and the adjacent base station takes over the edge user.

8. The 5G base station energy efficiency optimization system based on dynamic spectrum sharing according to claim 6, characterized in that: Further comprising a base station cluster coordination controller, configured to: Centralize control plane signaling to the central node through C / U separation architecture; During the off-peak period of business, user data is migrated to the designated base station, and the remaining base stations enter the second-level sleep mode.

9. The 5G base station energy efficiency optimization system based on dynamic spectrum sharing according to claim 6, characterized in that: The energy efficiency decision engine integrates an LSTM prediction model, which: Input the past 24 hours of traffic volume, weather events and base station location data; Output the spectrum demand forecast within the next 15 minutes with an error rate of less than 8%.

10. The 5G base station energy efficiency optimization system based on dynamic spectrum sharing according to claim 6, characterized in that: The RF front end supports dynamic antenna muting: Turn off the outer loop antenna array according to the user's location information; The beamforming codebook is adaptively adjusted to maintain the RSSI no less than -110dBm.

Images