Mobile communication load collaborative scheduling system and method based on energy efficiency perception

Abstract

The invention discloses a mobile communication load collaborative scheduling system and method based on energy efficiency perception, and belongs to the technical field of mobile communication, energy consumption data of each functional module in a base station and external network environment parameters are collected in real time through a multi-source sensor network in the base station, and a dynamic energy efficiency map is generated; based on historical energy efficiency data and real-time load characteristics, constructing an energy efficiency construction prediction model, and predicting an energy efficiency hotspot area; based on the historical movement track of the user and the spatial and temporal distribution characteristics of the service flow, the movement trend of the user and the change trend of the service flow are analyzed, and in combination with an energy efficiency prediction model, a potential overload base station is identified, and overload early warning is triggered; after receiving the early warning information, screening candidate target base stations according to a preset radius by taking the overload base station as a core; and generating a load idle list in a descending order according to the real-time energy efficiency ratio of the candidate base station, and performing energy efficiency sensitivity priority grading on the to-be-migrated service, thereby generating a collaborative scheduling strategy and executing load migration.

Description

Technical Field

[0001] The present invention relates to the field of mobile communication technology, and in particular to a mobile communication load collaborative scheduling system and method based on energy efficiency perception. Background Art

[0002] With the continuous advancement of mobile communication technology, especially the advancement of 5G and future 6G networks, base station clusters, as core components supporting large-scale user connections and data transmission, occupy a vital position in mobile communication networks. The energy efficiency performance of base stations directly affects the operating costs and environmental impact of the entire network. In dynamic load scenarios, the issue of base station energy efficiency is particularly prominent. Traditional mobile communication systems mainly focus on performance indicators such as network coverage quality and signal stability. However, in multi-user and high-load environments, network traffic exhibits dynamic changes in time and space. For example, business demand in urban areas during peak hours is significantly higher than during off-peak hours, and the load distribution of base stations in different geographical locations also fluctuates greatly. This load fluctuation makes it difficult for traditional static resource allocation methods to meet the needs of energy efficiency optimization.

[0003] Existing methods mainly rely on overall energy consumption data at the base station level to make decisions, and fail to deeply analyze the energy consumption characteristics of each functional module within the equipment. This energy efficiency evaluation method can easily lead to deviations between the scheduling strategy and the actual energy efficiency status of the equipment, thereby affecting the overall energy efficiency. In addition, existing scheduling mechanisms mostly focus on load balancing within the communication network, while ignoring the comprehensive impact of load migration on base station energy efficiency. There is a lack of perception and comprehensive evaluation of the dynamic energy efficiency characteristics of each module within the base station. As a result, existing solutions cannot flexibly adjust scheduling strategies when faced with sudden traffic changes or specific environmental factors, resulting in energy efficiency not being optimized. Summary of the Invention

[0004] The purpose of the present invention is to provide a mobile communication load collaborative scheduling system and method based on energy efficiency awareness to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solution: a method for coordinated scheduling of mobile communication loads based on energy efficiency perception, comprising: Step S100: Using a multi-source sensor network within the base station, real-time energy consumption data of each functional module within the base station and external network environment parameters are collected, and a dynamic energy efficiency map is generated through spatiotemporal joint modeling. Step S200: Based on historical energy efficiency data and real-time load characteristics, an energy efficiency prediction model is constructed to output prediction results of energy efficiency hotspot areas within a preset time period in the future; Step S300: Analyze user mobility trends and traffic flow change trends based on historical user movement trajectories and the spatiotemporal distribution characteristics of traffic flow. Based on this, and in conjunction with the base station energy efficiency prediction model, identify potentially overloaded base stations and establish an energy efficiency warning mechanism to provide overload warnings. Step S400: After receiving the warning information, the candidate target base stations are screened according to the preset radius with the overloaded base station as the core; a load idle list is generated in descending order according to the real-time energy efficiency ratio of the candidate base stations, and the energy efficiency sensitivity priority of the services to be migrated is prioritized to generate a collaborative scheduling strategy and execute load migration.

[0006] Furthermore, step S100 includes: Step S101: Data from the radio frequency unit, baseband unit, and environmental module are collected in real time through the sensor network built into each base station. The radio frequency unit includes the instantaneous transmission power, power amplifier efficiency, and energy efficiency ratio of the channel resource block of each antenna channel. The baseband unit includes the computing load rate, task-level energy consumption, and cache utilization of the CPU / GPU / DSP. The environmental module includes the internal temperature of the base station, the power consumption of the cooling system, and the power supply voltage fluctuation rate. External network parameters are synchronously acquired, including the location heat map generated by the user measurement report, the service traffic classification, and the base station topology relationship. The collected data is normalized, and timestamps and spatial tags are added, and stored in a distributed spatiotemporal database. The spatial tag is identified by the base station ID + geographic grid number. Step S102: Modeling spatiotemporal data based on a graph convolutional network generates a dynamic energy efficiency map. Base stations are used as nodes, and node features include real-time energy efficiency, load rate, and temperature. Edge weights are defined by the geographical distance between base stations and the service relevance. The time dimension uses dilated convolution to extract the periodic fluctuation characteristics of energy efficiency, and the spatial dimension uses Chebyshev polynomial approximation to achieve neighborhood base station feature aggregation. The final model outputs a dynamic energy efficiency map, including a device-level energy efficiency heat map, a network-level energy efficiency conduction matrix, and environmental correlation factors. The device-level energy efficiency heat map represents the energy efficiency status of each functional module within the base station; The network-level energy efficiency transfer matrix is a matrix that quantifies the impact of load migration between base stations on energy efficiency. In this matrix, each element represents the impact coefficient of one base station on the energy efficiency of the target base station when the base station migrates load to another base station. The environmental correlation factor represents the coefficient of the influence of external environmental factors on the energy efficiency of the base station.

[0007] Furthermore, step S200 includes: Step S201: Retrieve a sequence of dynamic energy efficiency graphs within a historical time window T1 from a distributed spatiotemporal database, the sequence including a device-level energy efficiency heat map, a network-level energy efficiency conduction matrix, and environmental correlation factors at each timestamp; extract historical data for each base station, including the average energy efficiency, energy efficiency standard deviation, and number of energy efficiency fluctuations of each functional module within the past T1 time period; a sequence of load migration influence coefficients with adjacent base stations in the network-level conduction matrix; temperature-energy efficiency change curve parameters corresponding to environmental correlation factors; synchronously collect real-time load characteristics at the current moment, including the current total service traffic volume of each base station and the traffic proportion of different service types; the number of user connections and the number of active users; the real-time calculation load rate of the CPU / GPU / DSP of the baseband unit; the current transmit power and channel resource block utilization rate of the radio frequency unit; Step S202: Fuse historical data with real-time data to calculate device-level energy efficiency statistics, including the ratio of current energy efficiency to historical average energy efficiency and the deviation between current load rate and historical load rate mean; construct a high-dimensional feature vector containing historical statistical features, real-time load features, and environmental correlation features, with the vector dimension denoted as D; and use the principal component analysis algorithm to reduce the dimensionality of the high-dimensional feature vector and map it to a low-dimensional space R d , retain the principal components whose cumulative variance contribution rate reaches β, and obtain the eigenvector X after dimensionality reduction; Step S203: Based on the fused feature data, a spatiotemporal graph attention network is constructed as an energy efficiency prediction model. The model consists of two parts: a spatiotemporal encoder and a hotspot decoder. The spatiotemporal encoder learns the energy efficiency conduction relationship between base stations through the graph attention mechanism, and captures the periodic law of energy efficiency in combination with the temporal convolutional network. The hotspot decoder maps the model output to a heat map probability distribution, and generates an energy efficiency hotspot area prediction map within the future time window T2. In the model training stage, the optimization objective function is used to optimize the heat map boundary prediction accuracy, and the first α ratio of historical data is used as the training set and the remaining 1-α ratio is used as the validation set. Finally, a dynamic threshold θ is set according to the energy efficiency health index, and the base stations predicted as hotspots are spatially clustered according to the geographic grid number to generate continuous hotspot area boundary coordinates to form a hotspot area prediction set.

[0008] Furthermore, step S300 includes: Step S301: Based on the location heat map and base station topology relationship in the user measurement report, the user's historical movement trajectory is extracted; the user's movement pattern is modeled using a hidden Markov model, where the state space is a set of geographical grids and the observation sequence is the user's position at each time point; the model parameters are estimated using the Baum-Welch algorithm, and the transition probability matrix P of the user migrating from the current grid to other grids is calculated. move ∈R (m×m) , where m is the number of grids; combined with the time factor, construct the time-dependent transition probability matrix Pmove (t), reflects the changes in user mobility patterns in different time periods; at the same time, it quantifies the volatility of business demand and calculates the spatiotemporal entropy value for business traffic classification data: spatial entropy H s Measures the distribution uniformity of business types in different regions at the same time. The calculation formula is: H s =-∑ m ∑ t p(m,t)log p (m, t), where p(m, t) represents the business proportion of grid m at time t; time entropy H t Measures the degree of change in the type of business in the same area at different times. The calculation formula is: H t =-∑ m ∑ t q(m,t)log q (m, t), where q(m, t) represents the service type distribution of grid m at time t; Step S302: User movement probability matrix P move (t) and the spatiotemporal entropy value H of the business traffic s 、H t As input, combined with the energy efficiency hotspot area prediction map output in step S200, potential overload risks are identified through overlay analysis: for each base station b, the expected load increment ΔL(b,t) in the future T2 time window is calculated =∑ m∈N(b) P move (m→b,t)·F f (m, t), where ΔL(b, t) represents the expected load increment of base station b in the future time window T2, N(b) is the neighborhood grid set of base station b, and P move (m→b,t) represents the transfer probability of the user migrating from grid m to base station b at time t, F f (m,t) represents the load of grid m at time t; the expected load increment ΔL(b,t) is input into the energy efficiency prediction model to predict the base station energy efficiency change trend E(b,t), which represents the energy efficiency prediction value of base station b at time t; when E(b,t) is continuously lower than the preset energy efficiency threshold θe and the load rate L(b,t) exceeds the critical value θL, base station b is determined to be a potentially overloaded base station; Step S303: Set a three-level dynamic energy efficiency warning threshold system. When the predicted base station energy efficiency E(b,t) satisfies θe·0.8≤E(b,t)<θe, a first-level warning is triggered; when θe·0.6≤E(b,t)<θe·0.8, a second-level warning is triggered; when E(b,t)<θe·0.6, a third-level warning is triggered; the warning signal includes the base station ID and the expected overload time; a warning diffusion network is constructed based on the base station topology relationship. When a base station triggers a warning, its correlation weight with the adjacent base station is calculated, and the correlation weight is calculated by the geographical distance, service correlation and historical collaborative load rate; the warning information is diffused to the adjacent base station set N(b) in descending order of the correlation weight, and the diffusion depth is d, which is used to control the propagation range of the warning information; the triggered warning is tracked in real time, and the warning is automatically lifted when the actual energy efficiency returns to the safe range.

[0009] Furthermore, step S400 includes: Step S401: After receiving an overload warning, a set of candidate target base stations C is screened based on a preset geographic radius R with the overloaded base station as the core, and the comprehensive compatibility of each candidate base station c with the overloaded base station is calculated: F(c) = α·ER(c) + β·(1-d(c,b) / R) + γ·C(c), where ER(c) is the real-time energy efficiency ratio, d(c,b) is the geographic distance, C(c) is the percentage of remaining available capacity, and α, β, and γ are weight coefficients. Candidate base stations with a transmission delay exceeding t_max, a remaining capacity below C_min, and an energy efficiency ratio below ER_min are selected, and the remaining base stations are sorted in descending order of F(c) to generate a load idle list L. t_max, C_min, and ER_min are preset transmission delay thresholds, remaining capacity thresholds, and energy efficiency ratio thresholds, respectively. Step S402: Perform energy efficiency sensitivity assessment on the service set S to be migrated, and calculate the energy efficiency sensitivity index (ESI(s)) of service s = (δ·RTD(s) + ε·BW(s)) / ER(s), where RTD(s) is the real-time requirement, BW(s) is the bandwidth requirement, ER(s) is the current energy efficiency ratio, and δ and ε are weight coefficients. Classify the services into three priority levels: high, medium, and low based on ESI(s). After completing the priority division, a matching scheme between services and base stations is constructed by comprehensively considering service priority, base station real-time energy efficiency ratio, and the geographical distance between the base station and the overloaded base station. For each service s to be migrated, the objective function is: max∑ s∈S ∑ c∈L M(s,c)·(w1·ESI(s)·ER(c)-w2·d(c,b)); Among them, M(s,c) is a decision variable, which takes the value of 1 when service s is assigned to base station c and 0 otherwise; w1 and w2 are weight coefficients used to balance the service energy efficiency sensitivity, base station energy efficiency ratio and geographical distance; the set constraints include capacity constraint, delay constraint and integrity constraint; the capacity constraint is to ensure that the remaining capacity of the target base station can carry the bandwidth requirements of the migrated service; the delay constraint is to meet the service's requirements for transmission delay; the integrity constraint is that each service must and can only be assigned to one target base station; by solving the optimization problem, a collaborative scheduling strategy containing the service migration path, migration data volume and execution time window is generated and sent to the relevant base stations through the SDN controller. Each base station performs load scheduling according to the scheduling strategy.

[0010] A mobile communication load collaborative scheduling system based on energy efficiency perception, including an energy efficiency data acquisition module, an energy efficiency prediction and hotspot identification module, an early warning trigger module, and a load scheduling module; The energy efficiency data acquisition module collects energy consumption data of each functional module in the base station and external network environment parameters in real time through the multi-source sensor network inside the base station, and generates a dynamic energy efficiency map through spatiotemporal joint modeling; The energy efficiency prediction and hotspot identification module builds an energy efficiency prediction model based on historical energy efficiency data and real-time load characteristics, and outputs the prediction results of energy efficiency hotspot areas within a preset time period in the future; The warning trigger module analyzes user mobility trends and traffic flow trends based on historical user movement trajectories and the spatiotemporal distribution characteristics of traffic flow. Based on this, it combines the base station energy efficiency prediction model to identify potentially overloaded base stations and establishes an energy efficiency warning mechanism to provide overload warnings. After receiving the warning information, the load scheduling module takes the overloaded base station as the core and screens the candidate target base stations according to the preset radius; generates a load idle list in descending order according to the real-time energy efficiency ratio of the candidate base stations, and prioritizes the energy efficiency sensitivity of the services to be migrated, thereby generating a collaborative scheduling strategy and executing load migration.

[0011] Compared with the prior art, the present invention has the following beneficial effects: This invention uses a multi-source sensor network within a base station and spatiotemporal joint modeling technology to collect and analyze energy consumption data and network environment parameters in real time, generate dynamic energy efficiency maps, and predict energy efficiency hotspots. This method can identify potentially overloaded base stations in advance and dynamically adjust load distribution based on the energy efficiency and load characteristics of the base stations to optimize energy efficiency, thereby improving the energy utilization efficiency of mobile communication networks and reducing overload risks. By combining users' historical movement trajectories, business traffic changes and base station energy efficiency prediction models, the present invention establishes an energy-efficiency-aware load collaborative scheduling strategy, and implements base station overload warning through a multi-level warning mechanism. It can effectively respond to network load changes, dynamically adjust resource allocation, achieve precise load migration, and optimize resource configuration. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 The present invention is a flow chart of a method for collaborative scheduling of mobile communication loads based on energy efficiency awareness. DETAILED DESCRIPTION

[0013] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 are within the scope of protection of the present invention.

[0014] See also Figure 1 The present invention provides a technical solution: a method for collaborative scheduling of mobile communication loads based on energy efficiency perception, comprising: Step S100: Using a multi-source sensor network within the base station, real-time energy consumption data of each functional module within the base station and external network environment parameters are collected, and a dynamic energy efficiency map is generated through spatiotemporal joint modeling. Step S200: Based on historical energy efficiency data and real-time load characteristics, an energy efficiency prediction model is constructed to output prediction results of energy efficiency hotspot areas within a preset time period in the future; Step S300: Analyze user mobility trends and traffic flow change trends based on historical user movement trajectories and the spatiotemporal distribution characteristics of traffic flow. Based on this, and in conjunction with the base station energy efficiency prediction model, identify potentially overloaded base stations and establish an energy efficiency warning mechanism to provide overload warnings. Step S400: After receiving the warning information, the candidate target base stations are screened according to the preset radius with the overloaded base station as the core; a load idle list is generated in descending order according to the real-time energy efficiency ratio of the candidate base stations, and the energy efficiency sensitivity priority of the services to be migrated is prioritized to generate a collaborative scheduling strategy and execute load migration.

[0015] Furthermore, step S100 includes: Step S101: Data from the radio frequency unit, baseband unit, and environmental module are collected in real time through the sensor network built into each base station. The radio frequency unit includes the instantaneous transmission power, power amplifier efficiency, and energy efficiency ratio of the channel resource block of each antenna channel. The baseband unit includes the computing load rate, task-level energy consumption, and cache utilization of the CPU / GPU / DSP. The environmental module includes the internal temperature of the base station, the power consumption of the cooling system, and the power supply voltage fluctuation rate. External network parameters are synchronously acquired, including the location heat map generated by the user measurement report, the service traffic classification, and the base station topology relationship. The collected data is normalized, and timestamps and spatial tags are added, and stored in a distributed spatiotemporal database. The spatial tag is identified by the base station ID + geographic grid number. Step S102: Modeling spatiotemporal data based on a graph convolutional network generates a dynamic energy efficiency map. Base stations are used as nodes, and node features include real-time energy efficiency, load rate, and temperature. Edge weights are defined by the geographical distance between base stations and the service relevance. The time dimension uses dilated convolution to extract the periodic fluctuation characteristics of energy efficiency, and the spatial dimension uses Chebyshev polynomial approximation to achieve neighborhood base station feature aggregation. The final model outputs a dynamic energy efficiency map, including a device-level energy efficiency heat map, a network-level energy efficiency conduction matrix, and environmental correlation factors. The device-level energy efficiency heat map represents the energy efficiency status of each functional module within the base station; The network-level energy efficiency transfer matrix is a matrix that quantifies the impact of load migration between base stations on energy efficiency. In this matrix, each element represents the impact coefficient of one base station on the energy efficiency of the target base station when the base station migrates load to another base station. The environmental correlation factor represents the coefficient of the influence of external environmental factors on the energy efficiency of the base station.

[0016] Furthermore, step S200 includes: Step S201: Retrieve a sequence of dynamic energy efficiency graphs within a historical time window T1 from a distributed spatiotemporal database, the sequence including a device-level energy efficiency heat map, a network-level energy efficiency conduction matrix, and environmental correlation factors at each timestamp; extract historical data for each base station, including the average energy efficiency, energy efficiency standard deviation, and number of energy efficiency fluctuations of each functional module within the past T1 time period; a sequence of load migration influence coefficients with adjacent base stations in the network-level conduction matrix; temperature-energy efficiency change curve parameters corresponding to environmental correlation factors; synchronously collect real-time load characteristics at the current moment, including the current total service traffic volume of each base station and the traffic proportion of different service types; the number of user connections and the number of active users; the real-time calculation load rate of the CPU / GPU / DSP of the baseband unit; the current transmit power and channel resource block utilization rate of the radio frequency unit; Step S202: Fuse historical data with real-time data to calculate device-level energy efficiency statistics, including the ratio of current energy efficiency to historical average energy efficiency and the deviation between current load rate and historical load rate mean; construct a high-dimensional feature vector containing historical statistical features, real-time load features, and environmental correlation features, with the vector dimension denoted as D; and use the principal component analysis algorithm to reduce the dimensionality of the high-dimensional feature vector and map it to a low-dimensional space R d , retain the principal components whose cumulative variance contribution rate reaches β, and obtain the eigenvector X after dimensionality reduction; Step S203: Based on the fused feature data, a spatiotemporal graph attention network is constructed as an energy efficiency prediction model. The model consists of two parts: a spatiotemporal encoder and a hotspot decoder. The spatiotemporal encoder learns the energy efficiency conduction relationship between base stations through the graph attention mechanism, and captures the periodic law of energy efficiency in combination with the temporal convolutional network. The hotspot decoder maps the model output to a heat map probability distribution, and generates an energy efficiency hotspot area prediction map within the future time window T2. In the model training stage, the optimization objective function is used to optimize the heat map boundary prediction accuracy, and the first α ratio of historical data is used as the training set and the remaining 1-α ratio is used as the validation set. Finally, a dynamic threshold θ is set according to the energy efficiency health index, and the base stations predicted as hotspots are spatially clustered according to the geographic grid number to generate continuous hotspot area boundary coordinates to form a hotspot area prediction set.

[0017] Furthermore, step S300 includes: Step S301: Based on the location heat map and base station topology relationship in the user measurement report, the user's historical movement trajectory is extracted; the user's movement pattern is modeled using a hidden Markov model, where the state space is a set of geographical grids and the observation sequence is the user's position at each time point; the model parameters are estimated using the Baum-Welch algorithm, and the transition probability matrix P of the user migrating from the current grid to other grids is calculated. move ∈R (m×m) , where m is the number of grids; combined with the time factor, construct the time-dependent transition probability matrix P move (t), reflects the changes in user mobility patterns in different time periods; at the same time, it quantifies the volatility of business demand and calculates the spatiotemporal entropy value for business traffic classification data: spatial entropy H s Measures the distribution uniformity of business types in different regions at the same time. The calculation formula is: H s =-∑ m ∑ t p(m,t)log p (m, t), where p(m, t) represents the business proportion of grid m at time t; time entropy H t Measures the degree of change in the type of business in the same area at different times. The calculation formula is: H t =-∑ m ∑ tq(m,t)log q (m, t), where q(m, t) represents the service type distribution of grid m at time t; Step S302: User movement probability matrix P move (t) and the spatiotemporal entropy value H of the business traffic s 、H t As input, combined with the energy efficiency hotspot prediction map output in step S200, potential overload risks are identified through overlay analysis: for each base station b, the expected load increment within the future T2 time window is calculated; the expected load increment is input into the energy efficiency prediction model to predict the base station energy efficiency change trend E(b,t); when E(b,t) is continuously lower than the preset energy efficiency threshold θe and the load rate L(b,t) exceeds the critical value θL, base station b is determined to be a potential overloaded base station; Step S303: Set a three-level dynamic energy efficiency warning threshold system. When the predicted base station energy efficiency E(b,t) satisfies θe·0.8≤E(b,t)<θe, a first-level warning is triggered; when θe·0.6≤E(b,t)<θe·0.8, a second-level warning is triggered; when E(b,t)<θe·0.6, a third-level warning is triggered; the warning signal includes the base station ID and the expected overload time; a warning diffusion network is constructed based on the base station topology relationship. When base station b triggers a warning, its correlation weight with the adjacent base station is calculated, and the correlation weight is calculated by the geographical distance, service correlation and historical collaborative load rate; the warning information is diffused to the adjacent base station set N(b) in descending order of the correlation weight, with a diffusion depth of d; the triggered warning is tracked in real time, and the warning is automatically lifted when the actual energy efficiency returns to the safe range.

[0018] Furthermore, step S400 includes: Step S401: After receiving an overload warning, a set of candidate target base stations C is screened based on a preset geographic radius R with the overloaded base station as the core, and the comprehensive compatibility of each candidate base station c with the overloaded base station is calculated: F(c) = α·ER(c) + β·(1-d(c,b) / R) + γ·C(c), where ER(c) is the real-time energy efficiency ratio, d(c,b) is the geographic distance, C(c) is the percentage of remaining available capacity, and α, β, and γ are weight coefficients. Candidate base stations with a transmission delay exceeding t_max, a remaining capacity below C_min, and an energy efficiency ratio below ER_min are selected, and the remaining base stations are sorted in descending order of F(c) to generate a load idle list L. t_max, C_min, and ER_min are preset transmission delay thresholds, remaining capacity thresholds, and energy efficiency ratio thresholds, respectively. Step S402: Perform energy efficiency sensitivity assessment on the service set S to be migrated, and calculate the energy efficiency sensitivity index (ESI(s)) of service s = (δ·RTD(s) + ε·BW(s)) / ER(s), where RTD(s) is the real-time requirement, BW(s) is the bandwidth requirement, ER(s) is the current energy efficiency ratio, and δ and ε are weight coefficients. Classify the services into three priority levels: high, medium, and low based on ESI(s). After completing the priority division, a matching scheme between services and base stations is constructed by comprehensively considering service priority, base station real-time energy efficiency ratio, and the geographical distance between the base station and the overloaded base station. For each service s to be migrated, the objective function is: max∑ s∈S ∑ c∈L M(s,c)·(w1·ESI(s)·ER(c)-w2·d(c,b)); Among them, M(s,c) is a decision variable, which takes the value of 1 when service s is assigned to base station c and 0 otherwise; w1 and w2 are weight coefficients used to balance the service energy efficiency sensitivity, base station energy efficiency ratio and geographical distance; the set constraints include capacity constraint, delay constraint and integrity constraint; the capacity constraint is to ensure that the remaining capacity of the target base station can carry the bandwidth requirements of the migrated service; the delay constraint is to meet the service's requirements for transmission delay; the integrity constraint is that each service must and can only be assigned to one target base station; by solving the optimization problem, a collaborative scheduling strategy containing the service migration path, migration data volume and execution time window is generated and sent to the relevant base stations through the SDN controller. Each base station performs load scheduling according to the scheduling strategy.

[0019] In the embodiment of the present invention, a 5G base station cluster in a commercial area of a city is used as an example, which includes 10 macro base stations and 20 micro base stations. Step S100: Deploy sensors at each base station and collect data every 10 minutes: the radio frequency unit collects antenna transmission power and power amplifier efficiency; the baseband unit obtains CPU load rate and task energy consumption; the environmental module monitors temperature and heat dissipation power consumption; and simultaneously obtains user location heat maps and service traffic classification data. After normalization, a dynamic energy efficiency map is constructed based on a graph convolutional network. Node features are set to real-time energy efficiency, load rate, and temperature. Edge weights are determined based on the geographical distance between base stations and service relevance. This generates a device-level energy efficiency heat map and a network-level energy efficiency conduction matrix. Step S200: Energy efficiency hotspot prediction retrieves a sequence of dynamic energy efficiency maps from the past seven days and, combined with real-time data such as the current service traffic of each base station and the number of user connections, constructs a vector containing 100-dimensional features. Principal component analysis is used to reduce the dimensionality to 20 dimensions, retaining 95% of the variance contribution rate. Based on spatiotemporal graph attention network modeling, the training set accounts for 80% of the historical data and the validation set accounts for 20%. The prediction time window T2 is set to 1 hour, and a prediction map of energy efficiency hotspots for the next hour is output. Step S300: Analyze historical user movement trajectories, divide the area into 100 geographic grids, and calculate the user transition probability matrix between grids; calculate the spatiotemporal entropy value of service traffic to quantify service demand fluctuations; combine with the energy efficiency hotspot prediction map to calculate the expected load increase of each base station in the next hour. When the base station energy efficiency is lower than the threshold θe (set to 80% of the baseline energy efficiency) for three consecutive monitoring cycles and the load rate exceeds the critical value θL (70%), an early warning is triggered. The early warning is divided into three levels: the first level (energy efficiency is between 80% and 90% of the baseline value) is prompted by the internal system, the second level (60% to 80%) notifies the operation and maintenance personnel, and the third level (below 60%) automatically initiates the scheduling process; Step S400: Load collaborative scheduling: When base station B triggers the third-level warning, 5 candidate target base stations are screened within a radius of 2 kilometers with base station B as the center; the comprehensive adaptability of the candidate base stations is calculated, and one base station with excessive transmission delay is eliminated. A load idle list is generated according to the adaptability, and the energy efficiency sensitivity index is calculated for the migrated services. Real-time services such as high-definition video streaming are set to high priority, file downloads are set to medium priority, and background synchronization is set to low priority; based on the objective function, the optimal matching solution between the service and the base station is solved, and a scheduling strategy is generated: high-priority services are migrated to base station C with the highest energy efficiency ratio, and medium and low-priority services are dispersed to other adaptive base stations; instructions are issued through the SDN controller to complete load migration within 10 minutes.

[0020] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A method for collaborative scheduling of mobile communication loads based on energy efficiency awareness, characterized by: The method comprises: Step S100: Using a multi-source sensor network within the base station, real-time energy consumption data of each functional module within the base station and external network environment parameters are collected, and a dynamic energy efficiency map is generated through spatiotemporal joint modeling. Step S200: Based on historical energy efficiency data and real-time load characteristics, an energy efficiency prediction model is constructed to output prediction results of energy efficiency hotspot areas within a preset time period in the future; Step S300: Analyze user mobility trends and traffic flow change trends based on historical user movement trajectories and the spatiotemporal distribution characteristics of traffic flow. Based on this, and in conjunction with the base station energy efficiency prediction model, identify potentially overloaded base stations and establish an energy efficiency warning mechanism to provide overload warnings. Step S400: After receiving the warning information, the candidate target base stations are screened according to the preset radius with the overloaded base station as the core; a load idle list is generated in descending order according to the real-time energy efficiency ratio of the candidate base stations, and the energy efficiency sensitivity priority of the services to be migrated is prioritized to generate a collaborative scheduling strategy and execute load migration.

2. The method for coordinated scheduling of mobile communication loads based on energy efficiency awareness according to claim 1, characterized in that: The step S100 includes: Step S101: Data from the radio frequency unit, baseband unit, and environmental module are collected in real time through the sensor network built into each base station. The radio frequency unit includes the instantaneous transmission power, power amplifier efficiency, and energy efficiency ratio of the channel resource block of each antenna channel. The baseband unit includes the computing load rate, task-level energy consumption, and cache utilization of the CPU / GPU / DSP. The environmental module includes the internal temperature of the base station, the power consumption of the cooling system, and the power supply voltage fluctuation rate. External network parameters are synchronously acquired, including the location heat map generated by the user measurement report, the service traffic classification, and the base station topology relationship. The collected data is normalized, and timestamps and spatial tags are added, and stored in a distributed spatiotemporal database. The spatial tag is identified by the base station ID + geographic grid number. Step S102: Modeling spatiotemporal data based on a graph convolutional network generates a dynamic energy efficiency map. Base stations are used as nodes, and node features include real-time energy efficiency, load rate, and temperature. Edge weights are defined by the geographical distance between base stations and the service relevance. The time dimension uses dilated convolution to extract the periodic fluctuation characteristics of energy efficiency, and the spatial dimension uses Chebyshev polynomial approximation to achieve neighborhood base station feature aggregation. The final model outputs a dynamic energy efficiency map, including a device-level energy efficiency heat map, a network-level energy efficiency conduction matrix, and environmental correlation factors. The device-level energy efficiency heat map represents the energy efficiency status of each functional module within the base station; The network-level energy efficiency transfer matrix is a matrix that quantifies the impact of load migration between base stations on energy efficiency. In this matrix, each element represents the impact coefficient of one base station on the energy efficiency of the target base station when the base station migrates load to another base station. The environmental correlation factor represents the coefficient of the influence of external environmental factors on the energy efficiency of the base station.

3. The method for coordinated scheduling of mobile communication loads based on energy efficiency awareness according to claim 1, characterized in that: The step S200 includes: Step S201: Retrieve a sequence of dynamic energy efficiency graphs within a historical time window T1 from a distributed spatiotemporal database, the sequence including a device-level energy efficiency heat map, a network-level energy efficiency conduction matrix, and environmental correlation factors at each timestamp; extract historical data for each base station, including the average energy efficiency, energy efficiency standard deviation, and number of energy efficiency fluctuations of each functional module within the past T1 time period; a sequence of load migration influence coefficients with adjacent base stations in the network-level conduction matrix; temperature-energy efficiency change curve parameters corresponding to environmental correlation factors; synchronously collect real-time load characteristics at the current moment, including the current total service traffic volume of each base station and the traffic proportion of different service types; the number of user connections and the number of active users; the real-time calculation load rate of the CPU / GPU / DSP of the baseband unit; the current transmit power and channel resource block utilization rate of the radio frequency unit; Step S202: Fuse historical data with real-time data to calculate device-level energy efficiency statistics, including the ratio of current energy efficiency to historical average energy efficiency and the deviation between current load rate and historical load rate mean; construct a high-dimensional feature vector containing historical statistical features, real-time load features, and environmental correlation features, with the vector dimension denoted as D; and use the principal component analysis algorithm to reduce the dimensionality of the high-dimensional feature vector and map it to a low-dimensional space R d , retain the principal components whose cumulative variance contribution rate reaches β, and obtain the eigenvector X after dimensionality reduction; Step S203: Based on the fused feature data, a spatiotemporal graph attention network is constructed as an energy efficiency prediction model. The model consists of two parts: a spatiotemporal encoder and a hotspot decoder. The spatiotemporal encoder learns the energy efficiency conduction relationship between base stations through the graph attention mechanism, and captures the periodic law of energy efficiency in combination with the temporal convolutional network. The hotspot decoder maps the model output to a heat map probability distribution, and generates an energy efficiency hotspot area prediction map within the future time window T2. In the model training stage, the optimization objective function is used to optimize the heat map boundary prediction accuracy, and the first α ratio of historical data is used as the training set and the remaining 1-α ratio is used as the validation set. Finally, a dynamic threshold θ is set according to the energy efficiency health index, and the base stations predicted as hotspots are spatially clustered according to the geographic grid number to generate continuous hotspot area boundary coordinates to form a hotspot area prediction set.

4. The method for coordinated scheduling of mobile communication loads based on energy efficiency awareness according to claim 1, characterized in that: The step S300 includes: Step S301: Based on the location heat map and base station topology relationship in the user measurement report, the user's historical movement trajectory is extracted; the user's movement pattern is modeled using a hidden Markov model, where the state space is a set of geographical grids and the observation sequence is the user's position at each time point; the model parameters are estimated using the Baum-Welch algorithm, and the transition probability matrix P of the user migrating from the current grid to other grids is calculated. move ∈R (m×m) , where m is the number of grids; combined with the time factor, construct the time-dependent transition probability matrix P move (t), reflects the changes in user mobility patterns in different time periods; at the same time, it quantifies the volatility of business demand and calculates the spatiotemporal entropy value for business traffic classification data: spatial entropy H s Measures the distribution uniformity of business types in different regions at the same time. The calculation formula is: H s =-∑ m ∑ t p(m,t)log p (m, t), where p(m, t) represents the business proportion of grid m at time t; time entropy H t Measures the degree of change in the type of business in the same area at different times. The calculation formula is: H t =-∑ m ∑ t q(m,t)log q (m, t), where q(m, t) represents the service type distribution of grid m at time t; Step S302: User movement probability matrix P move (t) and the spatiotemporal entropy value H of the business traffic s 、H t As input, combined with the energy efficiency hotspot prediction map output in step S200, potential overload risks are identified through overlay analysis: for each base station b, the expected load increment within the future T2 time window is calculated; the expected load increment is input into the energy efficiency prediction model to predict the base station energy efficiency change trend E(b,t); when E(b,t) is continuously lower than the preset energy efficiency threshold θe and the load rate L(b,t) exceeds the critical value θL, base station b is determined to be a potential overloaded base station; Step S303: Set a three-level dynamic energy efficiency warning threshold system. When the predicted base station energy efficiency E(b,t) satisfies θe·0.8≤E(b,t)<θe, a first-level warning is triggered; when θe·0.6≤E(b,t)<θe·0.8, a second-level warning is triggered; when E(b,t)<θe·0.6, a third-level warning is triggered; the warning signal includes the base station ID and the expected overload time; a warning diffusion network is constructed based on the base station topology relationship. When base station b triggers a warning, its correlation weight with the adjacent base station is calculated, and the correlation weight is calculated by the geographical distance, service correlation and historical collaborative load rate; the warning information is diffused to the adjacent base station set N(b) in descending order of the correlation weight, with a diffusion depth of d; the triggered warning is tracked in real time, and the warning is automatically lifted when the actual energy efficiency returns to the safe range.

5. The method for coordinated scheduling of mobile communication loads based on energy efficiency awareness according to claim 1, characterized in that: The step S400 includes: Step S401: After receiving an overload warning, a set of candidate target base stations C is screened based on a preset geographic radius R with the overloaded base station as the core, and the comprehensive compatibility of each candidate base station c with the overloaded base station is calculated: F(c) = α·ER(c) + β·(1-d(c,b) / R) + γ·C(c), where ER(c) is the real-time energy efficiency ratio, d(c,b) is the geographic distance, C(c) is the percentage of remaining available capacity, and α, β, and γ are weight coefficients. Candidate base stations with a transmission delay exceeding t_max, a remaining capacity below C_min, and an energy efficiency ratio below ER_min are selected, and the remaining base stations are sorted in descending order of F(c) to generate a load idle list L. t_max, C_min, and ER_min are preset transmission delay thresholds, remaining capacity thresholds, and energy efficiency ratio thresholds, respectively. Step S402: Perform energy efficiency sensitivity assessment on the service set S to be migrated, and calculate the energy efficiency sensitivity index (ESI(s)) of service s = (δ·RTD(s) + ε·BW(s)) / ER(s), where RTD(s) is the real-time requirement, BW(s) is the bandwidth requirement, ER(s) is the current energy efficiency ratio, and δ and ε are weight coefficients. Classify the services into three priority levels: high, medium, and low based on ESI(s). After completing the priority division, a matching scheme between services and base stations is constructed by comprehensively considering service priority, base station real-time energy efficiency ratio, and the geographical distance between the base station and the overloaded base station. For each service s to be migrated, the objective function is: max∑ s∈S ∑ c∈L M(s,c)·(w1·ESI(s)·ER(c)-w2·d(c,b)); Among them, M(s,c) is a decision variable, which takes the value of 1 when service s is assigned to base station c and 0 otherwise; w1 and w2 are weight coefficients used to balance the service energy efficiency sensitivity, base station energy efficiency ratio and geographical distance; the set constraints include capacity constraint, delay constraint and integrity constraint; the capacity constraint is to ensure that the remaining capacity of the target base station can carry the bandwidth requirements of the migrated service; the delay constraint is to meet the service's requirements for transmission delay; the integrity constraint is that each service must and can only be assigned to one target base station; by solving the optimization problem, a collaborative scheduling strategy containing the service migration path, migration data volume and execution time window is generated and sent to the relevant base stations through the SDN controller. Each base station performs load scheduling according to the scheduling strategy.

6. A mobile communication load collaborative scheduling system based on energy efficiency awareness, characterized by: The system includes an energy efficiency data acquisition module, an energy efficiency prediction and hotspot identification module, an early warning trigger module, and a load scheduling module; The energy efficiency data acquisition module collects energy consumption data of each functional module in the base station and external network environment parameters in real time through the multi-source sensor network inside the base station, and generates a dynamic energy efficiency map through spatiotemporal joint modeling; The energy efficiency prediction and hotspot identification module builds an energy efficiency prediction model based on historical energy efficiency data and real-time load characteristics, and outputs the prediction results of energy efficiency hotspot areas within a preset time period in the future; The warning trigger module analyzes user mobility trends and traffic flow trends based on historical user movement trajectories and the spatiotemporal distribution characteristics of traffic flow. Based on this, it combines the base station energy efficiency prediction model to identify potentially overloaded base stations and establishes an energy efficiency warning mechanism to provide overload warnings. After receiving the warning information, the load scheduling module takes the overloaded base station as the core and screens the candidate target base stations according to the preset radius; generates a load idle list in descending order according to the real-time energy efficiency ratio of the candidate base stations, and prioritizes the energy efficiency sensitivity of the services to be migrated, thereby generating a collaborative scheduling strategy and executing load migration.

7. The mobile communication load collaborative scheduling system based on energy efficiency awareness according to claim 6, characterized in that: The energy efficiency data acquisition module includes a multi-source data acquisition unit and an energy efficiency spectrum generation unit; the multi-source data acquisition unit collects radio frequency, baseband and environmental data in real time through base station sensors, synchronously obtains network parameters, and performs normalization processing; And add spatiotemporal labels and store them in a distributed database; the energy efficiency map generation unit extracts features from spatiotemporal data based on a graph convolutional network, generates a device-level energy efficiency heat map, a network-level conduction matrix, and environmental correlation factors, and constructs a dynamic energy efficiency map.

8. The mobile communication load collaborative scheduling system based on energy efficiency awareness according to claim 6, characterized in that: The energy efficiency prediction and hotspot identification module includes a data fusion and feature extraction unit and an energy efficiency hotspot prediction unit; the data fusion and feature extraction unit fuses historical energy efficiency graph sequences with real-time load features, and generates a high-dimensional spatiotemporal feature vector through feature engineering dimensionality reduction processing; the energy efficiency hotspot prediction unit constructs an energy efficiency prediction model based on the spatiotemporal graph attention network, predicts future energy efficiency hotspot areas, combines the spatial clustering algorithm to generate hotspot boundary coordinates, and outputs the prediction set results.

9. The mobile communication load collaborative scheduling system based on energy efficiency awareness according to claim 6, characterized in that: The warning trigger module includes a user trajectory modeling unit, an overload identification unit, and a hierarchical warning unit; the user trajectory modeling unit analyzes the user's historical movement trajectory and calculates the transition probability matrix based on the base station topology relationship to predict the user's future movement trend; The overload identification unit identifies potentially overloaded base stations by quantifying the spatiotemporal entropy of service traffic and combining it with energy efficiency prediction results; the hierarchical warning unit triggers multi-level warnings based on dynamic energy efficiency thresholds, and spreads warning information to adjacent base stations through base station association weights, forming a regional collaborative defense mechanism.

10. The mobile communication load collaborative scheduling system based on energy efficiency awareness according to claim 6, characterized in that: The load scheduling module includes a candidate base station screening unit, a service priority classification unit, and a load migration scheduling unit; the candidate base station screening unit takes the overloaded base station as the center, screens candidate base stations according to a preset radius, and generates a candidate list; the service priority classification unit classifies services into high, medium, and low priorities by evaluating their sensitivity to energy efficiency; The load migration scheduling unit is used to generate an optimal matching strategy between services and base stations, and issue migration instructions through the SDN controller to dynamically adjust the base station load.

Images