Dynamic energy-saving management method and system for NR base station

Abstract

The invention provides a dynamic energy-saving management method and system for an NR base station, and relates to the technical field of NR base station energy conservation, and the method comprises the steps: collecting network data, base station state data, external environment data and adjacent region relation data; predicting the network traffic in a future predetermined period to obtain a traffic prediction sequence and corresponding prediction confidence and prediction error range; respectively calculating comprehensive utility function values for the candidate energy-saving strategies, and selecting a target strategy; coordinating a target base station and an adjacent base station to carry out cooperative adjustment of network parameters through a software defined network (SDN) controller so as to carry out coverage compensation before executing an energy-saving strategy; and monitoring an actual energy-saving effect and a user experience index, and performing feedback optimization on decision parameters based on a monitoring result. According to the method provided by the invention, the energy-saving management of the NR base station can be upgraded from a passive, extensive and isolated mode to an active, fine and collaborative intelligent mode, and the network energy efficiency is maximized on the premise of ensuring the user experience.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of NR base station energy saving, in particular to a dynamic energy saving management method and system for an NR base station. BACKGROUND

[0002] A new radio base station (NR base station or gNodeB, gNB for short) is the core wireless access equipment of the fifth generation mobile communication network, responsible for wireless signal transmission and processing with 5G terminals. Compared with 4G base stations, NR base stations realize three application scenarios of enhanced mobile broadband, ultra-reliable low-latency communication and massive machine type communication by introducing large-scale antenna arrays, supporting higher frequency bands, and adopting more flexible frame structures, providing strong network support for the Internet of Things.

[0003] However, the great leap in performance of the NR base station is accompanied by a sharp rise in energy consumption. Its energy consumption can be several times that of a 4G base station under the same coverage conditions, and the high electricity bill has become a major burden for operators to deploy and operate 5G networks, which also violates the green and low-carbon communication development goal. Therefore, energy saving management of the NR base station has become a core problem to be solved in the evolution of 5G networks. At present, the mainstream energy saving scheme still faces many severe challenges in practical application, and the fundamental reason is that the existing technology has the following limitations: First, the energy saving strategy is single and extensive, and cannot achieve fine energy consumption control.

[0004] Second, each base station makes energy saving decisions in isolation, lacks network-level coordination, and is prone to coverage blind spots.

[0005] Third, passive response rather than active execution, energy saving efficiency is relatively low. SUMMARY

[0006] In order to solve the technical problems in the related art, the present application provides a dynamic energy saving management method and system for an NR base station.

[0007] In order to achieve the above purpose, the technical scheme adopted by the present application includes: According to the first aspect of the present application, a dynamic energy saving management method for an NR base station is provided, comprising the following steps: Step S1: Collect network data, base station state data, external environment data and adjacent area relationship data; Step S2: Based on the data collected in step S1, use an artificial intelligence prediction model to predict the network traffic of a future predetermined period to obtain a traffic prediction sequence and its corresponding prediction confidence and prediction error range; Step S3: Based on the traffic prediction sequence, prediction confidence and prediction error range, calculate a comprehensive utility function value for each of the multiple candidate energy-saving strategies, the comprehensive utility function is used to balance between energy-saving benefits, user experience risks and strategy stability, and select the candidate energy-saving strategy with the optimal comprehensive utility function value as the target strategy; Step S4: According to the selected target strategy, coordinate the target base station and its adjacent base stations to perform collaborative adjustment of network parameters through the software-defined network (SDN) controller, so as to compensate for the coverage before executing the energy-saving strategy; Step S5: After executing the target strategy, monitor the actual energy-saving effect and user experience indicators, and perform feedback optimization on the decision parameters based on the monitoring results.

[0008] Optionally, in the step S1, the network data includes historical traffic, real-time traffic, user number and throughput; the base station state data includes power consumption, carrier configuration and antenna parameters; the external environment data includes calendar information and event schedule; and the adjacent area relationship data is obtained based on a measurement report (MR) adjacent area coverage relationship matrix.

[0009] Optionally, in the step S2, the artificial intelligence prediction model is a model combining long short-term memory (LSTM) network and space-time attention mechanism, the prediction confidence and the prediction error range are output by the model; the traffic prediction sequence is , represents the traffic prediction value at time , represents the predicted time interval, represents the total number of steps of prediction.

[0010] Optionally, in step S3, the calculation formula of the comprehensive utility function is as follows: In the formula, represents the th candidate energy-saving strategy, represents the predicted energy-saving amount after adopting the th candidate energy-saving strategy, represents the maximum power consumption of a single base station, represents the comprehensive risk evaluation score brought by the th candidate energy-saving strategy, represents the strategy currently being executed by the base station, represents the strategy switching cost, , and are comprehensive weight coefficients and satisfy​ .

[0011] Optionally, the formula of the comprehensive risk assessment score is: wherein, represents the prediction confidence and is output by the artificial intelligence prediction model, represents the prediction error range and is output by the artificial intelligence prediction model, represents the traffic prediction value at the time of , represents the prediction error rate, represents the synergistic coverage compensation degree and is calculated by the SDN controller, represents the normalized wake-up delay of the th candidate energy-saving strategy, , and are risk weight coefficients respectively and satisfy .

[0012] Optionally, the synergistic coverage compensation degree is determined in the following manner: If the target strategy needs to hibernate part of the resources, the SDN controller calculates whether the surrounding adjacent base stations can form effective compensation to the coverage area of the target base station by adjusting the antenna parameters according to the adjacent area relationship data, and outputs a quantitative value between [0, 1].

[0013] Optionally, in step S5, the feedback optimization specifically includes: comparing the monitored actual energy-saving amount and the user experience index with the prediction value and risk assessment value at the time of decision-making, and adaptively adjusting the comprehensive weight coefficients , and and the risk weight coefficients , and based on the comparison results.

[0014] Optionally, in step S4, the synergistic adjustment of the network parameters includes: The SDN controller instructs the adjacent base stations to dynamically adjust the antenna downtilt angle, azimuth angle or transmission power to expand the adjacent area coverage range before the target base station executes the hibernation strategy, to compensate for the expected coverage holes.

[0015] According to a second aspect of the present invention, a dynamic energy-saving management system for an NR base station is also provided, for executing the dynamic energy-saving management method for an NR base station as described in any of the technical solutions of the first aspect of the present invention, the dynamic energy-saving management system for an NR base station comprising: The data acquisition module is used to collect network data, base station status data, external environment data, and neighbor cell relationship data. An artificial intelligence prediction module, connected to the data acquisition module, is used to predict network traffic volume for a predetermined period based on the data acquired by the data acquisition module and using an artificial intelligence prediction model to obtain a traffic volume prediction sequence and its corresponding prediction confidence and prediction error range. A multi-objective optimization decision module, connected to the artificial intelligence prediction module, is used to calculate a comprehensive utility function value for multiple candidate energy-saving strategies based on the business volume prediction sequence, prediction confidence, and prediction error range. The comprehensive utility function is used to weigh energy-saving benefits, user experience risks, and strategy stability, and selects the candidate energy-saving strategy with the optimal comprehensive utility function value as the target strategy. The SDN collaborative control module, connected to the multi-objective optimization decision module, is used to coordinate the network parameters of the target base station and its neighboring base stations through the software-defined network (SDN) controller according to the selected target strategy, so as to perform coverage compensation before executing the energy-saving strategy. The feedback optimization module is used to monitor the actual energy-saving effect and user experience indicators after the target strategy is executed, and to optimize the decision parameters based on the monitoring results.

[0016] Beneficial effects: 1. Through the above technical solution, firstly, the method of the present invention can achieve refined and forward-looking energy-saving control, overcoming the limitations of single, crude, and passive strategies. Specifically, firstly, unlike traditional solutions that trigger energy saving based on whether the current instantaneous business volume is below a fixed threshold, the method of the present invention uses an artificial intelligence model to predict the business volume for a future cycle. This allows the system to "see" future business trends, thereby making more forward-looking decisions. Secondly, based on the predicted sequence, the method of the present invention can pre-formulate energy-saving strategies before the actual arrival of a business trough, thereby extending the effective energy-saving time window and significantly improving energy-saving efficiency. Simultaneously, it can also prepare in advance before the predicted business peak arrives, avoiding performance degradation. Furthermore, in the present invention, by obtaining the prediction confidence level and prediction error range, the prediction results can be labeled with credibility and risk boundaries, so that subsequent decisions are no longer blindly trusting a prediction value, but can flexibly adjust the aggressiveness of the strategy according to the uncertainty of the prediction, thereby achieving unprecedented refined control.

[0017] Secondly, the method of the present application can realize multi-objective collaborative optimization and solve the contradiction between energy saving and performance. Specifically, first, step S3 explicitly raises the user experience risk and policy stability to the same important decision dimension as the energy saving benefit. This shows that the energy saving decision is no longer a simple binary choice, but a search for the optimal solution in a multi-dimensional target space. Secondly, by defining a comprehensive utility function to uniformly and quantitatively evaluate multiple conflicting objectives, the policy selection can be changed from experience-dependent to scientifically calculated with mathematical models. In this way, the final selected target strategy can ensure that it is a globally optimal balance solution for energy saving, user experience and network stability under the current prediction information and network state.

[0018] Thirdly, the method of the present application can realize network-level collaborative energy saving and effectively avoid the generation of coverage blind area. Specifically, in step S4 of the present application, the key point is to perform coverage compensation before executing the energy saving strategy. This means that when it is decided to let a certain base station enter the energy saving state, the system will not execute immediately, but first instruct its surrounding neighboring base stations through the SDN controller to expand its coverage range by adjusting the network parameters to fill the coverage hole that may be generated after the target base station sleeps. This mechanism can greatly reduce the impact of energy saving behavior on network coverage quality, making it possible to perform deep energy saving while maintaining continuous network coverage, thereby solving the contradiction between local energy saving and global coverage.

[0019] Fourthly, the method of the present application can realize self-learning and continuous optimization of the system, ensuring the long-term adaptability of the scheme. Specifically, by comparing the difference between the prediction and the actual situation, the deviation of the prediction model or the unreasonable setting of the utility function weight can be found, and the decision parameters can be optimized. In this way, the entire management system can continuously learn the new characteristics of the network and the new patterns of the business, continuously improve its prediction accuracy and decision effectiveness, adapt to the dynamic changes of long-term network evolution, and avoid the problem of scheme solidification.

[0020] Overall, the method of the present application, through the organic integration of steps S1 (multi-dimensional data collection), S2 (artificial intelligence prediction and uncertainty quantification), S3 (multi-objective utility decision), S4 (SDN collaborative coverage compensation) and S5 (feedback optimization), has the following synergistically enhanced beneficial effects: refinement and foresight, intelligence and science, collaboration and globality, and adaptability and sustainability.

[0021] 2, Other beneficial effects or advantages of the present application will be described in detail in the specific embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, the drawings described below only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0023] Wherein: Figure 1 is a step flow diagram of a dynamic energy management method for an NR base station provided by an exemplary embodiment of the present application; Figure 2 is a layout diagram of a dynamic energy management system for an NR base station provided by an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0024] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, not all embodiments.

[0025] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, not all embodiments.

[0026] A new radio base station (New Radio Base Station, abbreviated as NR base station or gNodeB, gNB) is a core wireless access device of the fifth generation mobile communication (5G) network, responsible for wireless signal reception and transmission and processing with 5G terminals (UE). Compared with the 4G base station (eNodeB), the NR base station realizes three application scenarios of enhanced mobile broadband (eMBB), ultra-reliable low-latency communication (URLLC) and massive machine type communication (mMTC) by introducing massive antenna arrays (Massive MIMO), supporting higher frequency bands (such as millimeter waves), and adopting more flexible frame structures, etc., providing strong network support for Internet of Things.

[0027] However, the great leap in performance of the NR base station is accompanied by a sharp rise in energy consumption. Its energy consumption can reach several times that of the 4G base station under the same coverage conditions, and the high electricity bill has become the main burden of operators to deploy and operate 5G networks, which also violates the green and low-carbon communication development goal. Therefore, the energy management of the NR base station has become a core problem to be solved in the evolution of the 5G network. At present, the mainstream energy saving scheme still faces many severe challenges in practical application, and the fundamental reason is that the existing technology has the following limitations: First, the energy-saving strategy is single and rough, and cannot achieve fine energy consumption control. The existing scheme mostly adopts a trigger-type energy-saving mechanism based on a fixed threshold, for example, when the traffic volume is lower than a certain preset threshold, the entire carrier is turned off or the base station enters a sleep state. This "one-size-fits-all" strategy is extremely rough. The traffic load of the NR base station presents a high degree of dynamic unevenness in time and space ("tide effect"). The fixed threshold cannot accurately adapt to such rapid and complex traffic changes. This leads to the energy-saving strategy being either too conservative (the threshold is set too high and does not dare to save energy), missing the energy-saving opportunity, or being too aggressive (the threshold is set too low and sleeps too early / too deep), failing to respond quickly when the traffic suddenly arrives, causing network throughput to decrease, time delay to increase, and even connection to be interrupted, seriously affecting user experience.

[0028] Second, the energy-saving decision of each base station is isolated, lacks network-level coordination, and is prone to produce coverage blind spots. The existing energy-saving technology usually makes independent decisions based on a single base station. When a base station enters deep sleep (such as turning off the entire cell) according to its own traffic conditions, signal voids may occur in its coverage area. The propagation characteristics of 5G high-frequency signals determine that their coverage range is relatively small. The coverage relationship between base stations is an organic whole, rather than independent islands from each other. The sleep behavior of a single node will destroy the continuous coverage of the network. The principle behind this is that there is a lack of a centralized intelligent control unit to perceive the global network topology and load state, and to command adjacent base stations to make coordinated coverage compensation (such as dynamically adjusting the antenna downtilt angle or power of the adjacent area). Therefore, local energy saving is at the expense of the overall coverage quality of the network.

[0029] Third, passive response rather than active prediction, low energy-saving efficiency. The scheme in the existing related technology is generally a "rearview mirror" type of passive energy saving, that is, after detecting a low traffic state, the energy-saving action is triggered, and the reaction is lagging. Network traffic, especially traffic based on human activities, often has predictable patterns (such as morning rush hour on weekdays and low traffic at night). The passive response mechanism cannot take advantage of this potential law, so it cannot make energy-saving strategies in advance before the traffic trough arrives, and cannot wake up resources in advance before the traffic peak arrives, resulting in a shortened energy-saving window and an inability to maximize energy-saving efficiency.

[0030] Overall, the existing NR base station energy-saving technology is difficult to achieve deep energy saving while ensuring network performance and user experience due to the rough strategy, lack of coordination, and passive response. Therefore, there is an urgent need for an intelligent, coordinated, and fine dynamic energy-saving management new scheme.

[0031] Therefore, the application provides a brand-new technical solution, that is, a dynamic energy-saving management method for an NR base station. The technical idea of the method is that through three technical pillars of artificial intelligence prediction, multi-objective optimization and SDN coordination, the energy-saving management of the NR base station is upgraded from a passive, extensive and isolated mode to an active, fine and coordinated intelligent mode. The core idea can be briefly summarized as the following three points: First, know before acting, make decisions based on AI prediction. That is, use an artificial intelligence model to predict future network traffic and its uncertainty, change passive response to active planning, and provide forward-looking basis for energy-saving decisions.

[0032] Second, quantitative trade-off, based on multi-objective optimization. That is, construct a comprehensive utility function to unify and quantify the three conflicting objectives of energy-saving benefits, user experience risks and strategy stability, scientifically select the optimal energy-saving strategy, and achieve fine balance.

[0033] Third, collaborative coverage, based on SDN control. That is, before executing a deep energy-saving strategy, use a software-defined network (SDN) controller to intelligently schedule adjacent base stations to adjust parameters to compensate for possible coverage holes, ensuring that the overall network performance is not affected, and solving the contradiction between single-station energy-saving and overall network coverage.

[0034] Finally, a prediction-decision-coordination-feedback closed-loop self-optimization system is formed, which maximizes network energy efficiency while ensuring user experience.

[0035] The technical solutions of the application will be described in detail below with reference to the accompanying drawings.

[0036] As shown in Figure 1 The present embodiment provides a dynamic energy-saving management method for an NR base station according to the first aspect of the application, comprising the following steps: Step S1: Collect network data, base station state data, external environment data and adjacent area relationship data; Step S2: Based on the data collected in step S1, use an artificial intelligence prediction model to predict the network traffic in a future predetermined period to obtain a traffic prediction sequence and its corresponding prediction confidence and prediction error range; Step S3: Based on the traffic prediction sequence, prediction confidence and prediction error range, calculate a comprehensive utility function value for a plurality of candidate energy-saving strategies, the comprehensive utility function is used to weigh between energy-saving benefits, user experience risks and strategy stability, and select the candidate energy-saving strategy with the optimal comprehensive utility function value as the target strategy; Step S4: According to the selected target strategy, the network parameter coordination is performed by the SDN controller to coordinate the target base station and its neighboring base stations to perform coverage compensation before the energy-saving strategy is executed. Step S5: After the target strategy is executed, the actual energy-saving effect and user experience index are monitored, and the decision parameters are feedback optimized based on the monitoring results.

[0037] Through the above technical solution, first, the method of the application can realize fine and forward-looking energy-saving control, and overcome the limitations of single and extensive strategy and passive response. Specifically, first, unlike the traditional scheme which triggers energy saving based on whether the current instantaneous traffic is lower than a fixed threshold, the method of the application performs sequence prediction on the traffic of a future period (not just the current moment) based on an artificial intelligence model. This enables the system to "see" the future traffic trend, thereby making more forward-looking decisions. Second, based on the predicted sequence, the method of the application can develop energy-saving strategies in advance before the traffic trough really arrives (such as after the evening peak, predicting that the night traffic will continue to decline, and planning deep energy saving in advance), thereby extending the effective energy-saving time window to significantly improve energy-saving efficiency. At the same time, it can also prepare in advance before the predicted traffic peak arrives to avoid performance impairment. In addition, in the application, by obtaining the prediction confidence and prediction error range, the prediction result can be labeled with credibility and risk boundary, so that the subsequent decision is no longer blindly believed in a prediction value, but can flexibly adjust the aggressiveness of the strategy according to the uncertainty of the prediction (for example, when the prediction confidence is high, a more aggressive energy-saving strategy can be adopted; when the confidence is low, a more conservative strategy is adopted to guarantee performance), thereby realizing unprecedented fine control.

[0038] Second, the method of the application can realize multi-target coordinated optimization and solve the contradiction between energy saving and performance. Specifically, first, step S3 explicitly raises the user experience risk and strategy stability (which can be understood as the network disturbance cost caused by switching strategies) to the same important decision dimension as energy-saving benefits. This shows that the energy-saving decision is no longer a simple binary choice, but a search for the optimal solution in a multi-dimensional target space. Second, by defining a comprehensive utility function to uniformly and quantitatively evaluate multiple conflicting targets, the strategy selection can be changed from experience-dependent to scientifically calculated with mathematical models. In this way, the final selected target strategy can be a balanced solution that is globally optimal for energy saving, user experience, and network stability under the current prediction information and network state.

[0039] Thirdly, the method of the present application can realize network-level cooperative energy saving, and can effectively avoid the generation of coverage blind area. Specifically, in step S4 of the present application, the key point is to perform coverage compensation before executing the energy saving strategy. This means that when it is decided to let a certain base station (target base station) enter the energy saving state (such as sleep), the system will not immediately execute, but first instruct its surrounding neighboring base stations through the SDN controller to expand its own coverage range by adjusting network parameters (such as power, antenna angle, etc.), to fill in advance the coverage hole that may be generated after the target base station sleeps. This mechanism can greatly reduce the impact of energy saving behavior on network coverage quality, making it possible to perform deep energy saving (such as turning off the entire cell) while maintaining continuous network coverage, thereby solving the contradiction between local energy saving and global coverage.

[0040] Fourthly, the method of the present application can realize self-learning and continuous optimization of the system, and ensure the long-term adaptability of the scheme. Specifically, by comparing the difference between the prediction and the actual situation, the deviation of the prediction model or the unreasonable setting of the utility function weight can be found, and the decision parameters can be optimized (for example, the weight coefficients of each term in the utility function are adjusted below). In this way, the entire management system can continuously learn the new characteristics of the network and the new patterns of the business, continuously improve its prediction accuracy and decision effectiveness, adapt to the dynamic changes of the long-term evolution of the network, and avoid the scheme from becoming outdated.

[0041] Overall, the method of the present application, through the organic integration of the following five steps: step S1 (multi-dimensional data collection), step S2 (artificial intelligence prediction and uncertainty quantification), step S3 (multi-objective utility decision), step S4 (SDN cooperative coverage compensation), and step S5 (feedback optimization), produces the following synergistically enhanced beneficial effects: refinement and foresight (changing the triggering mechanism of energy saving through artificial intelligence prediction, from passive response after the fact to proactive planning in advance), intelligence and scientificity (through the multi-objective utility function, changing the decision-making process from empiricism to scientific quantitative trade-off), cooperation and globality (through SDN centralized cooperation, breaking the base station island, and realizing the cooperation of energy saving and coverage protection), and self-adaptation and sustainability (through the feedback loop, enabling the system to have the ability of continuous learning and self-optimization).

[0042] The method of the present application will be described in detail below in conjunction with a specific embodiment.

[0043] Implementation scenario: A NR base station (identified as base station A) deployed by an operator in a central business district of a city. The traffic in this area presents a typical "tidal effect", i.e., the user density and traffic are high during the daytime on weekdays, and the personnel are sparse and the traffic is extremely low at night and in the early morning. The existing fixed threshold scheme is afraid of not being able to respond to sudden traffic at night, so it does not dare to save energy deeply, resulting in a large amount of energy being wasted during the low traffic period.

[0044] Applying the method of the present application: Step one, collect multi-dimensional data.

[0045] Execution subject: data collection module.

[0046] Specific actions: Collect network data: base station A's historical 24-hour traffic curve, current real-time traffic (only 10% of the peak), number of users (less than 10), throughput (lower than 100Mbps).

[0047] Collect base station state data: base station A's current power consumption 3.5kW, maximum power consumption 4.0kW, active carrier configuration (including 1 main carrier of 100MHz and 1 auxiliary carrier), antenna parameters (downward angle 6 degrees, azimuth angle 0 degrees, transmit power 200W).

[0048] Collect external environment data: calendar information shows that it is currently Wednesday night 11:30, event schedule shows that there are no large-scale night activities in this area within the next 8 hours.

[0049] Collect neighboring area relationship data: based on a large number of user measurement report (MR) data, generate a neighboring area coverage relationship matrix. The matrix shows that base station B (east 500 meters), base station C (south 400 meters) have strong overlapping coverage relationship with base station A, which are ideal cooperative compensation base stations.

[0050] Step two, artificial intelligence traffic prediction.

[0051] Execution subject: artificial intelligence prediction module.

[0052] Specific actions: Input the multi-dimensional data collected in step one into an artificial intelligence prediction model combining long short-term memory network (LSTM) and spatio-temporal attention mechanism.

[0053] Model output traffic prediction sequence : predict traffic in the next 8 hours , The sequence shows that the traffic will continue to decline and maintain at a very low level (for example, from to .

[0054] The model also outputs the prediction confidence and prediction error range of each prediction point. For the night stable period, the confidence is high (such as ), and the error range is small (such as ).

[0055] Step three, multi-objective optimization decision Execution subject: multi-objective optimization decision module.

[0056] Specific actions: Generate multiple candidate energy-saving strategies , candidate energy-saving strategies can be multiple, for example, (mild energy saving: close part of the idle radio frequency channel), (moderate energy saving: close auxiliary carrier), (deep energy saving: hibernate the entire cell, only basic monitoring function is retained).

[0057] Among them, and as two candidate energy-saving strategies. (moderate energy saving), close auxiliary carrier, expected power consumption 2.5kW, energy saving 1.0kW. (deep energy saving), hibernate the entire cell, expected power consumption 0.7kW, energy saving 2.8kW.

[0058] Calculate the comprehensive risk assessment score of each candidate energy-saving strategy . Use the preset risk weight coefficient , , .

[0059] For candidate energy-saving strategy : No compensation, coverage compensation degree , very short wake-up delay . Then =0.4×(1-0.92)×(0.5% / 10%)+0.4×(1-1.0)+0.2×0.1=0.0216.

[0060] For candidate energy-saving strategy : The SDN controller calculates and outputs the cooperative coverage compensation degree , the deep sleep wake-up delay is longer . Then =0.4×(1-0.92)×(0.5% / 10%)+0.4×(1-0.98)+0.2×0.8=0.1696.

[0061] Calculate the comprehensive utility function for each strategy . Use the preset comprehensive weight coefficient , , , the policy switching cost is defined as , .

[0062] For the candidate energy-saving policy : its comprehensive utility function = 0.5 x (1 / 4) - 0.3 x 0.0216 - 0.2 x 0.3 = 0.05852.

[0063] For the candidate energy-saving policy : its comprehensive utility function = 0.5 x (2.8 / 4) - 0.3 x 0.1696 - 0.2 x 0.9 = 0.11912.

[0064] Decision: after calculation Therefore, policy S3 (deep sleep) is selected as the target policy. Its high energy-saving benefits dominate in the utility function, and the collaborative coverage compensation effectively suppresses its coverage risk, making it the optimal solution.

[0065] Step four, SDN collaborative coverage compensation.

[0066] Execution subject: SDN collaborative control module.

[0067] Specific actions: Before instructing base station A to enter sleep, the SDN controller first issues instructions to adjacent base stations B and C.

[0068] The instruction content is to dynamically adjust the network parameters: Instruct base station B to adjust its antenna downtilt angle from 4 degrees to 2 degrees (increase coverage range).

[0069] Instruct base station C to increase its transmit power from 180W to 220W (enhance signal strength).

[0070] These adjustments ensure that the coverage area of base station A is effectively taken over, forming a seamless "coverage compensation network" that completely avoids the generation of coverage holes. After that, the SDN controller issues instructions to base station A to execute the S3 sleep policy.

[0071] Step five, feedback optimization.

[0072] Execution subject: feedback optimization module.

[0073] Specific actions: During the 8 hours of base station A's sleep, the system continuously monitors the actual energy-saving amount The power consumption is 2.75kW and the user experience metrics (such as neighboring cell signal strength, handover success rate, and no dropped calls).

[0074] Compare actual data with predicted values ​​at the time of decision-making ( A comparison was made between the actual energy saving (2.8kW) and the risk assessment value (extremely low risk). It was found that the actual energy saving was slightly lower than the prediction, but the user experience indicators (switching success rate, signal strength) were completely normal, and the coverage risk was 0.

[0075] Based on this result, the parameter optimization process is initiated: Record the energy saving prediction deviation to use in subsequent training of artificial intelligence models and improve prediction accuracy.

[0076] Since coverage compensation has proven to be highly effective (coverage risk is 0), the parameters of the AI ​​model can be fine-tuned to reduce prediction bias, and the weighting coefficients can be adaptively adjusted (e.g., slightly increasing them). Or fine-tuning (To more accurately assess coverage compensation). This allows the system to make more precise decisions when responding to the next energy-saving cycle.

[0077] In this embodiment, it should be noted that, For the comprehensive utility function In this regard, it includes energy-saving benefits. User experience risk items and strategy stability term .

[0078] Among them, for energy-saving benefits In other words, represent The absolute energy-saving effect of the strategy, and by dividing by Normalization is performed to eliminate the dimensional differences caused by different base station scales, making the indicator a dimensionless value between 0 and 1. This allows the algorithm to fairly compare the energy-saving efficiency of different base stations or different strategies. This reflects the operator's strategic preferences: The larger the value, the more aggressive the system decision-making, and the more inclined it is to adopt a high energy-saving strategy.

[0079] For user experience risk items Firstly, the negative sign indicates that risk is a cost that reduces total utility; at the same time, It can also be a comprehensive evaluation model (sub-function), the principle of which lies in the quantification and management of uncertainty. Specifically, it includes predicting uncertainty risk items. Coverage of void risk items and performance degradation risk items Among them, for the uncertainty risk item in forecasting, measures the degree of uncertainty of the model's own prediction, measures the relative size of the prediction error. The product of the two can more accurately depict the potential risk brought by the uncertainty of traffic volume prediction (e.g. when the confidence is low and the error range is large, the risk value increases sharply). This avoids the limitations of a single indicator. For the coverage hole risk term, the principle is to introduce the global view of network synergy, and through the pre-computation of the synergy coverage compensation degree by the SDN controller, the qualitative problem of "whether the hibernation of a single base station will cause coverage problems" is converted into a computable probability or confidence problem (a value between 0 and 1). For the performance degradation risk term, it is used to quantify the possible performance degradation after the execution of the strategy, usually referring to the wake-up delay, which is normalized and included in the evaluation. The above three risk terms are further integrated by risk weight coefficients 、 and to finally output a single score representing the overall risk level of the strategy . Then the system's aversion to risk is controlled.

[0080] For the strategy stability term , this term introduces the smoothness or inertia idea in control theory. Frequent and drastic strategy switching (such as repeatedly jumping between deep hibernation and full power operation) can cause a series of problems, such as signaling storm (a large amount of control signaling interaction is needed for base station state switching), network oscillation (unstable states can cause users to frequently switch between different base stations, which increases overall energy consumption and reduces experience), and hardware wear (frequent switching of radio frequency devices can affect their lifespan). This term quantifies the magnitude or cost of strategy switching. It encourages the system to prefer strategies that are closer to the current state when the energy-saving effects are similar, thereby ensuring the smooth and stable operation of the network. The system's preference for stability is controlled.

[0081] According to a second aspect of the present application, as Figure 2Also shown, there is provided a dynamic energy saving management system for an NR base station for performing the dynamic energy saving management method for an NR base station as in any of the first aspects of the application, the dynamic energy saving management system for an NR base station comprising a data acquisition module, an artificial intelligence prediction module, a multi-objective optimization decision module, an SDN collaborative control module, and a feedback optimization module. The data acquisition module is configured to acquire network data, base station state data, external environment data, and inter-cell relationship data. The artificial intelligence prediction module is connected to the data acquisition module and is configured to, based on the data acquired by the data acquisition module, use an artificial intelligence prediction model to predict network traffic in a future predetermined period to obtain a traffic prediction sequence and corresponding prediction confidence and prediction error range. The multi-objective optimization decision module is connected to the artificial intelligence prediction module and is configured to, based on the traffic prediction sequence, prediction confidence, and prediction error range, calculate a comprehensive utility function value for each of a plurality of candidate energy saving strategies, the comprehensive utility function being configured to weigh between energy saving benefits, user experience risks, and strategy stability, and select a candidate energy saving strategy with the optimal comprehensive utility function value as a target strategy. The SDN collaborative control module is connected to the multi-objective optimization decision module and is configured to, according to the selected target strategy, coordinate the target base station and its neighboring base stations through a software-defined network (SDN) controller to collaboratively adjust network parameters to compensate for coverage before executing the energy saving strategy. The feedback optimization module is configured to monitor actual energy saving effects and user experience indicators after executing the target strategy, and perform feedback optimization on decision parameters based on the monitoring results.

[0082] Through the above technical solutions, the dynamic energy saving management system for an NR base station of the present application can integrate the five core functions of data perception, intelligent prediction, scientific decision-making, collaborative execution, and effect evaluation necessary for NR base station energy saving management into a unified system. This integrated design breaks the pattern of scattered functions and reliance on manual intervention in traditional solutions, achieving high automation of the entire process from data analysis to strategy execution, greatly improving the efficiency and response speed of energy saving management.

[0083] At the same time, this structure of sequential connection with a feedback loop constitutes a complete "perception-decision-execution-learning closed loop". Instead of being an open-loop, one-time decision, the system can continuously monitor the effects of its own decisions (through the feedback optimization module) and use feedback information for optimization. This enables the system to have the ability of self-learning, self-optimization, and self-evolution, to continuously adapt to changes in network traffic patterns, and to maintain optimal energy saving efficiency in the long term, embodying its core intelligent characteristics.

[0084] The above merely illustrates the specific embodiments of the present application, but the protection scope of the present application is not limited thereto, any change or replacement within the technical scope disclosed by the present application should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A dynamic energy-saving management method for NR base stations, characterized in that, Includes the following steps: Step S1: Collect network data, base station status data, external environment data, and neighbor cell relationship data; Step S2: Based on the data collected in step S1, use an artificial intelligence prediction model to predict the network traffic volume for a predetermined period in the future, and obtain the traffic volume prediction sequence and its corresponding prediction confidence and prediction error range. Step S3: Based on the business volume prediction sequence, prediction confidence and prediction error range, calculate a comprehensive utility function value for each of the multiple candidate energy-saving strategies. The comprehensive utility function is used to weigh the energy-saving benefits, user experience risks and strategy stability, and select the candidate energy-saving strategy with the optimal comprehensive utility function value as the target strategy. Step S4: Based on the selected target strategy, the software-defined network (SDN) controller coordinates the target base station and its neighboring base stations to perform coordinated adjustment of network parameters in order to perform coverage compensation before implementing the energy-saving strategy; Step S5: After executing the target strategy, monitor the actual energy-saving effect and user experience indicators, and optimize the decision parameters based on the monitoring results.

2. The dynamic energy-saving management method for NR base stations according to claim 1, characterized in that, In step S1, the network data includes historical traffic volume, real-time traffic volume, number of users, and throughput; the base station status data includes power consumption, carrier configuration, and antenna parameters; and the external environment data includes calendar information and event schedules. The neighbor cell relationship data is obtained based on the neighbor cell coverage relationship matrix of the measurement report MR.

3. The dynamic energy-saving management method for NR base stations according to claim 1, characterized in that, In step S2, the artificial intelligence prediction model is a model combining a long short-term memory network and a spatiotemporal attention mechanism, and the prediction confidence level... and prediction error range The business volume prediction sequence is output by the model. , Indicates in Forecast of traffic volume at any given time Indicates the time interval for prediction. This indicates the total number of steps in the prediction.

4. The dynamic energy-saving management method for NR base stations according to claim 1, characterized in that, In step S3, the comprehensive utility function The calculation formula is: In the formula, Indicates the first One candidate energy-saving strategy Indicates the use of the first The projected energy savings after considering each candidate energy-saving strategy This indicates the maximum power consumption of a single base station. Indicates the use of the first The comprehensive risk assessment score resulting from each candidate energy-saving strategy. This indicates the strategy that the base station is currently executing. Indicates the cost of policy switching. , and These are the comprehensive weight coefficients and satisfy the following conditions: .

5. The dynamic energy-saving management method for NR base stations according to claim 4, characterized in that, The formula for calculating the comprehensive risk assessment score is as follows: In the formula, This represents the prediction confidence level and is output by the artificial intelligence prediction model. This represents the prediction error range and is output by the artificial intelligence prediction model. Indicates in Forecast of traffic volume at any given time Indicates the prediction error rate. The collaborative coverage compensation degree is represented and calculated by the SDN controller. Indicates the first Normalized wake-up latency of each candidate energy-saving strategy , and They are risk weight coefficients and satisfy the following conditions: .

6. The dynamic energy-saving management method for NR base stations according to claim 5, characterized in that, The collaborative coverage compensation degree Determined in the following ways: If the target strategy If some resources need to be put into hibernation, the SDN controller calculates whether the surrounding neighboring base stations can effectively compensate the coverage area of ​​the target base station by adjusting their antenna parameters based on the neighbor cell relationship data, and outputs a quantization value between [0,1].

7. The dynamic energy-saving management method for NR base stations according to claim 6, characterized in that, In step S5, the feedback optimization specifically includes: The actual energy savings monitored The user experience metrics are compared with the predicted values ​​and risk assessment values ​​at the time of decision-making, and the comprehensive weighting coefficients are adaptively adjusted based on the comparison results. , and and risk weight coefficient , and .

8. The dynamic energy-saving management method for NR base stations according to claim 1, characterized in that, In step S4, the coordinated adjustment of the network parameters includes: The SDN controller instructs the neighboring base stations to dynamically adjust their antenna downtilt angle, azimuth angle, or transmit power to expand the coverage of neighboring cells before the target base station executes a sleep strategy, thereby compensating for expected coverage holes.

9. A dynamic energy-saving management system for NR base stations, characterized in that, For executing the dynamic energy-saving management method for an NR base station as described in any one of claims 1-8, the dynamic energy-saving management system for the NR base station includes: The data acquisition module is used to collect network data, base station status data, external environment data, and neighbor cell relationship data. An artificial intelligence prediction module, connected to the data acquisition module, is used to predict network traffic volume for a predetermined period based on the data acquired by the data acquisition module and using an artificial intelligence prediction model to obtain a traffic volume prediction sequence and its corresponding prediction confidence and prediction error range. A multi-objective optimization decision module, connected to the artificial intelligence prediction module, is used to calculate a comprehensive utility function value for multiple candidate energy-saving strategies based on the business volume prediction sequence, prediction confidence, and prediction error range. The comprehensive utility function is used to weigh energy-saving benefits, user experience risks, and strategy stability, and selects the candidate energy-saving strategy with the optimal comprehensive utility function value as the target strategy. The SDN collaborative control module, connected to the multi-objective optimization decision module, is used to coordinate the network parameters of the target base station and its neighboring base stations through the software-defined network (SDN) controller according to the selected target strategy, so as to perform coverage compensation before executing the energy-saving strategy. The feedback optimization module is used to monitor the actual energy-saving effect and user experience indicators after the target strategy is executed, and to optimize the decision parameters based on the monitoring results.

Images