Multi-agent 5g base station intelligent energy-saving method and device based on size model fusion

Abstract

This application discloses a multi-agent intelligent energy-saving method and device for 5G base stations based on large-scale model fusion, relating to the field of energy saving in wireless communication networks. This application utilizes a technology that integrates large-scale language data models and generative AI models to automatically identify network deployment scenarios and configure initial energy-saving strategies corresponding to those scenarios. Based on the initial energy-saving strategies for each cell or region, energy-saving cells are selected, and their corresponding energy-saving strategies and energy-saving interaction instructions are obtained. The network equipment corresponding to these energy-saving cells is then controlled to enter an energy-saving state. Furthermore, for cells entering the energy-saving state, the corresponding energy-saving strategies are iteratively optimized online based on changes in energy-saving effects and load performance indicators, seeking a balance between energy-saving threshold parameters and network performance to maximize energy savings.

Description

Technical Field

[0001] This application relates to the field of energy-saving technology for wireless communication networks, and in particular to a smart energy-saving method and device for multi-agent 5G base stations based on large-scale model fusion. Background Technology

[0002] As communication networks continue to expand, the energy consumption of network equipment is also increasing, making network energy conservation a crucial issue in the communications field. The key challenge is how to utilize energy-saving technologies to minimize the power consumption of communication networks, ensuring high network performance while reducing power consumption and improving network energy efficiency.

[0003] The main means to improve network energy efficiency include: improving air interface wireless technology, optimizing equipment hardware energy efficiency, and better matching load with network capacity. Among these, better matching load with network capacity has become the main technical approach. However, the threshold parameters for entering and exiting energy saving in current network energy-saving solutions are defined by expert experience, which cannot maximize energy saving. Summary of the Invention

[0004] In view of the above problems, this application provides a multi-agent 5G base station intelligent energy-saving method and device based on big-small model fusion to solve at least some of the above technical problems. The specific solution is as follows:

[0005] The first aspect of this application provides a multi-agent intelligent energy-saving method for 5G base stations based on size model fusion, including:

[0006] Identify the network deployment scenario information corresponding to the current wireless network, and configure an initial energy-saving strategy that matches the network deployment scenario information;

[0007] Based on the initial energy-saving strategy, energy-saving areas in the current wireless network are selected, and the current energy-saving strategy corresponding to the energy-saving area is determined based on the existing network data corresponding to the energy-saving area.

[0008] Based on the current energy-saving strategy, an energy-saving interaction command is generated and sent to the wireless network device corresponding to the energy-saving area. The energy-saving interaction command is used to control the energy-saving status of the wireless network device.

[0009] Based on the changing trends of the energy-saving effect and key network performance indicators corresponding to the energy-saving zone, the current energy-saving strategy is iteratively optimized online.

[0010] In one possible implementation, the network deployment scenario information includes network coverage information and network configuration information, and the initial energy-saving strategy includes energy-saving type, energy-saving period and energy-saving threshold parameters;

[0011] The initial energy-saving strategy, which matches the configuration with the network deployment scenario information, includes:

[0012] Configure the power saving type of the current wireless network based on the network configuration information of the current wireless network. The power saving type includes symbol shutdown, channel shutdown, carrier shutdown, or deep sleep of radio frequency module.

[0013] Based on the historical traffic load performance indicators and network coverage information of the current wireless network, configure the energy-saving time period and energy-saving threshold parameters for each area of ​​the current wireless network.

[0014] In one possible implementation, configuring the energy-saving time period and energy-saving threshold parameters for each area of ​​the current wireless network by combining the historical traffic load performance indicators and network coverage information corresponding to the current wireless network includes:

[0015] For any area of ​​the current wireless network, the off-peak traffic load level is determined based on the historical traffic load of that area;

[0016] The energy-saving threshold parameter for any region is determined based on the off-peak call load level of any region.

[0017] Based on the traffic load forecast results for any given region and the energy-saving threshold parameters, determine the energy-saving time period for any given region.

[0018] In one possible implementation, the initial energy-saving strategy includes an energy-saving threshold parameter, which includes a traffic load threshold value for entering an energy-saving state or an energy-saving time period; wherein, selecting energy-saving areas from the current wireless network based on the initial energy-saving strategy includes:

[0019] Areas with predicted call load values ​​lower than the corresponding call load threshold for energy saving are designated as energy-saving zones.

[0020] or,

[0021] The area that is currently in the corresponding energy-saving time period is selected as the energy-saving area.

[0022] In one possible implementation, determining the current energy-saving strategy corresponding to the energy-saving region based on the existing network data corresponding to the energy-saving region includes:

[0023] Acquire the internal and external network data corresponding to the energy-saving area. The internal network data includes engineering parameter information and terminal measurement report data. The external network data includes external system data that affects traffic load performance indicators.

[0024] Based on the internal and external network data, the predicted values ​​of the traffic load performance indicators corresponding to the energy-saving area are obtained using the load forecasting model.

[0025] Based on the predicted values ​​of the call load performance indicators and the energy-saving threshold parameters corresponding to the energy-saving area, the optimal energy-saving period corresponding to the energy-saving area is determined;

[0026] The current energy-saving strategy is obtained based on the energy-saving type and optimal energy-saving period corresponding to the energy-saving area.

[0027] In one possible implementation, the step of iteratively optimizing the current energy-saving strategy online based on the changing trends of the energy-saving effect and key network performance indicators corresponding to the energy-saving region includes:

[0028] Increase the target adjustment step size by the current energy-saving threshold parameter corresponding to the energy-saving zone;

[0029] Energy-saving control is performed on the energy-saving region based on the energy-saving threshold parameter after adjusting the step size by increasing the target, and the corresponding key network performance indicators are obtained.

[0030] If the key network performance indicators do not decrease or decrease but do not exceed the allowable threshold, continue to execute the step of increasing the current energy-saving threshold parameter corresponding to the energy-saving zone by the target adjustment step size.

[0031] If the key network performance indicators drop and exceed the allowable threshold, the energy-saving threshold parameter is restored to the value before the target adjustment step size was increased, and the optimal energy-saving threshold parameter is obtained.

[0032] In one possible implementation, the method further includes:

[0033] If the energy-saving zone is not in energy-saving mode, the network performance KPIs of the energy-saving zone and the same coverage area are evaluated. If the evaluation results of the network performance KPIs of the energy-saving zone and the same coverage area are poor, the energy-saving zone is prevented from entering the energy-saving mode; if the evaluation results of the network performance KPIs of the energy-saving zone and the same coverage area are excellent or there is no result, the energy-saving zone is allowed to enter the energy-saving mode.

[0034] When the energy-saving zone has entered the energy-saving state, the network performance KPI of the same coverage area of ​​the energy-saving zone is evaluated. If the evaluation result of the network performance KPI of the same coverage area is poor, the energy-saving zone is activated; if the evaluation result of the network performance KPI of the same coverage area is excellent or there is no result, the energy-saving zone remains in the energy-saving state.

[0035] The same coverage area is an area whose coverage is the same as or overlaps with the energy-saving area by more than a preset threshold.

[0036] In one possible implementation, the method further includes:

[0037] Detect whether there is a conflict in the current energy-saving strategy. If there is a conflict, modify the current energy-saving strategy. If there is no conflict, generate an energy-saving interaction command based on the current energy-saving strategy.

[0038] Based on the execution result of the energy-saving interaction command, the system detects whether there are any omissions or conflicts in the current energy-saving strategy. If there are any omissions or conflicts, the system reports the conflict detection result of the current energy-saving strategy to the strategy management platform.

[0039] In one possible implementation, detecting whether the current energy-saving strategy conflicts includes:

[0040] Detect whether there is a semantic conflict in the current energy-saving strategy. If so, determine that the energy-saving area does not need to save energy. Otherwise, execute the step of generating energy-saving interaction instructions based on the current energy-saving strategy.

[0041] Check if the energy-saving interaction command is correct; if it is incorrect, modify the current energy-saving strategy.

[0042] If the energy-saving interaction command is correct, then check whether there is an implicit conflict in the current energy-saving strategy. If there is an implicit conflict, then modify the current energy-saving strategy.

[0043] If the current energy-saving strategy does not have implicit conflicts, then check if the current energy-saving strategy has explicit conflicts. If there are no explicit conflicts, then execute the step of generating energy-saving interaction instructions based on the current energy-saving strategy. If there are explicit conflicts, then modify the current energy-saving strategy.

[0044] A second aspect of this application provides a multi-agent 5G base station intelligent energy-saving device based on size model fusion, comprising:

[0045] The initial policy configuration module is used to identify the network deployment scenario information corresponding to the current wireless network and configure an initial energy-saving policy that matches the network deployment scenario information.

[0046] The current strategy determination module is used to filter energy-saving areas in the current wireless network based on the initial energy-saving strategy, and determine the current energy-saving strategy corresponding to the energy-saving area based on the existing network data corresponding to the energy-saving area.

[0047] The energy-saving control module is used to generate energy-saving interaction instructions based on the current energy-saving strategy and send them to the wireless network device corresponding to the energy-saving area. The energy-saving interaction instructions are used to control the energy-saving status of the wireless network device.

[0048] The energy-saving strategy update module is used to iteratively optimize the current energy-saving strategy online based on the changing trends of the energy-saving effect and key network performance indicators corresponding to the energy-saving zone.

[0049] A third aspect of this application provides an electronic device, comprising at least one processor and a memory connected to the processor, wherein:

[0050] The memory is used to store computer programs;

[0051] The processor is used to execute the computer program so that the electronic device can implement the above-described first aspect or any implementation thereof, the multi-agent 5G base station intelligent energy-saving method based on size model fusion.

[0052] The fourth aspect of this application provides a computer storage medium carrying one or more computer programs, which, when executed by an electronic device, enable the electronic device to implement the intelligent energy-saving method for multi-agent 5G base stations based on size model fusion, as described in the first aspect or any implementation thereof.

[0053] By employing the aforementioned technical solutions, the multi-agent 5G base station intelligent energy-saving method based on large-scale model fusion provided in this application can automatically identify network deployment scenarios and configure initial energy-saving strategies corresponding to those scenarios through the fusion of large-scale language data models and generative AI models. Based on the initial energy-saving strategies corresponding to each cell or region, energy-saving cells are selected, and the corresponding energy-saving strategies and energy-saving interaction instructions are obtained. The network equipment corresponding to these energy-saving cells is then controlled to enter an energy-saving state. Furthermore, for cells entering the energy-saving state, the corresponding energy-saving strategies are iteratively optimized online based on changes in energy-saving effects and load performance indicators, seeking a balance between energy-saving threshold parameters and network performance to maximize energy savings. Attached Figure Description

[0054] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0055] Figure 1 This application provides a schematic diagram of the architecture of a wireless energy-saving system.

[0056] Figure 2 A flowchart of a multi-agent 5G base station intelligent energy-saving method based on size model fusion provided in this application;

[0057] Figure 3A flowchart illustrating another intelligent energy-saving method for multi-agent 5G base stations based on size model fusion provided in this application embodiment;

[0058] Figure 4 A schematic diagram illustrating a knowledge base creation and usage process provided in an embodiment of this application;

[0059] Figure 5 A schematic diagram illustrating an online iterative optimization of energy-saving threshold parameters, provided as an embodiment of this application;

[0060] Figure 6 A flowchart illustrating an energy-saving strategy conflict management process provided in this application embodiment;

[0061] Figure 7 A schematic diagram of the structure of a multi-agent 5G base station intelligent energy-saving device based on size model fusion provided in this application embodiment;

[0062] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0063] The embodiments of this application are described below with reference to the accompanying drawings. The terminology used in the implementation section of this application is for explaining specific embodiments only and is not intended to limit the scope of this application.

[0064] The embodiments of this application will now be described with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are equally applicable to similar technical problems.

[0065] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms are interchangeable where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of elements is not necessarily limited to those elements, but may include other elements not explicitly listed or inherent to those processes, methods, products, or apparatuses.

[0066] First, let's introduce the technical terms used in this application:

[0067] 5G NR (New Radio) is a core component of fifth-generation mobile communication technology, providing wireless access with higher speeds, lower latency, greater connection density, and higher energy efficiency.

[0068] OFDM (Orthogonal Frequency Division Multiplexing) is a multi-carrier transmission technology that divides a channel into several orthogonal sub-channels, converts high-speed data signals into parallel low-speed sub-data streams, and modulates them onto each sub-channel for transmission.

[0069] RRC (Radio Resource Control) is the message configuration and control center of the access layer of the entire wireless communication protocol stack.

[0070] A PRB (Physical Resource Block) is the core resource unit in a wireless communication system and the smallest resource unit allocated to a user. A PRB consists of contiguous time-frequency resources, including a specific number of subcarriers and time slots. PRB utilization rate = number of PRBs already allocated to users / number of available PRBs. A higher PRB utilization rate indicates that physical resources are being used more effectively, meaning the system can better meet user needs.

[0071] An AAU (Active Antenna Unit) is a device that integrates radio frequency and digital processing functions. Its function is to amplify and process wireless signals, and then transmit and receive wireless signals through an antenna.

[0072] A BBU (Baseband Unit) is a component in a wireless base station system. It is a device that integrates digital signal processing and control functions. The BBU's function is to process the digital signals received from the AAU and control the operation of the wireless base station system.

[0073] An RRU (Remote Radio Unit) is a component of a wireless base station system; it is a device that integrates radio frequency (RF) functionality. Its function is to convert digital signals into RF signals and transmit and receive these signals via an antenna.

[0074] In the process of researching the solution proposed in this application, the inventors discovered that existing intelligent energy-saving methods for multi-agent 5G base stations based on large-scale model fusion rely on expert experience to determine key indicator thresholds for entering and exiting energy-saving states. When the cell-level load indicators are detected to meet the thresholds for the energy-saving strategy type, the network element management system issues the corresponding entry / exit instructions for the energy-saving strategy to the specific energy-saving network element to achieve wireless network energy saving. However, once the key indicator thresholds are determined, they cannot be adaptively adjusted. Furthermore, the set key indicator thresholds are relatively conservative, failing to maximize energy saving.

[0075] To address the aforementioned issues, this application provides a multi-agent intelligent energy-saving method for 5G base stations based on the fusion of large and small models. This method continuously iterates and optimizes energy-saving strategies online based on big data analysis of the trends of energy-saving effects and key network performance indicators (KPIs) in a full-scenario traffic model, seeking a balance between key indicator thresholds and network performance to maximize energy savings.

[0076] Please see Figure 1 This illustration shows a schematic diagram of an energy-saving system architecture for a wireless communication network according to an embodiment of this application. This system is an energy-saving system architecture based on an Open Radio Access Network (O-RAN) architecture, such as... Figure 1 As shown, the system mainly uses the service management orchestration (SMO) framework to interact with other subsystems in the wireless communication system. These other subsystems may include the network functions of O-RAN, the O-RAN operating platform O-Cloud, the core network, and external systems.

[0077] The service-based, open SMO framework differs significantly from traditional operator network management subsystems.

[0078] The SMO framework provides various services for network operation and management of network devices, rather than a comprehensive, integrated network management system. Specific network operation and management tasks are performed by these various services.

[0079] The SMO framework is an open operations management platform that can integrate network equipment products from any manufacturer that conforms to the O-RAN specification, and can apply network operations management software from any manufacturer that conforms to the O-RAN specification.

[0080] In some embodiments of this application, the SMO framework includes the following functions: cloud infrastructure OAM, RAN OAM, and non-real-time RAN intelligent controller (Non-RTRIC).

[0081] The Cloud Infrastructure Operations and Maintenance (OAM) is responsible for the operation, maintenance, and management of cloud infrastructure. The Radio Access Network (RAN-OAM) is responsible for the operation, maintenance, and management of radio access network elements.

[0082] Non-RTRIC is responsible for intelligent energy-saving control of network devices. Specifically, the Non-RTRIC framework deploys Energy Saving RAN Applications (ESrApp), energy-saving models, and energy-saving model management. Energy-saving model training extends beyond the Non-RTRIC framework; that is, energy-saving model training is deployed within the SMO framework, and the energy-saving model is managed within the Non-RTRIC framework. The energy-saving model supports the output of cell-level optimal energy-saving periods, optimal energy-saving threshold parameters, and energy-saving interaction commands related to energy saving.

[0083] Among them, ESrApp runs in the non-real-time RAN intelligent controller (Non-RTRIC) and is responsible for tasks such as network-level energy-saving strategy optimization and long-term data analysis (time span from seconds to hours).

[0084] The energy-saving model in this application is an intelligent energy-saving model that integrates the Large Language Model (LLM) and the Artificial Intelligence (AI) model.

[0085] In an O-RAN wireless energy-saving system, the various subsystems communicate with each other through corresponding interfaces, which may include the following interfaces:

[0086] O1 Interface: A new interface between the SMO framework and the internal network elements of O-RAN, used by SMO to intelligently manage and operate the logical network elements within O-RAN.

[0087] O2 Interface: This is the interface between the SMO and the O-RAN operating platform (O-Cloud), used by the SMO to intelligently manage and operate the various O-RAN network service nodes running on the O-Cloud.

[0088] The following is about Figure 1 A brief introduction to the overall working process of the energy-saving system shown:

[0089] Based on load forecasting results (which are obtained by a load forecasting model predicting the traffic load of a cell or region over a future period, where traffic load refers to the amount of communication services carried by the communication system per unit time, typically used to measure the efficiency of communication network resource utilization), ESrApp filters out energy-saving cells and initiates energy-saving related data requests to the database. The database responds to these requests by pushing current network data (including data from the internal and external networks for the current time period) to the AI / ML model. This allows the AI / ML model to identify the network deployment scenario based on the data pushed by the database and customize an energy-saving strategy that matches that scenario. Specifically, it determines the corresponding optimal energy-saving threshold parameters (i.e., the load threshold for entering and exiting energy-saving states) and the optimal energy-saving period, and generates energy-saving interaction instructions, which are then provided to ESrApp. ESrApp then uses the interface between RAN OAM and O-RAN (i.e., the O1 interface in the diagram) to interact with O-RAN via the AI / ML model, such as sending energy-saving strategies and instructions to O-RAN and receiving energy-saving instruction response information returned by O-RAN. Meanwhile, AI / ML and ESRApp are used to conduct real-time rapid assessments of communities that have entered energy-saving status to ensure continuous iteration and updates of energy-saving strategies.

[0090] As can be seen, the energy-saving system for wireless communication networks provided in this application deploys energy-saving applications on wireless network device nodes. The application deployment does not require a large amount of server hardware and network resources, and it does not need to interface with standard interfaces from different manufacturers, thus significantly reducing the development and operation costs of energy-saving devices.

[0091] Please see Figure 2 The diagram illustrates a flowchart of a multi-agent 5G base station intelligent energy-saving method based on size model fusion, provided in an embodiment of this application. This method is applied to... Figure 1 In the SMO of the system shown, the method may include the following steps:

[0092] S101 identifies network deployment scenario information for each cell or area in the current wireless network.

[0093] The energy-saving model can automatically identify the network deployment scenario information of the wireless network in the current community or area, such as network coverage, networking mode, and call load.

[0094] (1) Network coverage identification

[0095] Network coverage identification mainly relies on engineering parameter information and terminal measurement report (MR) data. Through big data analysis that integrates these two types of information, multi-frequency and multi-standard cell coverage identification is carried out to find the relationship between cells with the same coverage in the area (including most overlapping coverage).

[0096] Engineering parameter information is usually provided by the operator and is updated periodically with detailed data. It mainly includes information such as site name, latitude and longitude, antenna height and azimuth angle.

[0097] MR (Mean Access) is based on measurement results from wireless terminal devices (such as mobile phones), including information such as signal strength, signal-to-noise ratio, and modem parameters. MR data statistics are mainly based on cell identifiers, signal strength, and tracking areas (TA) carried in the terminal measurement information, and these are used to calculate the cell's coverage area and whether there are scenarios with weak uplink and downlink coverage.

[0098] (2) Network configuration identification

[0099] As networks continue to develop and evolve, various network topologies exist, including Standalone (SA) / Non-Standalone (NSA), co-construction and sharing, different hardware equipment models (including single-mode / hybrid-mode using the same hardware), and different network topologies (such as 4G / 5G shared BBU frames, 4G / 5G separate BBU frames, etc.).

[0100] The big data model in the energy-saving model distinguishes equipment types through base station and cell configuration identification, and obtains serving cell information and neighbor cell relationship information from the OMC backend data configuration table. This information is used to analyze how energy-saving types such as symbol shutdown, channel shutdown, carrier shutdown, and deep sleep of radio frequency modules are matched with the corresponding configuration information. OMC, or Operation and Maintenance Center, is a network management system specifically used to monitor and control the operation and performance of communication networks.

[0101] The serving cell information includes: resource data: standard, bandwidth, network element name, network element identifier, etc.; performance data: cell uplink PRB utilization, cell downlink PRB utilization, number of RRC connected users, etc.; alarm data: alarm name, alarm impact, alarm time, etc.; configuration data: power and other related configuration data.

[0102] Neighbor cell relationship information is a list of neighbor cells that have neighbor cell relationships with the serving cell, which may include resource data such as network type, bandwidth, network element name, and network element identifier.

[0103] The energy-saving model combines historical traffic load performance indicators (such as cell uplink PRB utilization, cell downlink PRB utilization, cell RRC connected users, and other traffic load-related performance indicators), network configuration analysis results, and network coverage identification results to identify the network deployment scenario of the current network equipment, including coverage scenario, networking mode, and network traffic load.

[0104] S102 configures initial energy-saving strategies that match the network deployment scenarios of each cell or region.

[0105] Initial energy-saving strategies may include energy-saving types (such as symbol shutdown, channel shutdown, carrier shutdown, or deep sleep of RF modules), energy-saving threshold parameters (traffic load performance index thresholds for entering and exiting energy-saving states), and energy-saving periods.

[0106] Among them, the energy-saving type can be configured according to the network configuration information of the current wireless network (such as serving cell information and neighbor cell relationship information).

[0107] Symbol shutdown involves disabling the transmission of one or more symbol cycles in the time dimension. A symbol is the basic time unit in the radio frame structure, such as the OFDM symbol in 5G NR. Taking a 5G NR low-frequency system as an example, a radio frame is 10ms long, each radio frame consists of 10 radio subframes, each radio subframe consists of 2 time slots, each radio subframe is 1ms long, and each time slot is 0.5ms long. Each time slot typically consists of 14 symbols. In actual communication, the base station is not always in a state of maximum traffic, so the symbols in a subframe are not always filled with valid information. The power amplifier (PA) and transceiver unit (TRX) switches are turned off during symbol cycles when no data transmission is predicted to reduce system power consumption, and the PA and TRX switches are turned on in advance during symbol cycles when data transmission is predicted to ensure that services are not affected.

[0108] Channel shutdown is the closure of a specific logical or physical channel, such as a control channel, a traffic channel, or a frequency band resource block.

[0109] Carrier shutdown involves turning off the entire carrier, i.e., a specific frequency bandwidth. For example, in carrier aggregation scenarios, secondary carriers are shut down to reduce power consumption when the load is insufficient.

[0110] Deep sleep mode for radio frequency modules refers to the ability of the base station radio frequency unit (AAU) to shut down the power supply to most active devices and enter a sleep state when the base station is in a prolonged idle state, thereby saving energy. This function is particularly suitable for 5G base stations because the RF unit of a 5G base station contains a large number of active devices and baseband processing units, resulting in a significant increase in its idle power consumption.

[0111] In an exemplary embodiment, automatically configuring an initial energy-saving strategy that matches the scenario through an AI algorithm model of the energy-saving model may include the following process:

[0112] A1, determine the off-peak call load level of a region or cell based on historical call load data of the region or cell;

[0113] A2, determine energy-saving threshold parameters based on the off-peak call load level of a region or community.

[0114] Energy-saving threshold parameters include thresholds for traffic load performance indicators when entering and exiting energy-saving states. For example, different energy-saving types correspond to different energy-saving threshold parameters. For instance, the energy-saving threshold parameter for symbol shutdown is higher than that for channel shutdown, and the threshold for channel shutdown is higher than that for carrier shutdown.

[0115] A3 determines the applicable energy-saving time period for each area or community based on the corresponding energy-saving threshold parameters and call load forecast results.

[0116] To enable the energy-saving strategy to adapt to traffic load, the AI ​​algorithm model in the energy-saving model introduces a load prediction kernel. It is based on traffic load-related performance indicators (such as uplink and downlink PRB utilization, RRC user count, etc.) and models different network deployment scenarios and time characteristics of weekdays / holidays. The time series prediction algorithm is used to complete the load prediction, drive the timely application of energy-saving periods and energy-saving threshold parameters, and optimize the energy-saving effect while ensuring network performance.

[0117] The energy-saving period can be determined based on the time period during which the predicted load meets the performance index threshold for energy-saving conditions. For example, if it is predicted that the traffic load-related performance index of a certain cell meets the threshold value for entering the energy-saving state during the time period t1~t2, then the energy-saving period for that cell is determined to be the time period t1~t2.

[0118] S103: Select energy-saving communities that meet the energy-saving conditions from each community or region, request the internal and external network data corresponding to the energy-saving communities to be provided to the energy-saving model, and obtain the current energy-saving strategy corresponding to the energy-saving communities output by the energy-saving model.

[0119] Current energy-saving strategies include optimal energy-saving time periods, optimal energy-saving threshold parameters, and energy-saving interaction commands.

[0120] In an exemplary embodiment, for any cell or region in the current wireless network, it is determined whether the current time period is included in the energy-saving time period (the energy-saving time period in the initial energy-saving strategy corresponding to the cell or region) of that cell or region. If so, the cell is determined to be an energy-saving cell.

[0121] In another exemplary embodiment, for any cell or area in the current wireless network, the actual measured value of the traffic load of that cell or area is obtained, and it is determined whether the measured value of the traffic load is lower than the energy-saving threshold parameter (e.g., traffic load threshold value) corresponding to that cell or area. If it is lower than or equal to, it is determined that the energy-saving condition is met, and access to the energy-saving state is allowed. If it is higher than, it is determined that the energy-saving condition is not met, and entry into the energy-saving state is not allowed.

[0122] If the current cell or region meets the corresponding energy-saving conditions, the database is requested to provide the energy-saving model with internal and external network data for that cell or region. Internal network data may include engineering parameter information and MR data, while external network data may include data from external systems, such as complaint data, large-scale population flow information, internet ticketing and major event information, and network security information. This allows the energy-saving model to output the energy-saving strategy for the current time period based on the internal and external network data corresponding to that cell or region, such as the optimal energy-saving period, optimal energy-saving threshold parameters, and energy-saving interaction commands. These energy-saving interaction commands are automatically generated by the energy-saving model according to the northbound command-line development requirements of the wireless network equipment corresponding to that cell or region.

[0123] S104 interacts with the wireless network equipment corresponding to the energy-saving community to exchange energy-saving commands.

[0124] See Figure 1 The ESrApp application can obtain the energy-saving strategies and energy-saving interaction commands output by the energy-saving model, and exchange energy-saving commands through the interface O1 between RAN OAM and O-RAN. That is, it controls the network devices corresponding to the energy-saving cell to enter the energy-saving state through energy-saving commands.

[0125] S105: For cells or regions that have entered the energy-saving state, the corresponding energy-saving strategy for the cell or region is iteratively optimized online based on the changing trends of energy-saving effect and traffic load performance indicators.

[0126] Energy-saving strategies can be dynamically optimized based on network performance KPIs over a period of time (e.g., one day). Specifically, the energy-saving duration can be dynamically optimized based on KPIs to achieve a balance between energy-saving duration and network performance. Furthermore, energy-saving threshold parameters (traffic load performance thresholds for entering and exiting energy-saving states) can be continuously optimized based on network performance KPIs.

[0127] Network performance KPIs mainly include key performance indicators related to network quality, such as KPIs related to call establishment, dropped calls, handover, and user experience, such as network connection rate, network drop rate, network handover success rate, and network traffic.

[0128] At the same time, it can also conduct real-time rapid assessments of communities or areas that have entered energy-saving status, and the results of the rapid assessments can be updated in real time to indicate whether they have entered or exited energy-saving status.

[0129] The intelligent energy-saving method for 5G base stations based on large-scale model fusion provided in this embodiment can automatically identify network deployment scenarios and configure initial energy-saving strategies corresponding to those scenarios by fusing large language data models and generative AI models. Based on the initial energy-saving strategies for each cell or region, energy-saving cells are selected, and their corresponding energy-saving strategies and interaction commands are obtained. The network equipment corresponding to these energy-saving cells is then controlled to enter an energy-saving state. Furthermore, for cells entering the energy-saving state, the corresponding energy-saving strategies are iteratively optimized online based on changes in energy-saving effects and load performance indicators, seeking a balance between energy-saving threshold parameters and network performance to maximize energy savings.

[0130] Please see Figure 3 This document illustrates a flowchart of another intelligent energy-saving method for multi-agent 5G base stations based on size model fusion, provided in an embodiment of this application. The method... Figure 2 The detailed process of energy-saving model training and energy-saving control has been added to the embodiment shown.

[0131] like Figure 3 As shown, the method may include the following steps:

[0132] S201, Building a knowledge base based on external data.

[0133] In one exemplary embodiment, the process of building a knowledge base may include the following steps:

[0134] (1) Collect external data

[0135] For example, external data may include documentation from wireless network manufacturers or operators, and data retrieved from external system API retrieval plugins.

[0136] Documentation from wireless network manufacturers or operators may include: documentation related to wireless network element models (mainly providing knowledge about wireless network hardware), technical requirements for the interface of the wireless network capability scheduling subsystem (mainly providing knowledge about wireless network resources, alarms, and performance), wireless MR documentation (mainly providing information on wireless network coverage at the cell level), and northbound command line development guide documentation (mainly providing operation and control commands for wireless network elements).

[0137] External system API retrieval plugin data may include: centralized complaint system (mainly providing information such as the location of the complaint and the associated wireless cell), Internet crowd flow analysis system (mainly providing information on large-scale crowd flow), Internet ticketing and major event information (mainly providing information on the attendance of people in entertainment, sports and other related projects), and network security information (information on the security scenario and detailed information on the base stations or cells included in the security).

[0138] (2) Data processing

[0139] Since the model has certain limitations on the size of the input data, after completing the above data collection, large documents need to be segmented into smaller text blocks. These text blocks should maintain semantic integrity as much as possible. For example, they can be segmented by chapter. The purpose of segmentation is to ensure that the size of each text block is suitable for the model to process, while minimizing the loss of contextual information.

[0140] Next, the segmented text blocks are converted into numerical vectors. This step can be done by calling a vector conversion server. Specifically, the text data of the text blocks can be sent to a server for vector conversion, and the server returns the vectors corresponding to the text data. These vectors can capture the semantic information of the text. Similar texts are close to each other in the vector space, meaning that semantically similar texts will also produce similar vector data.

[0141] (3) Result storage

[0142] It stores vectors, constructs and stores knowledge graphs based on text vectors, and stores the original documents.

[0143] The knowledge base built on a large language model can efficiently store a large number of energy-saving related text vectors, quickly retrieve the text vector most similar to a given vector, and support complex query operations.

[0144] S202 uses historical data and knowledge base from wireless networks to train the energy-saving model.

[0145] Historical data for wireless networks can include: performance data (such as cell uplink and downlink PRB utilization, RRC connected user count, etc.), resource data (such as network type, bandwidth, network element name, network element identifier, etc.), alarm data (such as alarm name, alarm impact, alarm time, etc.), parameter configuration data (such as power and other related configuration data), and MR data (such as signal strength, signal-to-noise ratio, etc.). Manual annotation of policy execution effect labels in historical data is required, such as energy-saving efficiency and network quality change indicators, to further construct the mapping relationship between the input (such as network status, MR data, external system data, etc.) and output (policy execution effect) of the energy-saving model.

[0146] The energy-saving model is built upon a Retrieval Augmented Generation (RAG) model. RAG is a model that combines retrieval and generation techniques, generating answers or content by referencing information from external knowledge bases, offering strong interpretability and customizability. RAG improves prediction quality and accuracy by retrieving relevant information from large-scale document collections and using this information to guide text generation. The retrieval function retrieves relevant information from external knowledge bases based on the user's query. Augmentation involves embedding the user's query and retrieved relevant knowledge into a pre-defined prompt template. Generation involves inputting the retrieval-enhanced prompt content into a large language model to generate the desired output.

[0147] The knowledge base, through structured storage of domain knowledge and practical experience, enables energy-saving models to overcome the limitations of the data-driven paradigm. Specifically, the knowledge base provides professional rules and experience to the energy-saving model, enhancing the interpretability of decisions. During the training process of the energy-saving model, the knowledge base can enhance the data quality and expand dynamic features; during the iterative optimization process of the energy-saving model, the knowledge base provides corresponding support for the model's continuous learning. Furthermore, in terms of ensuring energy-saving effects, the knowledge base provides corresponding support for the energy-saving model to implement policy conflict detection, anomaly recovery, and effect verification.

[0148] In addition, semantic tags need to be added to the text paragraphs in the knowledge base, such as power adjustment rules and traffic load prediction methods, in order to improve the prediction accuracy of RAG.

[0149] In an exemplary embodiment, the basic energy-saving model can be trained using an offline pre-training method. For example, the basic energy-saving model can be trained using historical network status data (such as data from the past three months) corresponding to the current base station node. Specifically, the historical network status data can be input into the initial model to obtain the corresponding energy-saving strategy and the corresponding strategy execution effect. The loss value between the predicted strategy execution effect and the actual execution effect is calculated based on the loss function. If the loss value is greater than a preset threshold, the model parameters in the basic model are adjusted according to the loss value. The above process is repeated until the loss value is less than or equal to the preset threshold to obtain the basic prediction model corresponding to each base station node.

[0150] Furthermore, contrastive learning is used to optimize the knowledge base vector representation. This process improves the accuracy of knowledge base retrieval by deeply integrating domain knowledge with representation learning. In the scenario of energy-saving strategy generation, it provides intelligent energy-saving technologies such as load forecasting energy-saving period optimization, energy-saving threshold parameter optimization, and rapid evaluation guarantee. Through intelligent orchestration of energy-saving time, energy-saving threshold parameters, and other dimensions, as well as the mutual coordination of multiple frequency layers, it fully explores the energy-saving capabilities of base stations and maximizes the duration of energy-saving effectiveness while ensuring user perception.

[0151] Contrastive learning is a self-supervised learning method that learns the vector representation of a text by comparing sample pairs. The core idea of ​​contrastive learning is that similar samples should be close to each other in the representation space, while dissimilar samples should be far apart. The process of contrastive learning optimizing vector representations can be summarized as follows:

[0152] (1) Constructing and generating samples

[0153] The samples involved in this embodiment include three types: anchor, positive, and negative. Anchor represents the original text paragraph (such as "AAU deep sleep triggering condition"); positive represents a manually rewritten synonymous description (such as "RF unit sleep activation rule"); and negative represents similar but unrelated text (such as "base station heat dissipation system start-up and shutdown logic"). Further, a loss function is constructed.

[0154] (2) Constructing the loss function

[0155] The loss function is used to calculate the loss between the historical vector representation and the current vector representation of the same sample. The historical vector representation refers to the vector representation of the sample obtained through the historical vector transformation model, while the current vector representation is the vector representation of the sample obtained based on the updated vector transformation model. The model parameters of the vector representation model are then updated based on the loss value.

[0156] (3) HyDE Enhanced Representation

[0157] HyDE (Hypothetical Document Embeddings) is a technique that enhances retrieval performance by generating hypothetical documents. The core idea of ​​HyDE is to improve retrieval accuracy by generating a hypothetical document that more closely approximates the embedding space of the target document than the original query. Specifically, HyDE first generates a hypothetical document that better captures the intent of the query. Then, it vectorizes this hypothetical document to generate an embedding representation. This vectorized hypothetical document is then used to retrieve similar documents from the database, finding the real documents that are closest to the hypothetical document.

[0158] Furthermore, each base station node updates its local model parameters through Personalized Federation Instruction Fine-tuning (PFIT), and a dynamic weighted average is used when a global model is aggregated to adapt to different network load scenarios.

[0159] (1) Model aggregation under the federated learning framework: By designing a dynamic weighted average algorithm model, the weight calculation is based on the number of network users, network traffic, resource utilization entropy value and user service perception quality score of the base station node. For heterogeneous base station equipment, feature space mapping technology is used to unify the uniqueness of the model input. Knowledge distillation is used to compress the differences in model parameters and retain the core energy-saving decision logic to achieve alignment of the parameter surface.

[0160] (2) Network energy-saving adaptation solution

[0161] A real-time adaptation engine is constructed using a two-layer decision-making mechanism and a load prediction module. The first layer is a fast energy-saving response layer based on a rule engine, and the second layer is a model inference-based fine-grained strategy generation. The load prediction module uses an LSTM+Attention model to predict the load for the next hour. The real-time adaptation engine, combined with a scene feature library, outputs a network energy-saving adaptation solution.

[0162] LSTM (Long Short-Term Memory) networks, whose core idea is to introduce a long-term memory unit (CellState) to store long-term information and selectively update or delete information through a set of carefully designed gating mechanisms (including input gates, forget gates, and output gates). These gating mechanisms can control the inflow and outflow of information into the cell state, thereby achieving effective learning of long-term dependencies. Attention models are mechanisms that dynamically allocate weights, enabling neural networks to focus on key information when processing sequential data.

[0163] like Figure 1 As shown, the trained energy-saving model is deployed in the Non-RTRIC framework of the SMO framework. In addition, the Non-RTRIC framework also deploys an energy-saving model management module for managing the energy-saving model, such as updating the energy-saving strategy of the energy-saving model.

[0164] Please see Figure 4 This diagram illustrates a process for establishing and using a knowledge base according to an embodiment of this application. Collected external data is preprocessed to obtain document objects. These document objects are then segmented into text paragraphs (i.e., text blocks). The text paragraphs are further converted into corresponding vectors, and corresponding indexes are built for the text vectors to generate a professional knowledge base.

[0165] Upon receiving a wireless energy-saving question, the system converts it into a corresponding energy-saving question vector. It then retrieves the top k relevant text vectors matching the question vector from the knowledge base. These text vectors are combined to form a context. Finally, the system calls the large language model within the energy-saving model to return results related to the question. These results may include operations without user interaction or operations with user interaction.

[0166] S203, ESrApp filters energy-saving communities and requests the database to push the required internal and external live network data to the energy-saving model.

[0167] After ESrApp selects the community that needs energy saving, it notifies the database to push the community's internal and external network data to the energy saving model, such as MR data, data from external systems, such as complaint data, large-scale population flow information, internet ticketing and major event information, and network security information.

[0168] S204, ESrApp obtains the energy-saving strategies of the energy-saving community output by the energy-saving model.

[0169] like Figure 1 As shown, ESrApp subscribes to the energy-saving strategy of the energy-saving model through the R1 interface. The energy-saving strategy may include the optimal energy-saving period, the optimal energy-saving threshold parameter, and the energy-saving interaction command.

[0170] (1) Optimal energy-saving period

[0171] In an exemplary embodiment, the optimal energy-saving period can be determined based on the predicted load corresponding to each predicted time granularity (e.g., 15 minutes) and the energy-saving threshold parameter corresponding to the energy-saving cell. Taking carrier shutdown as an example, the process of determining the optimal time period for carrier shutdown of the energy-saving cell is as follows:

[0172] The predicted average number of users per carrier RRC is less than or equal to the threshold for the number of carrier-level UEs that are turned off.

[0173] The predicted carrier uplink / downlink PRB utilization rate is ≤ the carrier shutdown uplink / downlink PRB utilization rate load threshold;

[0174] The predicted uplink / downlink PRB utilization rate of the energy-saving cell group is ≤ the wake-up threshold of the uplink / downlink PRB utilization rate of all basic coverage cells; where, the energy-saving cell group includes the currently selected energy-saving cells and co-coverage cells (cells with the same or largely overlapping coverage areas as the energy-saving cells). All basic coverage cells are the energy-saving cells and all co-coverage cells of the energy-saving cells.

[0175] If all the above indicators for the same time granularity simultaneously meet the corresponding conditions, then the predicted load for that granularity is considered to meet the carrier shutdown condition. At least two consecutive time granularities or higher are then filtered from the original energy-saving time periods as the final carrier shutdown energy-saving time. Simultaneously, at least four consecutive time granularities or higher are selected as the final deep sleep energy-saving time.

[0176] When deep sleep mode cannot meet the operator's pursuit of ultimate energy saving, in order to achieve even greater energy saving, the AAU / RRU can be directly powered down when the traffic load is low. This type of energy saving is called automatic start-stop energy saving. Automatic start-stop of AAU / RRU will power down the entire AAU / RRU (except for the power software module). The energy saving entry conditions are the same as the energy saving entry conditions for carrier shutdown / deep sleep in the 5G NR system. Energy saving wake-up only supports timed wake-up.

[0177] Automatic start-stop needs to consider the converted traffic load performance indicators. The conversion method is as follows: when the energy-saving threshold parameter corresponding to the carrier shutdown is lower than the preset value, the original energy-saving threshold parameter is kept unchanged. When the energy-saving threshold parameter corresponding to the carrier shutdown is higher than the preset threshold, the original energy-saving threshold parameter is reduced by a certain value as the automatic start-stop threshold. The time period with at least 9 consecutive event granularities where the traffic load is lower than the automatic start-stop threshold is the final energy-saving time period for automatic start-stop.

[0178] As for the channel shutdown method, only the energy-saving threshold parameters of the primary serving cell are considered, without needing to pay attention to the load of any neighboring cells. At least two consecutive time periods with different time granularities are filtered out, and the intersection of all carrier channel shutdown energy-saving periods with the same antenna group is taken as the final channel shutdown energy-saving period.

[0179] (2) Optimal energy-saving threshold parameters

[0180] For energy saving, the higher the threshold value for entering the shutdown state, the better the energy saving effect, that is, the easier it is to enter the energy saving state. However, in order to take into account the differences of various scenarios, traditional energy saving solutions set relatively conservative threshold values, that is, the threshold values ​​are low, which means it is more difficult to enter the energy saving state, resulting in poor energy saving effect.

[0181] In an exemplary embodiment, a rollback-based self-optimization strategy for traffic load performance indicators is adopted to obtain the optimal energy-saving threshold parameter. That is, based on the full-scenario traffic load prediction model, the energy-saving effect and network performance KPI change trend are continuously optimized online iteratively. Specifically, it may include the following process:

[0182] For each cell, relevant performance index data is extracted periodically (e.g., once a day). Clustering algorithms are used to find the optimal adjustment step size for different threshold parameters. The corresponding energy-saving threshold parameters are adjusted according to the optimal adjustment step size. After the energy-saving threshold parameters are optimized and refreshed, the network performance KPIs of the wireless network are monitored. Within the allowable fluctuation range, the energy-saving threshold parameters are continuously iterated to finally reach the best balance point between energy saving and network performance, i.e., the optimal energy-saving threshold parameters.

[0183] Please see Figure 5 After iteratively optimizing the threshold parameters, it is determined whether the network performance KPI has declined. If so, it is further determined whether the network performance KPI has exceeded the allowable threshold. If it has not exceeded the allowable threshold, the threshold parameters are iteratively optimized again. If it has exceeded the allowable threshold, the threshold parameters are rolled back to the values ​​before the iterative optimization, and the network performance KPI is iteratively evaluated again. If the network performance KPI has not declined, the threshold parameters are iteratively optimized again until the optimal balance between energy saving and performance is reached.

[0184] Furthermore, the aforementioned predicted load is obtained through the load prediction algorithm in the energy-saving model.

[0185] In an exemplary embodiment, when modeling the load forecasting algorithm, positive-effect cells, negative-effect cells, and no-effect cells are distinguished based on historical load data. The subsequence splitting forecasting method within the same period is adopted, and the impact of holiday factors on forecasting indicators is combined with the second-order exponential smoothing forecasting algorithm to obtain a forecasting model with optimal computational performance and the best optimization effect.

[0186] Cell scenario classification: Positive effect cell refers to a cell where the performance index value increases at the beginning of the holiday and decreases at the end of the holiday; no effect cell refers to a cell where the performance index value does not change significantly during the holidays; negative effect cell refers to a cell where the performance index value decreases at the beginning of the holiday and increases at the end of the holiday.

[0187] By comprehensively modeling holiday factors, load development trends, and periodicity, the accuracy of cell load forecasting can be significantly improved. Holiday effects and periodic changes are crucial features; incorporating trend modeling helps capture long-term trends, and combining this with machine learning or deep learning methods enables more comprehensive cell load forecasting.

[0188] Meanwhile, for some communities with strong load surge characteristics, probability leads to poor prediction results. In this case, we consider screening energy-saving communities based on load stability and load volatility, and smoothing the data.

[0189] Based on performance indicators related to load, the cell scenarios are divided. Except for scenarios where load indicators surge or plummet, the accuracy of cell-level traffic load prediction is significantly improved, reaching over 94.5%.

[0190] S205, ESrApp interacts with energy-saving commands through the O1 interface between RAN OAM and O-RAN.

[0191] S206, ESRApp provides real-time rapid assessment of energy-saving communities and updates the energy-saving status based on the assessment results.

[0192] In one exemplary embodiment, real-time rapid evaluation to ensure updated energy-saving status may include the following situations:

[0193] Before entering the energy-saving phase, a real-time network performance KPI quick evaluation is triggered for energy-saving cells and co-coverage cells (cells whose coverage areas are the same as or largely overlap with the energy-saving cell), using network performance KPI data from a historical time period (e.g., the previous 2 hours) for evaluation. If the network performance KPI evaluation result of the energy-saving cell or co-coverage cell is poor, the energy-saving cell is prevented from entering the energy-saving state; if the evaluation results of the energy-saving cell and co-coverage cell are excellent or there is no result, the energy-saving cell is allowed to enter the energy-saving state.

[0194] During the energy-saving preparation phase, a rapid evaluation of network performance KPIs for energy-saving cells and co-coverage cells is triggered, using network performance KPI data from the energy-saving preparation phase for assessment. If the evaluation result for an energy-saving cell or co-coverage cell is poor, the energy-saving cell is prevented from entering energy-saving mode; if the evaluation results for both the energy-saving cell and the co-coverage cell are excellent or there is no result, the energy-saving cell is allowed to enter energy-saving mode.

[0195] Once the energy-saving phase has begun, a periodic rapid assessment of the same-coverage cells is triggered. The same-coverage cells undergo network performance KPI evaluation at a specified interval (e.g., 1 minute). If the network performance KPI evaluation result of the same-coverage cells is poor, the energy-saving cells are activated; if the evaluation result of the same-coverage cells is excellent or there is no result, the energy-saving cells are not activated.

[0196] The multi-agent 5G base station intelligent energy-saving method based on big-and-small model fusion provided in this embodiment utilizes big-and-small language model capabilities to upload relevant documents such as model maps of wireless network devices to a knowledge base. The big-and-small language model autonomously learns the energy-saving strategies supported by the model and automatically generates corresponding energy-saving entry and exit commands, greatly reducing the development and maintenance costs of energy-saving products and avoiding the problem of energy-saving accidents affecting user experience due to untimely command updates caused by wireless device iteration and upgrades. Moreover, the energy-saving model based on big-and-small language models can effectively and promptly adjust energy-saving strategies in response to sudden abnormal events that affect user experience, such as complaints and sudden high-load events, achieving a balance between energy saving effect and wireless perception.

[0197] Furthermore, considering that mobile communication networks are highly coupled and complex systems, users of energy-saving systems may have limited knowledge of the system's internal details, and energy-saving strategy requirements may originate from different network departments; therefore, conflicts between energy-saving strategies are inevitable. This application also provides a cognitive-enhanced energy-saving strategy conflict management method, which manages energy-saving strategies by identifying semantic conflicts, explicit conflicts, implicit conflicts, and omitted conflicts using a large language model. Figure 6 As shown, this management method is in Figure 4 In addition to the above, the following steps are also included:

[0198] S301, check if there is a semantic conflict in the energy-saving strategy; if not, execute S302; if yes, determine that the community does not need energy saving.

[0199] If multiple energy-saving strategies affect overlapping objects and their energy-saving intentions are mutually exclusive, repetitive, or contained in each other, a semantic conflict is identified. Semantic conflicts can be detected using historical energy-saving strategies and clearly defined rules based on expert experience. When a semantic conflict is detected, the system will not execute the energy-saving strategy and will report the failure of energy-saving strategy creation to the initiator.

[0200] For example, if multiple sites need to enable energy-saving functions in combination and the energy-saving threshold parameters need to be automatically optimized, and there is a mutual exclusion between energy-saving duration and performance indicators, a semantic conflict is determined.

[0201] S302, determine if the energy-saving strategy is correct; if correct, execute S303; if incorrect, modify the energy-saving strategy.

[0202] S303, check if there is an implicit conflict in the energy-saving strategy; if not, execute S304; if so, modify the energy-saving strategy.

[0203] Implicit conflicts in energy-saving strategies refer to situations where multiple energy-saving strategies require non-exclusive energy-saving parameter values, but these values ​​are unlikely to be simultaneously achieved in a specific scenario. Implicit conflicts in energy-saving strategies can be identified using machine learning algorithms, or by simultaneously issuing the intents of these energy-saving strategies in a digital twin target network using a digital twin platform. If an implicit conflict is detected, the system will not execute the energy-saving strategy and will report the failure of intent creation / activation / modification to the initiator of the energy-saving strategy.

[0204] For example, multiple sites need to have their energy-saving functions enabled in combination and the energy-saving threshold parameters need to be automatically optimized, but the same coverage cell list includes other scenarios such as: important road scenarios, key protection scenarios (important meetings), etc.

[0205] S304, check if there is a display conflict in the energy-saving strategy; if not, proceed to S305; if so, modify the energy-saving strategy.

[0206] An explicit conflict in energy-saving strategies refers to a situation where multiple energy-saving strategy objects and their affected parameters overlap, and the required values ​​for these overlapping parameters are mutually exclusive. The key to detecting explicit conflicts in energy-saving strategies lies in the system's ability to accurately determine the parameter values ​​required for each energy-saving strategy to be implemented. Upon detecting an explicit conflict, the system will not execute the energy-saving strategy and will report a failure to the strategy creator (creation / activation / modification).

[0207] For example, when multiple sites need to enable energy-saving functions in combination and the energy-saving threshold parameters need to be automatically optimized, the influence of neighboring cells needs to be considered. If there are cells that need to determine the carrier load of neighboring cells, and the same cell is a neighboring cell of different primary serving cells, primary serving cell A determines that energy saving is possible and cell B determines that energy saving is not possible, then there is a mutual exclusion of parameter requirements.

[0208] S305 generates energy-saving interactive instructions.

[0209] S306 executes the energy-saving interactive command.

[0210] The energy-saving interactive command is sent to the network device corresponding to the energy-saving community for execution.

[0211] S307: Based on the execution result of the energy-saving interaction command, detect whether there are any omissions or conflicts in the energy-saving strategy; if so, request manual assistance; otherwise, continue to detect whether there are any omissions or conflicts.

[0212] Missing conflicts refer to conflicts between multiple executed energy-saving strategies discovered by ESRApp through real-time monitoring. Missing conflicts are mainly caused by the immaturity of semantic, explicit, and implicit conflict technologies, or by unforeseen events that make the energy-saving strategies incomplete. After missing conflicts are discovered, they can be resolved by escalating to energy-saving strategy management through collaboration between large and small models or by manual assistance. Energy-saving strategies can be updated and adjusted in real time, or conflicts can be eliminated by prioritizing a certain energy-saving intention.

[0213] For example, multiple sites need to activate energy-saving functions in combination and require automatic optimization of energy-saving threshold parameters. When a sudden large-scale event occurs around a community that has already entered the energy-saving state, such as a temporary gathering of live streaming users, etc.

[0214] Corresponding to the above-described embodiment of the intelligent energy-saving method for multi-agent 5G base stations based on big-small model fusion, this application also provides an embodiment of an intelligent energy-saving device for multi-agent 5G base stations based on big-small model fusion.

[0215] like Figure 7 As shown, the intelligent energy-saving device for multi-agent 5G base stations based on size model fusion provided in this embodiment includes:

[0216] The initial policy configuration module 101 is used to identify the network deployment scenario information corresponding to the current wireless network and configure an initial energy-saving policy that matches the network deployment scenario information.

[0217] The current strategy determination module 102 is used to filter energy-saving areas in the current wireless network based on the initial energy-saving strategy, and determine the current energy-saving strategy corresponding to the energy-saving area based on the existing network data corresponding to the energy-saving area.

[0218] The energy-saving control module 103 is used to generate an energy-saving interaction command based on the current energy-saving strategy and send it to the wireless network device corresponding to the energy-saving area. The energy-saving interaction command is used to control the energy-saving status of the wireless network device.

[0219] The energy-saving strategy update module 104 is used to iteratively optimize the current energy-saving strategy online based on the changing trends of the energy-saving effect and key network performance indicators corresponding to the energy-saving area.

[0220] Please see Figure 8 This application also provides an embodiment of an electronic device, which includes a bus 201, a processor 202, a communication interface 203, and a memory 204. The processor 202, the memory 204, and the communication interface 203 communicate with each other via the bus 201.

[0221] Bus 201 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, only one thick line is used in the diagram, but this does not indicate that there is only one bus or one type of bus.

[0222] The processor 202 can be any one or more of the following processors: central processing unit (CPU), graphics processing unit (GPU), microprocessor (MP), or digital signal processor (DSP).

[0223] Memory 204 may include volatile memory, such as random access memory (RAM). Memory 204 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid state drive (SSD).

[0224] The memory 204 can be used to store software code related to the intelligent energy-saving method for multi-agent 5G base stations based on size model fusion provided in this application. The processor 202 can execute the steps of the method in the memory, and can also schedule other units to achieve the corresponding functions.

[0225] This application also provides a computer program product including computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device enables the electronic device to implement any of the multi-agent 5G base station intelligent energy-saving methods based on size model fusion provided in this application.

[0226] This application also provides a computer storage medium carrying one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any of the multi-agent 5G base station intelligent energy-saving methods based on size model fusion provided in this application.

[0227] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the device embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0228] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0229] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.

[0230] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

Claims

1. A multi-agent intelligent energy-saving method for 5G base stations based on size-model fusion, characterized in that, include: Identify the network deployment scenario information corresponding to the current wireless network, and configure an initial energy-saving strategy that matches the network deployment scenario information; Based on the initial energy-saving strategy, energy-saving areas in the current wireless network are selected, and the current energy-saving strategy corresponding to the energy-saving area is determined based on the existing network data corresponding to the energy-saving area. Based on the current energy-saving strategy, an energy-saving interaction command is generated and sent to the wireless network device corresponding to the energy-saving area. The energy-saving interaction command is used to control the energy-saving status of the wireless network device. Based on the changing trends of the energy-saving effect and key network performance indicators corresponding to the energy-saving zone, the current energy-saving strategy is iteratively optimized online. The initial energy-saving strategy includes energy-saving threshold parameters, which include a traffic load threshold value for entering an energy-saving state; the step of filtering energy-saving areas in the current wireless network based on the initial energy-saving strategy includes: Areas where the predicted call load is lower than the corresponding call load threshold for entering energy-saving mode are defined as energy-saving areas. The step of determining the current energy-saving strategy corresponding to the energy-saving region based on the existing network data corresponding to the energy-saving region includes: Acquire the internal and external network data corresponding to the energy-saving area. The internal network data includes engineering parameter information and terminal measurement report data. The external network data includes external system data that affects traffic load performance indicators. Based on the internal and external network data, the predicted values ​​of the traffic load performance indicators corresponding to the energy-saving area are obtained using the load forecasting model. Based on the predicted values ​​of the call load performance indicators and the energy-saving threshold parameters corresponding to the energy-saving area, the optimal energy-saving period corresponding to the energy-saving area is determined; The current energy-saving strategy is obtained based on the energy-saving type and optimal energy-saving period corresponding to the energy-saving area. The step of iteratively optimizing the current energy-saving strategy online based on the energy-saving effect and the changing trends of key network performance indicators corresponding to the energy-saving zone includes: Increase the target adjustment step size by the current energy-saving threshold parameter corresponding to the energy-saving zone; Energy-saving control is performed on the energy-saving region based on the energy-saving threshold parameter after adjusting the step size by increasing the target, and the corresponding key network performance indicators are obtained. If the key network performance indicators do not decrease or decrease but do not exceed the allowable threshold, continue to execute the step of increasing the current energy-saving threshold parameter corresponding to the energy-saving zone by the target adjustment step size. If the key network performance indicators drop and exceed the allowable threshold, the energy-saving threshold parameter will be restored to the value before the target adjustment step size is increased, and the optimal energy-saving threshold parameter will be obtained. The method further includes: If the energy-saving zone is not in energy-saving mode, the network performance KPIs of the energy-saving zone and the same coverage area are evaluated. If the evaluation results of the network performance KPIs of the energy-saving zone and the same coverage area are poor, the energy-saving zone is prevented from entering the energy-saving mode; if the evaluation results of the network performance KPIs of the energy-saving zone and the same coverage area are excellent or there is no result, the energy-saving zone is allowed to enter the energy-saving mode. When the energy-saving zone has entered the energy-saving state, the network performance KPI of the same coverage area of ​​the energy-saving zone is evaluated. If the evaluation result of the network performance KPI of the same coverage area is poor, the energy-saving zone is activated; if the evaluation result of the network performance KPI of the same coverage area is excellent or there is no result, the energy-saving zone remains in the energy-saving state. The same coverage area is an area whose coverage is the same as or overlaps with the energy-saving area by more than a preset threshold. The method further includes: A knowledge base is built based on external data, and an energy-saving model is trained using the knowledge base and historical data of the wireless network. The knowledge base is built based on a large language model, and the energy-saving model is used to output the current energy-saving strategy based on internal network data and external network data.

2. The method according to claim 1, characterized in that, The network deployment scenario information includes network coverage information and network configuration information, and the initial energy-saving strategy includes energy-saving type, energy-saving period and energy-saving threshold parameters; The initial energy-saving strategy, which matches the configuration with the network deployment scenario information, includes: Configure the power saving type of the current wireless network based on the network configuration information of the current wireless network. The power saving type includes symbol shutdown, channel shutdown, carrier shutdown, or deep sleep of radio frequency module. Based on the historical traffic load performance indicators and network coverage information of the current wireless network, configure the energy-saving time period and energy-saving threshold parameters for each area of ​​the current wireless network.

3. The method according to claim 2, characterized in that, The step of configuring energy-saving time periods and energy-saving threshold parameters for each area of ​​the current wireless network, by combining the historical traffic load performance indicators and network coverage information of the current wireless network, includes: For any area of ​​the current wireless network, the off-peak traffic load level is determined based on the historical traffic load of that area; The energy-saving threshold parameter for any region is determined based on the off-peak call load level of any region. Based on the traffic load forecast results for any given region and the determined energy-saving threshold parameters, the energy-saving time period for any given region is determined.

4. The method according to claim 1, characterized in that, The method further includes: Detect whether there is a conflict in the current energy-saving strategy. If there is a conflict, modify the current energy-saving strategy. If there is no conflict, generate an energy-saving interaction command based on the current energy-saving strategy. Based on the execution result of the energy-saving interaction command, the system detects whether there are any omissions or conflicts in the current energy-saving strategy. If there are any omissions or conflicts, the system reports the conflict detection result of the current energy-saving strategy to the strategy management platform.

5. A smart energy-saving device for a multi-agent 5G base station based on size model fusion for performing the method according to any one of claims 1-4, characterized in that, include: The initial policy configuration module is used to identify the network deployment scenario information corresponding to the current wireless network and configure an initial energy-saving policy that matches the network deployment scenario information. The current strategy determination module is used to filter energy-saving areas in the current wireless network based on the initial energy-saving strategy, and determine the current energy-saving strategy corresponding to the energy-saving area based on the existing network data corresponding to the energy-saving area. The energy-saving control module is used to generate energy-saving interaction instructions based on the current energy-saving strategy and send them to the wireless network device corresponding to the energy-saving area. The energy-saving interaction instructions are used to control the energy-saving status of the wireless network device. The energy-saving strategy update module is used to iteratively optimize the current energy-saving strategy online based on the changing trends of the energy-saving effect and key network performance indicators corresponding to the energy-saving zone.

6. An electronic device, characterized in that, It includes at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is used to execute the computer program so that the electronic device can implement the intelligent energy-saving method for multi-agent 5G base stations based on size model fusion as described in any one of claims 1 to 4.

7. A computer storage medium, characterized in that, The storage medium carries one or more computer programs, which, when executed by an electronic device, enable the electronic device to implement the intelligent energy-saving method for multi-agent 5G base stations based on size model fusion as described in any one of claims 1 to 4.

Images