A base station power scheduling optimization method, apparatus, and electronic device

Abstract

This invention provides a base station power scheduling optimization method, apparatus, and electronic device, relating to the field of broadband technology. The method includes: dividing multiple base stations into at least one base station cluster based on computing power capacity and the corresponding transmission access point numbers; receiving basic information transmitted by each base station; acquiring real-time operating data and current scheduling strategies of each base station upon receiving a scheduling optimization request; determining whether to initiate federated training; and, if federated training is initiated, using real-time operating data and global parameters for learning and training to obtain a power scheduling optimization strategy for each base station. This invention designs a novel base station power optimization system architecture, overcoming the deficiency of insufficient computing power on the base station side and solving the problem of data silos for individual base stations. It forms a globally optimized scheduling optimization strategy to optimize and improve base station power, thereby reducing base station energy consumption and improving base station energy efficiency.

Description

Technical Field

[0001] This invention relates to the field of broadband technology, and in particular to a base station power scheduling optimization method, apparatus, and electronic equipment. Background Technology

[0002] With the arrival of the 5G era, the types of communication services and the number of terminal connections are experiencing explosive growth, posing unprecedented challenges to energy conservation and consumption reduction in the communications industry. Data shows that base stations are the biggest power consumers in mobile communication networks, accounting for approximately 80% of energy consumption. More densely packed base stations mean higher energy consumption, which is a major cost challenge for 5G networks.

[0003] Most energy-saving solutions for base stations utilize artificial intelligence technology. Based on historical service data from cells within a selected network area, a base station service load prediction model is constructed, generating a network energy-saving strategy. The energy-saving tasks for each base station are then executed according to this strategy. However, due to the complexity of service loads and the requirements for communication network assurance, while base station energy saving has achieved some results, there is still significant room for optimization and improvement.

[0004] The current method still has the following problems:

[0005] 1) Due to the significant reduction in the coverage area of ​​base stations and the high density of base station deployment, power optimization based on the service situation of individual base stations lacks global management and comprehensive consideration of the coverage and power management of neighboring base stations, making it difficult to achieve a global optimization effect.

[0006] 2) Adopting a direct communication and coordination method between base stations and neighboring base stations increases the computational complexity of the base station itself, raises the computational capability requirements of the base station, and is not conducive to global power optimization.

[0007] 3) Training and prediction using service models from multiple base stations to generate a globally optimized power management strategy requires mastering global data and having strong computing power. It also requires intelligent computing support from a computing center, which is not currently supported in base station management. Summary of the Invention

[0008] In view of the above problems, the present invention is proposed to provide a base station power scheduling optimization method, apparatus and electronic device to overcome the above problems or at least partially solve the above problems.

[0009] Firstly, a base station power scheduling optimization method is provided, which is applied to a cloud center. The base station power scheduling optimization method includes:

[0010] Based on the computing power capacity and the transmission access point numbers corresponding to the multiple base stations, the multiple base stations are divided into at least one base station cluster;

[0011] Receive basic information sent by each base station in each base station cluster, and use the basic information as a global parameter;

[0012] Upon receiving a scheduling optimization request from the target base station, the system obtains the real-time operating data and current scheduling strategy of each base station in the target base station cluster to which the target base station is located;

[0013] Determine whether to start federated training based on the real-time running data, the current scheduling strategy, and preset conditions;

[0014] If the federated training is initiated, the federated training is used to learn and train based on the real-time running data, the global parameters, and the current scheduling policy to obtain the power scheduling optimization policy for each base station in the target base station cluster, and then the policy is sent to each base station.

[0015] If it is determined that the federated training will not be initiated, a rejection request is sent to the target base station cluster.

[0016] Optionally, based on computing power capacity and the transmission access point numbers of multiple base stations, the multiple base stations are divided into at least one base station cluster, including:

[0017] Upon receiving a registration request from each base station, the transmission access point number carried in each registration request is parsed to obtain the information.

[0018] Based on the transmission access point number, base stations corresponding to the same transmission access point number are grouped together, and the number of base stations in a group is determined.

[0019] Based on the number of base stations in a group, determine whether the computing power capacity is sufficient to perform federated training on all base stations in the group.

[0020] If the computing power capacity meets the required computing power, then the group of base stations is determined to form a base station cluster.

[0021] Optionally, the preset conditions include: the policy validity period and the average power threshold per household;

[0022] Determining whether to initiate federated training based on the operational data, the current scheduling policy, and preset conditions includes:

[0023] The real-time average power per user of each base station in the target base station cluster is determined based on the real-time operating data.

[0024] If the real-time average power per user of any base station exceeds the average power per user threshold, then the federated training is initiated; or,

[0025] If the validity period of the current scheduling policy of any base station exceeds the validity period of the policy, then the federated training is initiated.

[0026] Optionally, based on the real-time operating data, the global parameters, and the current scheduling policy, the federated training is used to learn and train, obtaining the power scheduling optimization policy for each base station in the target base station cluster, including:

[0027] Using the parameter matrices corresponding to the real-time operating data and the current scheduling strategy as a training pair, and combining the global parameters to update the parameters of the federated training model, the power scheduling optimization strategy for each base station in the target base station cluster is obtained through learning and training.

[0028] Optionally, after dividing the plurality of base stations into at least one base station cluster, the method further includes:

[0029] Based on the base station cluster, a periodic instruction is sent to all base stations in each base station cluster. The periodic instruction is used to instruct each base station to send real-time operation data to the cloud center periodically.

[0030] Receive real-time operational data sent by each base station periodically.

[0031] Secondly, a base station power scheduling optimization method is provided, which is applied to a base station and includes:

[0032] After determining that it will join the base station cluster, it sends its basic information to the cloud center as global parameters. The base station cluster is divided by the cloud center according to the computing power capacity and the transmission access point number of multiple base stations.

[0033] Under the condition that the preset conditions are met, a scheduling optimization request is sent to the cloud center;

[0034] Based on the request from the cloud center, the current running data is preprocessed to obtain standardized real-time running data, which is then sent to the cloud center along with the current scheduling strategy.

[0035] Receive a rejection request from the cloud center, which is sent by the cloud center based on the real-time operating data, the current scheduling policy, and the preset conditions, when it determines that federated training should not be started; or

[0036] The system receives and executes a power scheduling optimization strategy from the cloud center. This power scheduling optimization strategy is obtained and sent by the cloud center through learning and training based on the running data and the global parameters, when the cloud center determines to start the federated training.

[0037] Optionally, before joining a base station cluster, the base station power scheduling optimization method includes:

[0038] The registration application is sent to the cloud center through its corresponding transmission access point, and the registration application carries the transmission access point number.

[0039] If joining the base station cluster is successful, a key is received from the cloud center, and data communication with the cloud center is performed based on the key;

[0040] If the user does not join the base station cluster, they will not receive the key and will be unable to communicate with the cloud center.

[0041] Optionally, the preset conditions include: the policy validity period and the average power threshold per household;

[0042] Under preset conditions, a scheduling optimization request is sent to the cloud center, including:

[0043] Determine your own real-time average household power based on the real-time operating data;

[0044] If the real-time average power per household exceeds the average power per household threshold, a scheduling optimization request is sent to the cloud center; or

[0045] If the validity period of the current scheduling policy exceeds the validity period of the policy, a scheduling optimization request is sent to the cloud center.

[0046] Optionally, after determining to join the base station cluster, the following steps are also included:

[0047] Receive periodic instructions from the cloud center, the periodic instructions being used to instruct the base station to send real-time operation data to the cloud center periodically;

[0048] According to the periodic instructions, real-time operation data is sent to the cloud center periodically.

[0049] Thirdly, a base station power scheduling optimization method is provided, which is applied to a cloud-edge collaborative system. The cloud-edge collaborative system includes a cloud center and at least one base station cluster. The base station power scheduling optimization method includes:

[0050] Multiple base stations send registration applications to the cloud center through their respective transmission access points, and the registration applications carry the transmission access point number;

[0051] The cloud center divides the multiple base stations into at least one base station cluster based on the computing power capacity and the transmission access point numbers corresponding to the multiple base stations;

[0052] Each base station in each base station cluster sends its basic information to the cloud center, and the cloud center uses the basic information as a global parameter.

[0053] Under the condition that the target base station meets the preset conditions, it sends a scheduling optimization request to the cloud center;

[0054] The cloud center receives the scheduling optimization request and obtains the real-time operating data and current scheduling strategy of each base station in the target base station cluster where the target base station is located.

[0055] The cloud center determines whether to start federated training based on the real-time operating data, the current scheduling strategy, and the preset conditions.

[0056] If the federated training is initiated, the cloud center uses the federated training to learn and train based on the real-time running data, the global parameters, and the current scheduling policy to obtain the power scheduling optimization policy for each base station in the target base station cluster, and then distributes it to each base station.

[0057] If it is determined that the federated training will not be initiated, the cloud center sends a rejection request to the target base station cluster;

[0058] The target base station receives the rejection request; or

[0059] The target base station receives and executes the power scheduling optimization strategy.

[0060] Fourthly, a base station power scheduling optimization device is provided, which is applied in a cloud center. The base station power scheduling optimization device includes:

[0061] The clustering module is used to divide the multiple base stations into at least one base station cluster based on the computing power capacity and the transmission access point number corresponding to the multiple base stations;

[0062] The basic information receiving module is used to receive basic information sent by each base station in each base station cluster, and use the basic information as a global parameter.

[0063] The data acquisition strategy module is used to acquire the real-time operating data and current scheduling strategy of each base station in the target base station cluster where the target base station is located when a scheduling optimization request is received from the target base station.

[0064] The judgment module is used to determine whether to start federated training based on the real-time running data, the current scheduling strategy, and preset conditions.

[0065] The training module is used to learn and train the power scheduling optimization strategy for each base station in the target base station cluster by using the federated training based on the real-time running data, the global parameters and the current scheduling strategy when the federated training is determined to be started, and to distribute the strategy to each base station.

[0066] The module for sending a rejection request is used to send a rejection request to the target base station cluster if it is determined that the federated training will not be initiated.

[0067] Fifthly, a base station power scheduling optimization device is provided, which is applied to a base station and includes:

[0068] The basic information sending module is used to send its own basic information to the cloud center as global parameters after determining that it has joined the base station cluster. The base station cluster is divided by the cloud center according to the computing power capacity and the transmission access point number of multiple base stations.

[0069] The optimization request sending module is used to send a scheduling optimization request to the cloud center when preset conditions are met.

[0070] The data transmission strategy module is used to send real-time operation data and the current scheduling strategy according to the request from the cloud center;

[0071] The execution module is used to receive a rejection request from the cloud center, which is sent by the cloud center when it determines that federated training will not be started based on the real-time running data, the current scheduling strategy, and the preset conditions.

[0072] The execution module is further configured to receive and execute a power scheduling optimization strategy from the cloud center. The power scheduling optimization strategy is obtained and sent by the cloud center through learning and training based on the running data and the global parameters when the cloud center determines to start the federated training.

[0073] Sixthly, an electronic device is provided, comprising:

[0074] One or more processors; and

[0075] One or more machine-readable media storing instructions thereon, which, when executed by the one or more processors, cause the electronic device to perform the base station power scheduling optimization method as described in any of the first aspects; or,

[0076] When executed by the one or more processors, the electronic device performs the base station power scheduling optimization method as described in any of the second aspects.

[0077] This application has the following advantages:

[0078] In this invention, the cloud center divides multiple base stations into at least one base station cluster based on computing power capacity and the transmission access point numbers corresponding to multiple base stations, and implements a method for global scheduling optimization based on the base station clusters. The cloud center receives basic information sent by each base station in each base station cluster, and uses the basic information as global parameters. When a scheduling optimization request is received from a target base station, the cloud center obtains the real-time operating data and current scheduling strategy of each base station in the target base station cluster. Based on the real-time operating data, the current scheduling strategy, and preset conditions, the cloud center determines whether to start federated training. If federated training is started, the cloud center uses federated training to learn and train based on the real-time operating data and global parameters to obtain the power scheduling optimization strategy of each base station in the target base station cluster, and distributes it to each base station. If federated training is not started, the cloud center sends a rejection request to the target base station cluster, without increasing the computational load of the cloud center, thus improving the efficiency of the cloud center in handling other matters.

[0079] This method forms a base station cluster by connecting all base stations connected to the same access point via fiber optic cables. Using this cluster as an optimization area, a novel base station power optimization system architecture is designed. Based on a cloud-edge collaborative architecture, the base stations act as the edge to preprocess data, while the cloud center trains and predicts based on the comprehensive data from each base station within the cluster, forming a power scheduling optimization method. This reduces the management network costs associated with inter-base station collaboration, allowing for the reuse of existing transmission network architecture while ensuring data security. It also overcomes the limitation of insufficient computing power at the base station level, resolving the issue of data silos at individual base stations. The new method of cloud center scheduling edge nodes to participate in federated learning offers higher scalability compared to traditional optimization models. The cloud center intelligently trains and predicts service models from multiple base stations at the access center, forming a globally optimized scheduling strategy to improve base station power, thereby reducing energy consumption and improving energy efficiency. Attached Figure Description

[0080] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0081] Figure 1 This is a flowchart of a base station power scheduling optimization method according to an embodiment of the present invention;

[0082] Figure 2 This is an overview flowchart of the base station cluster initialization in an embodiment of the present invention;

[0083] Figure 3 This is an overview flowchart of the process by which the base station initiates a scheduling optimization request and the cloud center decides whether to start federated training in an embodiment of the present invention.

[0084] Figure 4 This is an overview flowchart of cloud center training and execution in an embodiment of the present invention;

[0085] Figure 5 This is another flowchart of a base station power scheduling optimization method according to an embodiment of the present invention;

[0086] Figure 6 This is a block diagram of a base station power scheduling optimization device according to an embodiment of the present invention;

[0087] Figure 7 This is another block diagram of a base station power scheduling optimization device according to an embodiment of the present invention;

[0088] Figure 8 This is an exemplary embodiment of the present invention, showing a comparison of the overall structure between the method proposed in this invention and current conventional methods. Detailed Implementation

[0089] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention, and are only some, not all, embodiments of the present invention, and are not intended to limit the present invention.

[0090] The inventors discovered that with the arrival of the 5G era, the types of communication services and the number of terminal connections are experiencing explosive growth, posing unprecedented challenges to energy conservation and consumption reduction in the communications industry. Further research revealed that the current idle power consumption of operators' 5G base station main equipment is approximately 2.2-2.3kW, and the full-load power consumption is approximately 3.7-3.9kW, which is about 3-4 times that of 4G base stations.

[0091] In mobile communication networks, base stations are major power consumers, accounting for approximately 80% of energy consumption. A denser network of base stations means higher energy consumption, which is a significant cost challenge for 5G networks.

[0092] The internal structure of base station equipment mainly includes: BBU, radio frequency (RF) unit, power amplifier (PA), main power supply, antenna interface, and cooling system. The BBU contains control unit, transmission unit, and baseband processing unit, primarily responsible for signal filtering, OFDM, modulation and demodulation, frequency domain processing (symbol mapping / demapping and MIMO equalization, etc.), CPRI, and DPD (digital predistortion processing). Based on this structure, base station power consumption can be divided into three main types: transmission power consumption, computing power consumption, and additional power consumption.

[0093] Furthermore, as 5G frequency bands increase, macro base station signal transmission faces greater link loss issues, resulting in insufficient indoor signal coverage. On one hand, the higher frequency bands of 5G make it difficult for outdoor macro base station signals to reach indoor areas; on the other hand, more and more services and traffic demands occur indoors. Simply increasing macro base station coverage is prohibitively costly. Faced with this dilemma, small and micro base stations have become a powerful tool for solving the pain points of 5G networks, providing a new solution for achieving deep 5G coverage.

[0094] Macro base stations are the largest in size and have the widest signal coverage, primarily addressing long-distance continuous outdoor coverage. Micro base stations have a coverage range of 50-200 meters and are used in urban or rural areas where macro base station deployment is limited. Meanwhile, the coverage radius of 4G base stations is approximately 1 to 3 kilometers, while that of 5G base stations is approximately 100 to 300 meters. This significant reduction in coverage radius not only increases the complexity of network planning and base station construction but also places higher demands on inter-site coordination. With base station energy consumption costs becoming a significant component of operating costs, refined base station function management has become an urgent issue that needs to be addressed.

[0095] The inventors' research revealed that current energy-saving solutions for base stations primarily utilize artificial intelligence technology. Based on historical service data from cells within a selected network area, a base station service load prediction model is constructed, generating a network energy-saving strategy. This strategy is then applied to the energy-saving tasks of each base station. However, due to the complexity of service loads and the requirements for communication network security, while base station energy saving has achieved some success, there is still significant room for optimization and improvement.

[0096] Further research by the inventors revealed that current methods for solving the high energy consumption problem of base stations still have the following issues:

[0097] 1) Due to the significant reduction in the coverage area of ​​base stations and the high density of base station deployment, power optimization based on the service situation of individual base stations lacks global management and comprehensive consideration of the coverage and power management of neighboring base stations, making it difficult to achieve global power optimization results.

[0098] 2) Adopting a direct communication and coordination method between base stations and neighboring base stations increases the computational complexity of the base station itself, raises the computational capability requirements of the base station, and is not conducive to global power optimization.

[0099] 3) Training and prediction using service models from multiple base stations to generate a globally optimized power optimization strategy requires mastering global data and having strong computing power. It also requires intelligent computing support from a computing center, which is currently not supported by existing base station management systems.

[0100] To address the aforementioned problems, the inventors, through extensive research, creatively proposed the base station power scheduling optimization method of this invention. The following provides a detailed explanation and description of the base station power scheduling optimization method proposed in this invention.

[0101] Reference Figure 1 The flowchart illustrates a base station power scheduling optimization method according to an embodiment of the present invention. This base station power scheduling optimization method is applied to a cloud center and includes:

[0102] Step 101: Based on the computing power capacity and the transmission access point numbers corresponding to the multiple base stations, divide the multiple base stations into at least one base station cluster.

[0103] In this embodiment of the invention, in order to achieve global management and coverage and power management of neighboring base stations, obtain global power optimization effect, and reduce the computational complexity of a single base station, not only are a large amount of computation performed in the cloud center, but multiple base stations are also divided into base station clusters to form optimization areas.

[0104] The cloud center can divide multiple base stations into at least one base station cluster based on computing power capacity and the transmission access point numbers corresponding to the multiple base stations. In one possible embodiment, this step may specifically include:

[0105] Step S1: Upon receiving a registration request from each base station, parse the transmission access point number carried in each registration request.

[0106] Generally, after each base station is put into operation, in addition to data communication with neighboring base stations, all data communication with higher-level equipment or networks requires transmission via optical fiber to a transmission access point. The transmission access point then transmits the data to the higher-level equipment or network. All base stations within a certain range communicate with higher-level equipment or networks through a single transmission access point. For example, all base stations within a 5-kilometer radius may be connected to a single transmission access point via optical fiber. Alternatively, they may be divided by administrative or commercial areas; for instance, all base stations within an industrial park may be connected to a single transmission access point via optical fiber.

[0107] Based on the above structure, when each base station registers with the cloud center, the registration application sent needs to be transmitted through the transmission access point. Therefore, the registration application can carry the unique number of the transmission access point. In this embodiment of the invention, the transmission access point number is used to represent the unique number of the transmission access point. When the cloud center receives the registration application from each base station, it can parse it to obtain the transmission access point number carried in each registration request.

[0108] Step S2: Based on the transmission access point number, divide the base stations corresponding to the same transmission access point number into a group and determine the number of base stations in a group;

[0109] Step S3: Based on the number of base stations in a group, determine whether the computing power capacity is sufficient to perform federated training on all base stations in the group.

[0110] Step S4: If the computing power capacity meets the required computing power, determine that the group of base stations forms a base station cluster.

[0111] After obtaining the transmission access point numbers, the cloud center can group the base stations corresponding to the same transmission access point number into a group, thereby determining the number of base stations in a group. Then, based on the number of base stations in a group, it can determine whether the cloud center's computing power capacity is sufficient to perform federated training on all base stations in that group.

[0112] Generally, to ensure computing power and cloud center efficiency, the cloud center's computing capacity is compared based on a doubling of resource predictions. That is, the cloud center's computing capacity needs to be twice the computing power required for federated training of all base stations in a group. If both conditions are met, these base stations are allowed to form a base station cluster. For example, if a transmission access point corresponds to 30 base stations, when these 30 base stations register with the cloud center, the cloud center groups them into a group based on the transmission access point number, determining the number of base stations in this group to be 30. Assuming the average computing power required for federated training of data from one base station is 10, then 30 base stations would require a total of 300 computing power. If the cloud center determines the computing power capacity to be no less than 600, then these 30 base stations can form a base station cluster. If the cloud center determines that the computing power capacity is less than 600, then these 30 base stations cannot form a base station cluster. The cloud center will issue an instruction to inform the relevant staff of the situation. The relevant staff can adjust the equipment to increase the computing power capacity of the cloud center, or change the transmission access point of some of the 30 base stations to another transmission access point to reduce the number of base stations forming a base station cluster, so as to meet the quantity and capacity requirements of the cloud center. The specific method can be operated according to actual needs.

[0113] Using the above method, all base stations connected to the same access point via optical fiber form a base station cluster. This cluster serves as an optimized area, leading to a novel base station power optimization system architecture based on cloud-edge collaboration. In this architecture, base stations act as the edge layer to preprocess data, while the cloud center trains and predicts based on the comprehensive data from each base station within the cluster, forming a power scheduling optimization method. This reduces the management network costs associated with inter-base station collaboration, allowing for the reuse of existing transmission network architecture while ensuring data security. Furthermore, it overcomes the limitation of insufficient computing power at the base station level, resolving the issue of isolated data silos for individual base stations.

[0114] It should be noted that after a base station cluster is successfully formed, the cloud center will issue a management ID and a key to each base station according to a preset process. The cloud center will use this management ID to manage all base stations in the base station cluster, and use the key to encrypt data transmission during data communication.

[0115] The above process can be used Figure 2 The following is a flowchart of the base station cluster entry initialization process: First, the base station initiates an initial registration application to the cloud center. The cloud center then determines whether the application meets the cluster entry conditions. The main conditions include whether the base station's transmission access site is within the cluster and whether the current computing power capacity of the cloud center is sufficient.

[0116] If both of the above conditions are met, the base station will be allowed to join the cluster, and a management ID and key will be issued to the base station. The cloud center will use this management ID to manage the base station and use the key for encrypted data transmission. After receiving the management ID and key from the cloud center, the base station will use the key pair to send basic information to the cloud center, including the base station location, base station altitude, base station maximum power, antenna height, power supply mode, etc. Figure 2 For the sake of simplicity, not all basic information is shown. Complete the basic information for the base station. After registration, the base station will periodically send operational data to the cloud center according to the schedule issued by the cloud center, including the number of connected users, current power, interference-to-noise ratio, terminal distance, etc. Figure 2 For the sake of simplicity, not all running data is shown.

[0117] Step 102: Receive the basic information sent by each base station in each base station cluster, and use the basic information as global parameters.

[0118] In this embodiment of the invention, after a base station cluster is successfully formed, the cloud center receives the basic information sent by each base station in each base station cluster and uses this basic information as a global parameter.

[0119] As mentioned earlier, the cloud center sends management IDs and keys to the base stations. Upon receiving these, the base stations need to use the key pair to send basic information back to the cloud center to ensure data security. This basic information includes: base station location, base station height, antenna height, coordinates of neighboring base station overlay areas, maximum base station power, and power supply mode, enabling the cloud center to complete the base station's basic information. The base station location, height, antenna height, and coordinates of neighboring base station overlay areas are sent to the cloud center to address the data silo problem of individual base stations. This is not addressed in existing methods, as current methods are based on data from a single base station and do not require information such as location, height, antenna height, and coordinates of neighboring base station overlay areas for computation. However, the method proposed in this invention is based on global optimization, thus requiring consideration of all base stations within a base station cluster and necessitating the use of this information in the computation.

[0120] In addition, after dividing multiple base stations into at least one base station cluster, the cloud center also needs to issue periodic instructions to all base stations in each base station cluster. These periodic instructions are used to instruct each base station to send real-time operation data to the cloud center periodically. In this way, each base station sends real-time operation data to the cloud center when each period arrives. The cloud center receives the real-time operation data sent by each base station periodically. This real-time operation data includes: the number of access users, current power, interference plus noise ratio, terminal distance, etc., so that the cloud center can perform other business processing based on this real-time operation data.

[0121] Step 103: Upon receiving a scheduling optimization request from the target base station, obtain the real-time operating data and current scheduling strategy of each base station in the target base station cluster.

[0122] In routine operation, all base stations in the base station cluster operate according to set parameters and scheduling strategies. Each base station also has preset conditions, including a strategy validity period and an average power threshold per user. When a target base station in the cluster determines that scheduling optimization is needed based on these preset conditions, it sends a scheduling optimization request to the cloud center. The process by which a target base station determines the need for scheduling optimization based on the preset conditions is as follows: the target base station determines its real-time average power per user based on its real-time operational data; if the real-time average power per user exceeds the average power threshold in the preset conditions, the target base station sends a scheduling optimization request to the cloud center. Alternatively, if the target base station determines that the validity period of its current scheduling strategy exceeds the policy validity period in the preset conditions, the target base station also sends a scheduling optimization request to the cloud center.

[0123] When the cloud center receives a scheduling optimization request from the target base station, it obtains the real-time operating data and current scheduling policies of all base stations in the target base station cluster. It is understandable that if a base station determines that scheduling optimization is not needed based on preset conditions, it will not send a scheduling optimization request to the cloud center, and the cloud center will not perform subsequent federated training. That is, in the method proposed in this invention, whether federated training needs to be performed is initiated by the base station side, i.e., the edge side; if it is not initiated, federated training will not be performed. The final decision on whether to perform federated training is made by the cloud center.

[0124] In a preferred embodiment, the base station at the edge or the cloud center, with sufficient computing power, can respectively employ a model algorithm to update the current scheduling strategy in real time. Assume the parameters of the real-time running data are as follows:

[0125]

[0126]

[0127] The parameters of the current scheduling policy are as follows:

[0128] PRB utilization rate Total terminal traffic Q Average RRC Connections Reference signal received power RSRP Packet loss rate Plr Received Signal Strength (RSSI) Average delay T Reference signal reception quality (RSRQ) Downlink Channel Quality (CQI) Signal-to-Interference-plus-Noise Ratio (SINR) Power scheduling strategy parameters Example data Start time Time_start 8:30am End time (Time_end) 12:00am Base station power BS_pow 2000kw Frequency domain channel resources Channel FID[i] = 1 / 0 Time-domain channel resources Channel TID[j] = 1 / 0

[0129] By properly processing the above parameters and generating their corresponding feature vectors, a feature space is formed. Historical resource usage data is used as the input vector, and a backpropagation (BP) neural network is employed for training to generate a new scheduling policy vector, enabling real-time updates to the current scheduling policy. The training model formula is as follows: in, w i x is the weight of the i-th parameter feature. i Let y0 be the i-th running data parameter, n be the number of training cycles up to the present, and y0 be the scheduling vector for the first scheduling, which is usually a scheduling vector set manually based on experience. The initial values ​​of each parameter are also usually set manually based on experience. In this way, the edge base station can update the current scheduling policy in real time when there is sufficient computing power, and then send it to the cloud center. The cloud center receives multiple updated current scheduling policies and uses federated training to learn and train them.

[0130] Step 104: Determine whether to start federated training based on real-time running data, current scheduling strategy, and preset conditions.

[0131] After receiving any scheduling optimization request, the cloud center obtains the real-time operating data and current scheduling policy of each base station, and then determines whether to start federated training based on the real-time operating data, the current scheduling policy and preset conditions.

[0132] The cloud center also has preset conditions, which it then uses to determine whether to initiate federated training. Specifically:

[0133] The cloud center determines the real-time average power per user for each base station in the target base station cluster based on real-time operational data; if the real-time average power per user of any base station exceeds the average power per user threshold in the preset conditions, the cloud center determines to start federated training.

[0134] Alternatively, if the validity period of the current scheduling policy of any base station exceeds the policy validity period in the preset conditions, the cloud center will determine to start federated training.

[0135] The methods for the aforementioned base station to initiate scheduling optimization requests and for the cloud center to decide whether to initiate federated training can be referred to... Figure 3 Overview:

[0136] The scheduling optimization request is initiated by the base station. The base station periodically checks whether the current scheduling policy has expired or whether the current average power per user exceeds a threshold. If either condition is met, a scheduling optimization request is sent to the cloud center. Upon receiving the request, the cloud determines whether policy optimization is necessary, including whether the current scheduling policy has expired or whether the current average power per user exceeds the threshold. If either condition is met, federated training is initiated. If the relevant conditions are not met, the base station's scheduling optimization request is rejected, and federated training is not started.

[0137] Step 105: If federated training is initiated, the power scheduling optimization strategy for each base station in the target base station cluster is obtained by learning and training based on real-time running data, global parameters and current scheduling strategy, and then distributed to each base station.

[0138] Based on the judgment in step 104 above, when the cloud center determines to start federated training, it uses federated training to learn and train based on real-time running data, global parameters and current scheduling policies, to obtain the power scheduling optimization policy for each base station in the target base station cluster, and then distributes it to each base station.

[0139] In one possible embodiment, the specific method for the above-mentioned federated training to perform learning training is as follows:

[0140] The training pair consists of the parameter matrices corresponding to real-time operational data and the current scheduling strategy. The base station needs to preprocess the data. For example, it needs to de-identify the terminal user IDs, normalize the data to a single dimension (Dim), and generate corresponding matrices according to the set parameter sequence, such as the parameter matrix {t1, t2, ... t} corresponding to the real-time operational data. n The parameter matrix {I1, I2, ..., I} corresponding to the current scheduling strategy n}

[0141] These two parameter matrices serve as a training pair in the federated training process. They are then combined with global parameters to update the model parameters for further training, resulting in a power scheduling optimization strategy for each base station in the target base station cluster. Federated learning (FL) is a machine learning method. When training a model, it is necessary to set parameters such as participation rate, learning rate, momentum, batch size, and regularization. These are fundamental parameters in machine learning, and their settings can be referenced from existing parameter settings.

[0142] The above cloud center training and deployment process can be referenced. Figure 4 In summary:

[0143] The cloud center initiates policy optimization training based on requests from the base stations. First, the cloud sends data requests to all base stations within the cluster to collect the latest real-time operational data from each base station. Second, the operational data and global parameters are input into the training model to start training. After training using a federated learning algorithm, the scheduling optimization policy for each base station is obtained, and the trained scheduling optimization policy is distributed. Finally, the base stations execute the policies based on the acquired scheduling information.

[0144] Step 106: If it is determined that federated training will not be initiated, send a rejection request to the target base station cluster.

[0145] As mentioned above, based on the judgment in step 104, the cloud center sends a rejection request to the target base station cluster if it determines that federated training will not be initiated.

[0146] The above explanation and description of the base station power scheduling optimization method proposed in this invention is based on the cloud center side. For the edge side, i.e., the base station side, refer to... Figure 5 Base station power scheduling optimization methods include:

[0147] Step 501: After determining to join the base station cluster, send its own basic information to the cloud center as global parameters. The base station cluster is divided by the cloud center based on computing power capacity and the transmission access point number of multiple base stations.

[0148] First, the base station sends a registration request to the cloud center through its corresponding transmission access point. This registration request carries the transmission access point number. If it successfully joins the base station cluster, it receives a key from the cloud center and communicates with the cloud center based on the key. If it does not join the base station cluster, it does not receive the key and cannot communicate with the cloud center. After confirming that it has joined the base station cluster, it sends its basic information to the cloud center as global parameters based on the key.

[0149] Simultaneously, after determining whether to join the base station cluster, each base station also receives periodic instructions from the cloud center. These periodic instructions instruct the base station to send real-time operational data to the cloud center periodically. Each base station, according to the periodic instructions, preprocesses the current operational data according to the periodic schedule before sending the real-time operational data to the cloud center. The specific methods for this part can be found in steps 101-102 above, and will not be repeated here.

[0150] Step 502: If the preset conditions are met, send a scheduling optimization request to the cloud center;

[0151] Step 503: Based on the request from the cloud center, preprocess the current running data to obtain standardized real-time running data, and send it to the cloud center along with the current scheduling strategy.

[0152] All base stations in the base station cluster operate routinely according to set parameters and scheduling policies. Each base station has preset conditions, including policy validity period and average power threshold per user. When a target base station in the cluster determines that scheduling optimization is needed based on these preset conditions, it sends a scheduling optimization request to the cloud center. The process by which a target base station determines the need for scheduling optimization based on the preset conditions is as follows: the target base station determines its real-time average power per user based on its real-time operational data; if the real-time average power per user exceeds the average power threshold in the preset conditions, the target base station sends a scheduling optimization request to the cloud center. Alternatively, if the target base station determines that the validity period of its current scheduling policy exceeds the policy validity period in the preset conditions, the target base station also sends a scheduling optimization request to the cloud center.

[0153] After sending a scheduling optimization request, each base station preprocesses its current operational data according to the cloud center's request to obtain standardized real-time operational data. This preprocessing is for de-identifying user information and normalizing operational data, measurement parameters, and other data, unifying dimensions to facilitate more efficient and accurate calculations by the cloud center. Since the cloud center receives data from each base station, performing such preprocessing operations on its own would involve a large volume of data and computational demands, increasing energy consumption and reducing operational efficiency. However, having each base station perform preprocessing itself reduces computational load and energy consumption increases, indirectly improving the cloud center's processing efficiency.

[0154] After each base station preprocesses and obtains standardized real-time operational data, it sends the data, along with its respective current scheduling strategy, to the cloud center. The specific method for this part can be found in step 103 above, and will not be repeated here.

[0155] Step 504: Receive a rejection request from the cloud center. The rejection request is sent by the cloud center when it determines not to start federated training based on real-time running data, the current scheduling policy, and preset conditions; or, receive and execute a power scheduling optimization policy from the cloud center. The power scheduling optimization policy is obtained by the cloud center through learning and training using federated training based on running data, global parameters, and the current scheduling policy when it determines to start federated training.

[0156] If the cloud center decides not to initiate federated training, it sends a rejection request to the base stations. If the cloud center decides to initiate federated training, it uses the running data and global parameters to learn and train a power scheduling optimization strategy, which is then sent to each base station. Each base station receives the strategy and executes it accordingly. The specific methods for this part can be found in steps 104 to 106 above, and will not be repeated here.

[0157] Using the above method, all base stations connected to the same access point via optical fiber form a base station cluster. This cluster serves as an optimization area, leading to a novel base station power optimization system architecture. Based on a cloud-edge collaborative architecture, the base stations act as the edge, performing data preprocessing. The cloud center trains and predicts based on the comprehensive data from each base station within the cluster, forming a power scheduling optimization method. This reduces the management network costs associated with inter-base station collaboration, allowing for the reuse of existing transmission network architecture while ensuring data security. Furthermore, it overcomes the limitation of insufficient computing power at the base station level, resolving the issue of data silos within individual base stations. The new method of cloud center scheduling edge nodes for federated learning offers higher scalability compared to traditional optimization models. The cloud center intelligently trains and predicts service models from multiple base stations at the access center, forming a globally optimized scheduling strategy to improve base station power, thereby reducing energy consumption and increasing energy efficiency.

[0158] Based on the above-mentioned base station power scheduling optimization method, this invention also proposes another base station power scheduling optimization method, which is applied to a cloud-edge collaborative system. The cloud-edge collaborative system includes a cloud center and at least one base station cluster. The base station power scheduling optimization method includes:

[0159] Step V1: Multiple base stations send registration applications to the cloud center through their respective transmission access points, with the registration applications carrying the transmission access point number;

[0160] Step V2: The cloud center divides multiple base stations into at least one base station cluster based on the computing power capacity and the transmission access point numbers corresponding to the multiple base stations.

[0161] The specific methods for steps V1 to V2 can be found in the aforementioned steps 101 or 501, and will not be repeated here.

[0162] Step V3: Each base station in each base station cluster sends its basic information to the cloud center, which uses the basic information as a global parameter.

[0163] The specific method for step V3 can be found in the aforementioned steps 102 or 501, and will not be repeated here.

[0164] Step V4: If the target base station meets the preset conditions, it sends a scheduling optimization request to the cloud center.

[0165] The specific method for step V4 can be found in step 502 above, and will not be repeated here.

[0166] Step V5: The cloud center receives the scheduling optimization request and obtains the real-time operating data and current scheduling strategy of each base station in the target base station cluster.

[0167] Step V6: The cloud center determines whether to start federated training based on real-time operational data, the current scheduling strategy, and preset conditions;

[0168] Step V7: If the federated training is initiated, the cloud center uses the federated training to learn and train based on the real-time running data, the global parameters and the current scheduling policy, to obtain the power scheduling optimization policy for each base station in the target base station cluster, and then distributes it to each base station.

[0169] Step V8: If it is determined that the federated training will not be initiated, the cloud center sends a rejection request to the target base station cluster.

[0170] The specific methods for steps V5 to V8 can be found in the aforementioned steps 103 to 106 or step 503, and will not be repeated here.

[0171] Step V9: The target base station receives the rejection request; or the target base station receives and executes the power scheduling optimization strategy.

[0172] The specific method for step V9 can be found in step 504 above, and will not be repeated here.

[0173] Based on the above-described base station power scheduling optimization method, this invention also provides a base station power scheduling optimization device, referring to... Figure 6 The diagram illustrates a block diagram of a base station power scheduling optimization device according to an embodiment of the present invention. The base station power scheduling optimization device is applied in a cloud center and includes:

[0174] The clustering module 610 is used to divide the multiple base stations into at least one base station cluster based on the computing power capacity and the transmission access point number corresponding to the multiple base stations.

[0175] The basic information receiving module 620 is used to receive basic information sent by each base station in each base station cluster, and use the basic information as a global parameter.

[0176] The data acquisition strategy module 630 is used to acquire the real-time operating data and current scheduling strategy of each base station in the target base station cluster where the target base station is located when a scheduling optimization request is received from the target base station.

[0177] The judgment module 640 is used to determine whether to start federated training based on the real-time running data, the current scheduling strategy, and preset conditions.

[0178] The training module 650 is used to, when the federated training is determined to be started, learn and train using the federated training based on the real-time running data, the global parameters and the current scheduling policy, to obtain the power scheduling optimization policy of each base station in the target base station cluster, and distribute it to each base station.

[0179] The rejection request sending module 660 is used to send a rejection request to the target base station cluster if it is determined that the federated training will not be started.

[0180] Optionally, the cluster partitioning module 610 is specifically used for:

[0181] Upon receiving a registration request from each base station, the transmission access point number carried in each registration request is parsed to obtain the information.

[0182] Based on the transmission access point number, base stations corresponding to the same transmission access point number are grouped together, and the number of base stations in a group is determined.

[0183] Based on the number of base stations in a group, determine whether the computing power capacity is sufficient to perform federated training on all base stations in the group.

[0184] If the computing power capacity meets the required computing power, then the group of base stations is determined to form a base station cluster.

[0185] Optionally, the preset conditions include: strategy validity period and average household power threshold; the judgment module 640 is specifically used for:

[0186] The real-time average power per user of each base station in the target base station cluster is determined based on the real-time operating data.

[0187] If the real-time average power per user of any base station exceeds the average power per user threshold, then the federated training is initiated; or,

[0188] If the validity period of the current scheduling policy of any base station exceeds the validity period of the policy, then the federated training is initiated.

[0189] Optionally, the training module 650 is specifically used for:

[0190] Using the parameter matrices corresponding to the real-time operating data and the current scheduling strategy as a training pair, and combining the global parameters to update the parameters of the federated training model, the power scheduling optimization strategy for each base station in the target base station cluster is obtained through learning and training.

[0191] Optionally, the base station power scheduling optimization device further includes:

[0192] The periodic instruction module is used to issue periodic instructions to all base stations in each base station cluster based on the base station cluster. The periodic instructions are used to instruct each base station to send real-time operation data to the cloud center periodically.

[0193] The receiving periodic data module is used to receive real-time operational data sent by each base station according to a periodic schedule.

[0194] Based on the above-described base station power scheduling optimization method, this embodiment of the invention also provides another base station power scheduling optimization apparatus, referring to... Figure 7 The diagram illustrates a block diagram of another base station power scheduling optimization device according to an embodiment of the present invention. The base station power scheduling optimization device is applied to a base station and includes:

[0195] The basic information sending module 710 is used to send its own basic information to the cloud center as global parameters after determining that it has joined the base station cluster. The base station cluster is divided by the cloud center according to the computing power capacity and the transmission access point number of multiple base stations.

[0196] The optimization request sending module 720 is used to send a scheduling optimization request to the cloud center when preset conditions are met.

[0197] The data transmission strategy module 730 is used to send real-time operation data and the current scheduling strategy according to the request from the cloud center.

[0198] The execution module 740 is used to receive a rejection request from the cloud center, which is sent by the cloud center when it determines that federated training will not be started based on the real-time running data, the current scheduling strategy and the preset conditions.

[0199] The execution module 740 is further configured to receive and execute a power scheduling optimization strategy from the cloud center. The power scheduling optimization strategy is obtained and sent by the cloud center through learning and training based on the running data and the global parameters when the cloud center determines to start the federated training.

[0200] Optionally, the base station power scheduling optimization device includes:

[0201] The registration application module is used to send a registration application to the cloud center through its corresponding transmission access point, and the registration application carries the transmission access point number;

[0202] The key receiving module is used to receive a key issued by the cloud center if the user successfully joins the base station cluster, and to communicate with the cloud center based on the key; if the user does not join the base station cluster, the user will not receive the key and will be unable to communicate with the cloud center.

[0203] Optionally, the preset conditions include: policy validity period and average power threshold per household; the optimization request sending module 720 is specifically used for:

[0204] Determine your own real-time average household power based on the real-time operating data;

[0205] If the real-time average power per household exceeds the average power per household threshold, a scheduling optimization request is sent to the cloud center; or

[0206] If the validity period of the current scheduling policy exceeds the validity period of the policy, a scheduling optimization request is sent to the cloud center.

[0207] Optionally, the base station power scheduling optimization device further includes:

[0208] A periodic instruction receiving module is used to receive periodic instructions from the cloud center, the periodic instructions being used to instruct the base station to send real-time operation data to the cloud center periodically;

[0209] The periodic data transmission module is used to send real-time operation data to the cloud center according to the periodic instructions.

[0210] Based on the above-described base station power scheduling optimization method, this embodiment of the invention also provides an electronic device, including:

[0211] One or more processors; and

[0212] One or more machine-readable media storing instructions thereon, when executed by the one or more processors, cause the electronic device to perform the base station power scheduling optimization method as described in any one of steps 101 to 106; or,

[0213] When executed by one or more processors, the electronic device performs the base station power scheduling optimization method as described in any one of steps 501 to 504.

[0214] In summary, combining Figure 8An exemplary comparison diagram of the overall structure between the method proposed in this invention and current conventional methods is shown, which allows for a clear understanding: Figure 8 The diagram on the left represents the structure of the traditional method, where multiple terminals connect to a single base station, which in turn connects to the transmission network via transmission equipment. The power optimization control module, responsible for power scheduling optimization, is located within this single base station. Power optimization is performed based on the service conditions of each individual base station, lacking global management and comprehensive consideration of the coverage and power management of neighboring base stations, making it difficult to achieve overall optimization. Furthermore, the direct communication and coordination between a single base station and its neighbors increases the computational complexity of the individual base station, raising the computational requirements and hindering global power optimization.

[0215] Figure 8 The structural diagram on the right represents the structure of the method proposed in this invention. Multiple terminals are connected to one base station, and other base stations are also connected to other terminals. These multiple base stations form a base station cluster through a single transmission device (i.e., the same access point), connected to the transmission network. The power optimization control module, which enables power scheduling optimization, is located in the cloud center. Using this base station cluster as an optimization area, a novel base station power optimization system architecture is designed. Based on a cloud-edge collaborative architecture, the base stations act as the edge side to complete data preprocessing. The cloud center trains and predicts based on the comprehensive data from each base station within the cluster, forming a power scheduling optimization method. This reduces the management network cost of inter-base station collaboration, allowing the use of the existing transmission network architecture while also achieving data security protection. Simultaneously, it overcomes the deficiency of insufficient computing power on the base station side, solving the problem of data silos for individual base stations. The new method of scheduling edge nodes to participate in federated learning by the cloud center offers higher scalability compared to traditional optimization models. The cloud center intelligently trains and predicts the service models of multiple base stations at the access center, forming a globally optimized scheduling strategy to optimize and improve base station power, thereby reducing base station energy consumption and improving base station energy efficiency.

[0216] Through the above embodiments, the base station power scheduling optimization method of the present invention involves a cloud center dividing multiple base stations into at least one base station cluster based on computing power capacity and the transmission access point numbers corresponding to multiple base stations, and implementing a global scheduling optimization method based on the base station clusters. The cloud center receives basic information sent by each base station in each base station cluster, using the basic information as global parameters. When a scheduling optimization request is received from a target base station, the cloud center obtains the real-time operating data and current scheduling strategy of each base station in the target base station cluster. Based on the real-time operating data, the current scheduling strategy, and preset conditions, the cloud center determines whether to start federated training. If federated training is started, the cloud center uses federated training to learn and train based on the real-time operating data and global parameters, obtains the power scheduling optimization strategy of each base station in the target base station cluster, and distributes it to each base station. If federated training is not started, the cloud center sends a rejection request to the target base station cluster, without increasing the computational load of the cloud center and improving the efficiency of the cloud center in handling other matters.

[0217] This method forms a base station cluster by connecting all base stations connected to the same access point via fiber optic cables. Using this cluster as an optimization area, a novel base station power optimization system architecture is designed. Based on a cloud-edge collaborative architecture, the base stations act as the edge to preprocess data, while the cloud center trains and predicts based on the comprehensive data from each base station within the cluster, forming a power scheduling optimization method. This reduces the management network costs associated with inter-base station collaboration, allowing for the reuse of existing transmission network architecture while ensuring data security. It also overcomes the limitation of insufficient computing power at the base station level, resolving the issue of data silos at individual base stations. The new method of cloud center scheduling edge nodes to participate in federated learning offers higher scalability compared to traditional optimization models. The cloud center intelligently trains and predicts service models from multiple base stations at the access center, forming a globally optimized scheduling strategy to improve base station power, thereby reducing energy consumption and improving energy efficiency.

[0218] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0219] The technical solutions provided by the embodiments of the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A base station power scheduling optimization method, characterized in that, The base station power scheduling optimization method is applied to a cloud center, and the base station power scheduling optimization method includes: Based on the computing power capacity and the transmission access point number corresponding to the multiple base stations, the multiple base stations are divided into at least one base station cluster. When each base station communicates with the upper-level device or the upper-level network, it transmits data to the upper-level device or the upper-level network through the transmission access point. Receive basic information sent by each base station in each base station cluster, and use the basic information as global parameters. The basic information includes: base station location, base station height, base station maximum power, antenna mounting height, and power supply mode. Upon receiving a scheduling optimization request from the target base station, the system obtains the real-time operating data and current scheduling strategy of each base station in the target base station cluster to which the target base station is located; Determine whether to start federated training based on the real-time running data, the current scheduling strategy, and preset conditions; If the federated training is initiated, the federated training is used to learn and train based on the real-time running data, the global parameters, and the current scheduling policy to obtain the power scheduling optimization policy for each base station in the target base station cluster, and then the policy is sent to each base station. If it is determined that the federated training will not be initiated, a rejection request is sent to the target base station cluster.

2. The base station power scheduling optimization method according to claim 1, characterized in that, Based on computing power capacity and the transmission access point numbers of multiple base stations, the multiple base stations are divided into at least one base station cluster, including: Upon receiving a registration request from each base station, the transmission access point number carried in each registration request is parsed to obtain the information. Based on the transmission access point number, base stations corresponding to the same transmission access point number are grouped together, and the number of base stations in a group is determined. Based on the number of base stations in a group, determine whether the computing power capacity is sufficient to perform federated training on all base stations in the group. If the computing power capacity meets the required computing power, then the group of base stations is determined to form a base station cluster.

3. The base station power scheduling optimization method according to claim 1, characterized in that, The preset conditions include: the policy validity period and the average power threshold per household; Determining whether to initiate federated training based on the operational data, the current scheduling policy, and preset conditions includes: The real-time average power per user of each base station in the target base station cluster is determined based on the real-time operating data. If the real-time average power per user of any base station exceeds the average power per user threshold, then the federated training is initiated; or, If the validity period of the current scheduling policy of any base station exceeds the validity period of the policy, then the federated training is initiated.

4. The base station power scheduling optimization method according to claim 1, characterized in that, Based on the real-time operating data, the global parameters, and the current scheduling policy, the federated training is used to learn and train, obtaining the power scheduling optimization policy for each base station in the target base station cluster, including: Using the parameter matrices corresponding to the real-time operating data and the current scheduling strategy as a training pair, and combining the global parameters to update the parameters of the federated training model, the power scheduling optimization strategy for each base station in the target base station cluster is obtained through learning and training.

5. The base station power scheduling optimization method according to claim 1, characterized in that, After dividing the multiple base stations into at least one base station cluster, the method further includes: Based on the base station cluster, a periodic instruction is sent to all base stations in each base station cluster. The periodic instruction is used to instruct each base station to send real-time operation data to the cloud center periodically. Receive real-time operational data sent by each base station periodically.

6. A base station power scheduling optimization method, characterized in that, The base station power scheduling optimization method is applied to a base station, and the base station power scheduling optimization method includes: After determining that it will join the base station cluster, it sends its basic information to the cloud center as global parameters. The base station cluster is divided by the cloud center according to the computing power capacity and the transmission access point number of multiple base stations. When each base station communicates with the upper-level device or the upper-level network, it transmits data to the upper-level device or the upper-level network through the transmission access point. The basic information includes: base station location, base station height, base station maximum power, antenna height, and power supply mode. Under the condition that the preset conditions are met, a scheduling optimization request is sent to the cloud center; Based on the request from the cloud center, the current running data is preprocessed to obtain standardized real-time running data, which is then sent to the cloud center along with the current scheduling strategy. Receive a rejection request from the cloud center, which is sent by the cloud center based on the real-time operating data, the current scheduling strategy, and the preset conditions, when it determines that federated training should not be started; The system receives and executes a power scheduling optimization strategy from the cloud center. This power scheduling optimization strategy is obtained and sent by the cloud center through learning and training based on the running data and the global parameters, when the cloud center determines to start the federated training.

7. The base station power scheduling optimization method according to claim 6, characterized in that, Before joining a base station cluster, the base station power scheduling optimization method includes: The registration application is sent to the cloud center through its corresponding transmission access point, and the registration application carries the transmission access point number. If joining the base station cluster is successful, a key is received from the cloud center, and data communication with the cloud center is performed based on the key; If the user does not join the base station cluster, they will not receive the key and will be unable to communicate with the cloud center.

8. The base station power scheduling optimization method according to claim 6, characterized in that, The preset conditions include: the policy validity period and the average power threshold per household; Under preset conditions, a scheduling optimization request is sent to the cloud center, including: Determine your own real-time average household power based on the real-time operating data; If the real-time average power per household exceeds the average power per household threshold, a scheduling optimization request is sent to the cloud center; or If the validity period of the current scheduling policy exceeds the validity period of the policy, a scheduling optimization request is sent to the cloud center.

9. The base station power scheduling optimization method according to claim 6, characterized in that, After determining whether to join the base station cluster, the following is also included: Receive periodic instructions from the cloud center, the periodic instructions being used to instruct the base station to send real-time operation data to the cloud center periodically; According to the periodic instructions, real-time operation data is sent to the cloud center periodically.

10. A base station power scheduling optimization method, characterized in that, The base station power scheduling optimization method is applied to a cloud-edge collaborative system, which includes a cloud center and at least one base station cluster. The base station power scheduling optimization method includes: Multiple base stations send registration applications to the cloud center through their respective transmission access points. The registration applications carry the transmission access point number. When each base station communicates with the upper-level equipment or the upper-level network, it transmits data to the upper-level equipment or the upper-level network through the transmission access point. The cloud center divides the multiple base stations into at least one base station cluster based on the computing power capacity and the transmission access point numbers corresponding to the multiple base stations; Each base station in the base station cluster sends its own basic information to the cloud center. The cloud center uses the basic information as a global parameter. The basic information includes: base station location, base station height, base station maximum power, antenna mounting height, and power supply mode. Under the condition that the target base station meets the preset conditions, it sends a scheduling optimization request to the cloud center; The cloud center receives the scheduling optimization request and obtains the real-time operating data and current scheduling strategy of each base station in the target base station cluster where the target base station is located. The cloud center determines whether to start federated training based on the real-time operating data, the current scheduling strategy, and the preset conditions. If the federated training is initiated, the cloud center uses the federated training to learn and train based on the real-time running data, the global parameters, and the current scheduling policy to obtain the power scheduling optimization policy for each base station in the target base station cluster, and then distributes it to each base station. If it is determined that the federated training will not be initiated, the cloud center sends a rejection request to the target base station cluster; The target base station receives the rejection request; The target base station receives and executes the power scheduling optimization strategy.

11. A base station power scheduling optimization device, characterized in that, The base station power scheduling optimization device is applied in a cloud center, and the base station power scheduling optimization device includes: The clustering module is used to divide the multiple base stations into at least one base station cluster according to the computing power capacity and the transmission access point number corresponding to the multiple base stations. When each base station communicates with the upper-level device or the upper-level network, it transmits data to the upper-level device or the upper-level network through the transmission access point. The basic information receiving module is used to receive basic information sent by each base station in each base station cluster, and use the basic information as a global parameter. The basic information includes: base station location, base station height, base station maximum power, antenna mounting height, and power supply mode. The data acquisition strategy module is used to acquire the real-time operating data and current scheduling strategy of each base station in the target base station cluster where the target base station is located when a scheduling optimization request is received from the target base station. The judgment module is used to determine whether to start federated training based on the real-time running data, the current scheduling strategy, and preset conditions. The training module is used to learn and train the power scheduling optimization strategy for each base station in the target base station cluster by using the federated training based on the real-time running data, the global parameters and the current scheduling strategy when the federated training is determined to be started, and to distribute the strategy to each base station. The module for sending a rejection request is used to send a rejection request to the target base station cluster if it is determined that the federated training will not be initiated.

12. A base station power scheduling optimization device, characterized in that, The base station power scheduling optimization device is applied to a base station, and the base station power scheduling optimization device includes: The basic information sending module is used to send its own basic information to the cloud center as global parameters after determining that it has joined the base station cluster. The base station cluster is divided by the cloud center according to the computing power capacity and the transmission access point number of multiple base stations. When each base station communicates with the upper-level device or the upper-level network, it transmits data to the upper-level device or the upper-level network through the transmission access point. The basic information includes: base station location, base station height, base station maximum power, antenna mounting height, and power supply mode. The optimization request sending module is used to send a scheduling optimization request to the cloud center when preset conditions are met. The data transmission strategy module is used to send real-time operation data and the current scheduling strategy according to the request from the cloud center; The execution module is used to receive a rejection request from the cloud center, which is sent by the cloud center when it determines that federated training will not be started based on the real-time running data, the current scheduling strategy, and the preset conditions. The execution module is further configured to receive and execute a power scheduling optimization strategy from the cloud center. The power scheduling optimization strategy is obtained and sent by the cloud center through learning and training based on the running data and the global parameters when the cloud center determines to start the federated training.

13. An electronic device, characterized in that, include: One or more processors; and One or more machine-readable media on which instructions are stored, which, when executed by the one or more processors, cause the electronic device to perform the base station power scheduling optimization method as described in any one of claims 1 to 5; or, When executed by one or more processors, the electronic device performs the base station power scheduling optimization method as described in any one of claims 6 to 9.

Images