Base station load balancing method, device, storage medium and electronic equipment

Abstract

The application discloses a base station load balancing method and device, a storage medium and an electronic device, and relates to the technical field of communication. The method comprises the following steps: predicting the distribution characteristics of network service demand of a target scene in a target time period based on historical service data; adjusting the resource distribution of the base station according to the distribution characteristics; receiving a target service combination in the target time period, wherein the target service combination comprises M services and the priority of each service, and M is an integer greater than or equal to 1; based on the resource distribution of the base station, adjusting the target model according to the target service combination and a target algorithm, and generating a load balancing strategy for the target service combination based on the adjusted target model. The application solves the technical problem that the traditional load balancing algorithm is inefficient and cannot realize active load management when the base station load balancing scheme in the prior art faces rapid changes and burst conditions.

Description

Technical Field

[0001] This application relates to the field of communication technology, and more specifically, to a base station load balancing method, apparatus, storage medium, and electronic device. Background Technology

[0002] In modern communication networks, especially in 5G and future 6G network environments, base station load balancing solutions have become a key technology for ensuring network service quality and user experience. With the widespread adoption of mobile internet and the rise of the Internet of Things (IoT), wireless networks face unprecedented service demand challenges. In densely populated or frequently active areas, such as large sporting events, concerts, and shopping malls during holidays, the surge in user data traffic within a short period quickly overwhelms the base station's capacity, leading to network congestion and service interruptions.

[0003] The core of this problem lies in the limitations of existing base station load balancing algorithms. Traditional load balancing methods, such as rule-based static allocation strategies and dynamic allocation strategies that only passively adjust when overload is detected, are no longer able to cope with rapidly changing and highly dynamic service scenarios. When the network environment or user behavior changes abruptly, these algorithms often need to be retrained or undergo complex adjustments, which not only consumes a lot of time and computing resources, but may also cause service interruptions and degrade user experience during the adjustment process.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This application provides a base station load balancing method, apparatus, storage medium, and electronic device to at least solve the technical problems of low efficiency and inability to achieve proactive load management in traditional load balancing algorithms when facing rapid changes in services and sudden situations in the prior art.

[0006] According to one aspect of this application, a base station load balancing method is provided, comprising: predicting the distribution characteristics of network service demand for a target scenario within a target time period based on historical service data; adjusting the resource distribution of the base station according to the distribution characteristics; receiving a target service combination within the target time period, wherein the target service combination includes M services and the priority of each service, wherein M is an integer greater than or equal to 1; adjusting a target model based on the resource distribution of the base station, according to the target service combination and a target algorithm, and generating a load balancing strategy for the target service combination based on the adjusted target model, wherein the target algorithm is used to adjust the model parameters of the target model based on the target service combination without adjusting the structure of the target model, so that the target model generates a load balancing strategy that conforms to a constraint model, wherein the constraint model includes at least a strategy objective, wherein the strategy objective is used to characterize that the service rate of the service with a priority less than a preset value among the M services is greater than a preset rate, and the service time variance of the target service combination is less than a preset threshold.

[0007] Optionally, the constraint model is obtained through the following steps: setting an objective function, where the objective function is used to quantify the policy objective; determining N constraints based on the first model, where N is an integer greater than or equal to 1, and the first model is used to characterize the performance indicators of M services; and determining the constraint model based on the objective function and the N constraints.

[0008] Optionally, the target model is obtained through the following steps: obtaining S historical service combinations, wherein each historical service combination includes multiple historical services, and each historical service includes a corresponding priority; performing multiple target operations on the initial model based on the S historical service combinations until the number of target operations is greater than or equal to the number of the first iteration, thereby obtaining the target model, wherein each target operation is used to train the initial model based on each historical service combination.

[0009] Optionally, the target operation includes the following steps: performing multiple target trainings on the initial model based on the i-th historical service combination until a first parameter combination of the initial model is obtained, where i is an integer greater than or equal to 1 and less than or equal to S, and the first parameter combination is used to characterize the model parameters of the initial model under the constrained model; updating the initial model based on the first parameter combination to obtain the first model; determining the second parameter combination based on the first model using a target algorithm, where the second parameter combination is a guiding parameter determined by the target algorithm and used to adjust the model parameters of the first model; and determining the target model based on the second parameter combination and the first model.

[0010] Optionally, the target training includes the following steps: adjusting the initial parameters of the initial model based on the i-th historical service combination to obtain the third parameter combination; updating the initial model according to the third parameter combination to obtain the second model; training the i-th historical service combination according to the second model to generate the first policy; and determining whether the third parameter combination is the first parameter combination according to the first policy and the constraint model.

[0011] Optionally, determining whether the third parameter combination is the first parameter combination based on the first strategy and the constraint model includes: processing the i-th historical service combination according to the first strategy to obtain a processing result, wherein the processing result includes at least the performance indicators for completing the i-th historical service combination; determining whether the processing result conforms to the constraint model; and determining that the third parameter combination is the first parameter combination of the initial model if the processing result conforms to the constraint model.

[0012] Optionally, after adjusting the target model based on the target service composition and generating a load balancing strategy for the target service composition based on the adjusted target model, the method further includes: allocating resources to the target service composition according to the load balancing strategy to obtain allocation results; executing the target service composition based on the allocation results to obtain execution results, wherein the execution results include at least the performance indicators of completing the target service composition; and updating the target model based on the execution results.

[0013] According to another aspect of this application, a base station load balancing device is also provided, comprising: a prediction unit, configured to predict the distribution characteristics of network service demand for a target scenario within a target time period based on historical service data; an adjustment unit, configured to adjust the resource distribution of the base station according to the distribution characteristics; a receiving unit, configured to receive a target service combination within the target time period, wherein the target service combination includes M services and the priority of each service, wherein M is an integer greater than or equal to 1; and a generation unit, configured to adjust a target model based on the resource distribution of the base station, the target service combination, and a target algorithm, and generate a load balancing strategy for the target service combination based on the adjusted target model, wherein the target algorithm is configured to adjust the model parameters of the target model based on the target service combination without adjusting the structure of the target model, so that the target model generates a load balancing strategy that conforms to a constraint model, wherein the constraint model includes at least a strategy objective, wherein the strategy objective is configured to characterize that the service rate of the service with a priority less than a preset value among the M services is greater than a preset rate, and the service time variance of the target service combination is less than a preset threshold.

[0014] According to another aspect of this application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the above-described base station load balancing method.

[0015] According to another aspect of this application, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to perform the base station load balancing method described above.

[0016] In this application, the distribution characteristics of network service demand for a target scenario within a target time period are first predicted based on historical service data. Then, the resource distribution of the base station is adjusted according to the distribution characteristics. Next, a target service combination within the target time period is received. The target service combination includes M services and the priority of each service, where M is an integer greater than or equal to 1. Finally, based on the resource distribution of the base station, the target model is adjusted according to the target service combination and the target algorithm. A load balancing strategy for the target service combination is generated based on the adjusted target model. The target algorithm is used to adjust the model parameters of the target model based on the target service combination without adjusting the structure of the target model, so that the target model generates a load balancing strategy that conforms to the constraint model. The constraint model includes at least a policy objective, which is used to characterize that the service rate of the service with a priority less than a preset value among the M services is greater than a preset rate, and the service time variance of the target service combination is less than a preset threshold. This approach combines preliminary adjustments to base station resource distribution based on historical service data prediction with dynamic parameter adjustments. It achieves the goal of quickly adapting to and optimizing model parameters to cope with rapid changes in service demands and sudden load situations without altering the model structure. This results in efficient and proactive base station load balancing, thus solving the technical problem of inefficiency and inability to achieve proactive load management in traditional load balancing algorithms when facing rapid service changes and sudden situations. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0018] Figure 1 This is a flowchart of an optional base station load balancing method according to an embodiment of this application;

[0019] Figure 2 This is a schematic diagram of an optional base station load balancing device according to an embodiment of this application. Detailed Implementation

[0020] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0021] It should be noted that the terms "first," "second," etc., 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 data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0022] It should be noted that the information collected in this application (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) are information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of this data all comply with relevant laws, regulations, and standards, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding access points are provided for users to choose to authorize or refuse. For example, interfaces are set up between this system and relevant users or organizations, providing users with corresponding access points to choose to agree to or refuse automated decision-making results; if the user chooses to refuse, the process proceeds to the expert decision-making stage.

[0023] According to an embodiment of this application, a method embodiment of a base station load balancing method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0024] It should be noted that an intelligent processing system can serve as the execution subject of the base station load balancing method in the embodiments of this application. It is understood that the base station load balancing method provided in the embodiments of this application can also be executed by other systems or devices, and the embodiments of this application do not specifically limit this.

[0025] Figure 1 This is a flowchart of an optional base station load balancing method according to an embodiment of this application, such as... Figure 1 As shown, the method includes the following steps:

[0026] Step S101: Based on historical service data, predict the distribution characteristics of network service demand for the target scenario within the target time period.

[0027] Optionally, historical service data refers to the usage records of different types of services in the network over a period of time, including the request frequency, data volume, SLA level (Service Level Agreement, which refers to the level of service quality standards agreed upon between the service provider and the user) and actual network performance data (such as latency, packet loss rate, transmission rate, etc.) for each service.

[0028] Optionally, in the field of communication networks, especially for advanced application scenarios such as ultra-reliable low-latency communication, SLA levels define various performance indicators of the service, including but not limited to latency, packet loss rate, bandwidth, service availability, and the specific numerical requirements for these indicators. Different SLA levels correspond to different levels of service quality commitments.

[0029] Optionally, distribution characteristics involve the statistical features of service requests, such as the distribution of request frequency, the distribution of data packet size, and the distribution of service priority. These characteristics can help predict future service demand patterns.

[0030] Optionally, the intelligent processing system analyzes historical service data to predict the distribution of network service demand within a specific future target time period (e.g., the next few hours) under a particular scenario, including key information such as the request frequency and data volume of different types of services. This step is the foundation of proactive load balancing, enabling base stations to prepare resources before peak demand arrives.

[0031] Step S102: Adjust the resource distribution of the base station according to the distribution characteristics.

[0032] Alternatively, resource distribution refers to how a base station allocates its limited resources (such as spectrum, bandwidth, transmission power, etc.) to different services to meet service requirements and SLA levels.

[0033] Optionally, based on the service demand distribution predicted in the above steps, the base station adjusts its resource allocation strategy, giving priority to the demand for high-priority services, ensuring the effective use of resources, and leaving room for possible traffic surges.

[0034] Optionally, the intelligent processing system can predict traffic hotspots in advance based on the predicted traffic distribution, enabling base stations in the hotspot areas to proactively perform load balancing, thereby improving the average throughput of the system.

[0035] Step S103: Receive the target service combination within the target time period.

[0036] In step S103, the target service combination includes M services and the priority of each service.

[0037] In step S103, M is an integer greater than or equal to 1.

[0038] Optionally, the target service combination refers to the set of all services that the base station needs to process within the target time period and their priorities (referring to SLA levels), where M represents the number of services.

[0039] Optionally, the intelligent processing system receives and processes service requests in real time. These requests constitute a target service combination, including different service types and their respective priorities. This is crucial for achieving dynamic load balancing, ensuring that base stations can respond promptly to network changes and optimize resource allocation.

[0040] Step S104: Based on the resource distribution of the base station, adjust the target model according to the target service combination and the target algorithm, and generate a load balancing strategy for the target service combination based on the adjusted target model.

[0041] In step S104, the target algorithm is used to adjust the model parameters of the target model based on the target service combination without adjusting the structure of the target model, so that the target model generates a load balancing strategy that conforms to the constraint model.

[0042] In step S104, the constraint model includes at least the policy objective.

[0043] In step S104, the strategy objective is used to characterize that the service rate of the service with a priority lower than a preset value among the M services is greater than a preset rate, and the service time variance of the target service combination is less than a preset threshold.

[0044] Optionally, the target algorithm refers to a specific algorithm used to adjust the parameters of the target model. In this embodiment, the target algorithm is a meta-learning algorithm. It can quickly adjust the model parameters according to the target service combination without changing the model structure, so as to generate a load balancing strategy that conforms to the constraint model.

[0045] It should be noted that meta-learning algorithms are an advanced machine learning technique. Their core idea is to train a model on a series of related but different tasks, enabling the model to quickly adapt to new tasks. Specifically, in this embodiment, the meta-learning algorithm is designed to solve the load balancing problem among services within a base station, especially in scenarios facing rapid business changes and requiring proactive management. Meta-learning, also known as learning how to learn, can be broadly divided into optimization-based and probabilistic-based methods. Optimization-based meta-learning methods, such as MAML (Model-Agnostic Meta-Learning) and gradient-based meta-learning algorithms, find a set of parameters that achieve good performance on new tasks with only a few gradient updates. Probabilistic meta-learning methods, such as Bayesian Optimization and Variational Inference, construct a prior distribution of parameters between tasks and then perform posterior inference on new tasks to quickly obtain effective parameter configurations.

[0046] Optionally, the constraint model defines the conditions that the load balancing strategy must meet. For example, the strategy objective requires that the service rate of the service with a priority lower than a certain preset value among the M services must be greater than the preset rate, and the service time variance of the entire service combination must be less than a preset threshold, so as to ensure service quality and load balancing.

[0047] Optionally, the target model refers to a model specifically designed to generate load balancing strategies after parameter tuning using a meta-learning algorithm. It comprehensively considers historical data predictions, current resource distribution, service composition, and priority information, enabling it to provide customized solutions for real-time network environments.

[0048] Optionally, the intelligent processing system rapidly adjusts the target model using a target algorithm (meta-learning), enabling it to quickly adapt to changes in the target service composition and generate a load balancing strategy that meets the requirements of the constrained model. After parameter adjustment, the target model can intelligently maintain the quality of high-priority services while improving the service rate of low-priority services, and simultaneously control service time variance to avoid excessive service time fluctuations that could harm the user experience. Through this dynamic and intelligent adjustment process, the base station can achieve efficient and fair resource allocation, maintaining good service quality even in complex scenarios with rapidly changing network demands.

[0049] Optionally, the intelligent processing system utilizes meta-learning to tune parameters based on the trained model, thereby accelerating model training and significantly reducing costs.

[0050] As can be seen from steps S101 to S104, in this application, firstly, the distribution characteristics of network service demand of the target scenario within the target time period are predicted based on historical service data. Then, the resource distribution of the base station is adjusted according to the distribution characteristics. Next, the target service combination within the target time period is received. The target service combination includes M services and the priority of each service, where M is an integer greater than or equal to 1. Finally, based on the resource distribution of the base station, the target model is adjusted according to the target service combination and the target algorithm. Based on the adjusted target model, a load balancing strategy for the target service combination is generated. The target algorithm is used to adjust the model parameters of the target model based on the target service combination without adjusting the structure of the target model, so that the target model generates a load balancing strategy that conforms to the constraint model. The constraint model includes at least a strategy objective. The strategy objective is used to characterize that the service rate of the service with a priority less than a preset value among the M services is greater than a preset rate, and the service time variance of the target service combination is less than a preset threshold. This approach combines preliminary adjustments to base station resource distribution based on historical service data prediction with dynamic parameter adjustments. It achieves the goal of quickly adapting to and optimizing model parameters to cope with rapid changes in service demands and sudden load situations without altering the model structure. This results in efficient and proactive base station load balancing, thus solving the technical problem of inefficiency and inability to achieve proactive load management in traditional load balancing algorithms when facing rapid service changes and sudden situations.

[0051] In one alternative embodiment, the intelligent processing system first sets an objective function, which is used to quantify the policy objective, then determines N constraints based on a first model, where N is an integer greater than or equal to 1, the first model is used to characterize the performance indicators of M services, and finally determines a constraint model based on the objective function and the N constraints.

[0052] Optionally, the intelligent processing system defines an objective function whose purpose is to quantify the load balancing strategy objectives for different service combinations within the base station. Specific objectives include maximizing the service rate of the lowest service level while minimizing service time variance to ensure the QoS (Quality of Service) of all services.

[0053] Optionally, the first model refers to the URLLC model (Ultra-Reliable and Low Latency Communications), which is a mathematical model used to describe the characteristics and behavior of ultra-reliable, low-latency communication services. In this embodiment, the URLLC model is used not only to describe high-priority services, but also to characterize the performance indicators and interactions of all M services within the base station.

[0054] Optionally, the intelligent processing system sets N constraints based on the URLLC model, where N represents the number of constraints and is an integer greater than or equal to 1. These constraints may include minimum service rate, maximum latency, minimum packet reliability requirements, etc., to ensure that the service SLA level is met under any load balancing strategy.

[0055] Optionally, based on the predefined objective function and N constraints, the intelligent processing system constructs a constraint model. The constraint model is a mathematical framework for a multi-objective optimization problem; it combines the objective function and constraints to guide the training process of the meta-learning algorithm, ensuring that the final generated load balancing strategy not only maximizes the objective function but also satisfies all preset SLA requirements and performance metrics.

[0056] Optionally, this embodiment models the downlink transmission scenario in a heterogeneous network. All users served by the base station adopt the URLLC model. Within the base station, the service is classified into SLA levels based on latency, packet reliability, and downlink user rate. Different services have different SLA service levels, thereby generating different service combinations.

[0057] Optionally, the set of base stations is represented as Available wireless resources are divided into N wireless resources in the time-frequency domain. Each wireless resource has a bandwidth of B in the frequency domain. Each transmission time interval lasts for 1ms in the time domain. Therefore, there are a total of N available wireless resources in a time slot. The available time slot can be further divided into k smaller units.

[0058] Optionally, an SLA is a formal agreement between a service provider and a tenant, or between service providers themselves, based on which the level of service provided is clearly defined. Each SLA contains a specific number of elements, called metrics, which describe the level and quantity of communication services and measure the performance characteristics of the service objects. This embodiment divides different standards for different URLLC services according to the network SLA metric capability levels. Load balancing for service combinations of different levels is considered, such as a service combination of SLA1 and SLA2. The higher the capability level number, the higher the priority of the service. This embodiment stipulates that when service traffic with a higher SLA metric capability level arrives, the base station will have the probability of delaying ongoing transmissions of lower-level services.

[0059] Optionally, in this embodiment, the users served by the base station uniformly adopt the URLLC model. First, the symbol u is used... s This represents the URLLC service with an SLA level of 's'. Since services with different SLA levels have different priorities, transmitting URLLC traffic with a higher SLA level can affect the rate of ongoing services with a lower SLA level. Therefore, a decision variable is introduced. The service decision for the w-th URLLC service at the current moment is shown in formula (1):

[0060]

[0061] Where b represents the base station number, n represents the current radio resource number, and k represents the current time slot.

[0062] Optionally, define the signal-to-noise ratio (SNR) of the w-th URLLC service. As shown in formula (2):

[0063]

[0064] in, and These represent the transmission power and channel gain of the w-th URLLC service at the current time on the n-th radio resource with base station sequence number b, respectively. and σ represents the transmission power and channel gain of the w-th URLLC service at the current time on the n-th radio resource with a base station sequence number other than b. 2 It is noise power.

[0065] Optionally, to avoid transmission latency, the block size of the URLLC service should be limited. Define the achievable data rate of the URLLC service. The formula is shown in (3):

[0066]

[0067] in, Q represents the number of symbols in each micro-slot. -1 (x) represents the Gaussian inverse cumulative distribution function. The dispersion of the channel determines the randomness of the service channel and can be expressed as formula (4):

[0068]

[0069] Optionally, since different services have different priorities, when a service with a higher service priority arrives, a service with a lower service priority may be delayed, which can affect the system's capacity and reliability. Therefore, considering the service rates and service times of different services, this embodiment establishes a multi-objective optimization problem (strategy objective) to maximize the service rate of the lowest service priority while minimizing the service time variance, and must satisfy constraints such as latency limits, reliability limits, and downlink transmission rates for the corresponding services. For URLLC services, it is assumed that users create small data packet fragments, and the data packets are in micro-time slots within time slot t. The distribution follows a Poisson process. Using the random variable ψ... k (t) represents the number of data packets that arrived, as shown in formula (5):

[0070]

[0071] Where ψ(t) represents the total number of URLLC packets arriving within time slot t. Based on this, the reliability of the URLLC service can be obtained from formula (6):

[0072]

[0073] Where κ represents the packet size of the URLLC service. Equation (6) indicates that the interruption probability of the URLLC service should not exceed the threshold of the corresponding service level. Therefore, the constraint model of multiple URLLC service combinations can be expressed mathematically as follows:

[0074]

[0075] In formula (7), σ(t) represents the service time variance of different services, and the overall objective is to maximize the service rate of the lowest service level while minimizing the service time variance; formula (8) indicates that at any time t, for any user equipment n and base station b, only one service k can be transmitting on any given radio resource w; formula (9) explains the decision variables. The value can only be 0 or 1, which reflects the binary nature of resource allocation: resources are either used or not used, with no intermediate state; Formula (10) ensures the reliability of URLLC and specifies the upper limit of the probability of service interruption; Formula (11) is the transmission rate constraint for the current service level, ensuring the service rate of base station b and user equipment n on the service combination (s,w). The minimum service rate must be met or exceeded as specified in the Service Level Agreement (SLA). s .

[0076] As described above, the intelligent processing system can achieve proactive load balancing among services within a base station in a systematic and mathematical manner. The objective function ensures that the Service Level Agreement (SLA) of all services is optimally considered, while the N constraints provide the minimum service quality threshold for the system, including but not limited to minimum service rate requirements and service time variance limitations. The establishment of the constraint model enables the intelligent processing system to quickly generate and execute load balancing strategies that meet SLA levels and service performance indicators when facing real-time changing service combinations. This method not only improves the transmission efficiency of high-priority services but also takes into account the basic performance requirements of low-priority services, thereby achieving efficient utilization of network resources and improving overall service quality, enhancing the system's adaptability and robustness. In rapidly changing mobile network environments, this method can significantly improve user experience, reduce service interruption rates and latency, and optimize spectrum efficiency and network resource allocation.

[0077] In an optional embodiment, the target model is obtained through the following steps: The intelligent processing system first obtains S historical service combinations, wherein each historical service combination includes multiple historical services, and each historical service includes a corresponding priority. Then, based on the S historical service combinations, the initial model is subjected to multiple target operations until the number of target operations is greater than or equal to the number of the first iteration, thereby obtaining the target model. Each target operation is used to train the initial model based on each historical service combination.

[0078] Optionally, historical service portfolios refer to records of past service requests collected by the intelligent processing system. Each portfolio contains different service types and quantities requested within a specific time window. A historical service is a single service instance that makes up a historical service portfolio, and each service instance is labeled with its service type and priority information.

[0079] Optionally, the initial model refers to a model that has not yet been trained on historical service data. The target operation involves using a historical service combination as input to train the initial model, adjusting its parameters to better predict and handle similar service combinations. The target model, after multiple rounds of training with target operations, is a performance-optimized and parameter-tuned model used to generate a load balancing strategy that meets the constraints of the model.

[0080] Optionally, the intelligent processing system extracts S historical service combinations from the historical database. These combinations record the specific details of service requests within the base station at different times and scenarios in the past, including key information such as service type, service volume, and service priority. Here, S is the number of historical service combinations, and its value is greater than or equal to 1. More historical service combinations mean that the system can learn and extract experience from richer historical scenarios.

[0081] Optionally, the intelligent processing system uses the acquired S historical service combinations to train the initial model in multiple rounds, i.e., target operations. Each target operation involves training the model once using one historical service combination. Through continuous iteration, the model gradually learns how to make optimal load balancing decisions when faced with different service priorities and combinations. The first iteration count is a preset number of training rounds. When the number of target operations reaches or exceeds this preset value, or when all S historical service combinations have been trained, the training process ends, and the system obtains a target model. This model can more accurately predict and respond to service demands when dealing with load balancing issues between services within a base station, while also considering the priorities of different services.

[0082] As described above, iterative training of the model based on historical service data enables the intelligent processing system to generate a highly adaptive and intelligent load balancing target model. The training process of the target model allows the system to fully learn and understand service demand patterns and priority changes in historical service combinations, thus enabling it to quickly make decisions and generate load balancing strategies when faced with new service combinations. The training process based on S historical service combinations also enhances the model's robustness, allowing it to handle various complex service combination scenarios and avoiding performance limitations caused by a single training scenario.

[0083] In one optional embodiment, the target operation includes the following steps: The intelligent processing system first performs multiple target trainings on the initial model based on the i-th historical service combination until a first parameter combination of the initial model is obtained, where i is an integer greater than or equal to 1 and less than or equal to S, and the first parameter combination is used to characterize the model parameters of the initial model under the constrained model; the initial model is updated based on the first parameter combination to obtain a first model; then, based on the first model, a target algorithm is used to determine a second parameter combination, where the second parameter combination is a guiding parameter determined by the target algorithm and used to adjust the model parameters of the first model; finally, a target model is determined based on the second parameter combination and the first model.

[0084] Optionally, the first parameter combination refers to the set of model parameters that, after multiple training iterations, can satisfy the requirements of the constrained model. These parameters are the core of the model, determining its behavior and performance in a specific scenario.

[0085] Optionally, the intelligent processing system reads and analyzes data from the i-th historical service combination. This data may include past service request frequencies, packet sizes, service priorities, and actual resource allocation. The system then uses this data to train an initial model, adjusting the model parameters through backpropagation to minimize prediction errors and optimize the objective function. This training process may involve thousands or millions of iterations until the model parameters stabilize and meet the requirements of the constrained model, thus obtaining the first parameter combination.

[0086] Optionally, the intelligent processing system uses a first parameter combination to update the parameters of the initial model, resulting in a first model that more closely reflects the characteristics of the historical service composition. This update process may involve weight adjustments and bias corrections to ensure that the model can maintain its original structure while exhibiting better prediction and decision-making capabilities. Next, the intelligent processing system utilizes a target algorithm to perform a meta-learning process based on the first model. This process involves rapid iteration, adjusting the model's hyperparameters according to the characteristics of the current target service composition and the requirements of the constrained model, resulting in a second parameter combination. Through this adjustment, the first model will better adapt to the current scenario and generate a more accurate load balancing strategy. Finally, the intelligent processing system uses the second parameter combination to update the parameters of the first model, generating a target model. This model will be used to process the target service composition in real time and dynamically adjust resource allocation according to the requirements of the constrained model to generate the optimal load balancing strategy.

[0087] As shown above, by training with historical service combinations and rapidly tuning the parameters using meta-learning algorithms, the intelligent processing system can quickly generate the optimal load balancing strategy that meets the requirements of the constrained model. This method improves the model's generalization ability and adaptability, enabling the system to maintain efficient and fair resource allocation even under rapidly changing service demands and network environments, meeting the SLA levels of all services, and improving overall network performance and user experience. Specifically, the first parameter combination obtained through training with historical data ensures the model's stability and effectiveness in historical scenarios; while the second parameter combination obtained through meta-learning algorithms enables the model to converge quickly in new scenarios, generating a strategy that meets the constrained model. The entire process not only improves the model's training efficiency but also enhances the model's application flexibility under different service combinations, achieving dynamic optimization of the load balancing strategy, thereby significantly improving the utilization efficiency of network resources and the overall performance of the system.

[0088] In one optional embodiment, the target training includes the following steps: First, the intelligent processing system adjusts the initial parameters of the initial model based on the i-th historical service combination to obtain a third parameter combination. Then, the initial model is updated according to the third parameter combination to obtain a second model. Next, the i-th historical service combination is trained according to the second model to generate a first policy. Finally, the third parameter combination is determined to be the first parameter combination based on the first policy and the constraint model.

[0089] Optionally, the intelligent processing system reads data from the i-th historical service combination, including information such as service request type, frequency, SLA level, and service priority. The system then uses this information to train an initial model, adjusting the initial parameters through machine learning techniques such as backpropagation to obtain a third parameter combination that can better predict and handle the historical service combination. Next, the intelligent processing system uses the third parameter combination to update the parameters of the initial model, generating a second model. This step involves reinitializing certain layers of the model using the third parameter combination, or directly using them to update the corresponding model weights. The updated second model will be used in subsequent training processes to generate a more effective load balancing strategy. The intelligent processing system then uses the second model to train the i-th historical service combination using deep learning or reinforcement learning. During training, the model will attempt to predict and handle different service types and priorities within the service combination to minimize service time variance while maximizing the rate of low-priority services, satisfying the policy objective in the constrained model. After training, the system obtains a first policy, which is the specific performance of the second model on the i-th historical service combination, used to guide resource allocation within the base station. Finally, the intelligent processing system compares the first strategy with the constraint model, checking whether the service rate and service time variance of all services are within the constraints, especially whether the service rate of services with lower SLA levels exceeds the preset value and the service time variance is less than the preset threshold. If the first strategy meets all the requirements of the constraint model, the third parameter combination will be confirmed as the first parameter combination and used for the next model update; otherwise, the system will readjust the third parameter combination based on feedback information until the generated strategy fully meets the constraint model.

[0090] As can be seen from the above, by conducting phased model training and parameter adjustment, the intelligent processing system can automatically generate the optimal load balancing strategy that meets the requirements of the constrained model. This method not only improves the efficiency of model training but also enhances the accuracy of the strategy, effectively handling different service combination scenarios and ensuring the SLA level of all services. It is particularly outstanding in terms of rate optimization of low-priority services and overall system latency control. The overall effect is reflected in the following points: (1) It improves the training speed and quality of the model. Through gradual parameter adjustment and strategy optimization, the model can find the first parameter combination that meets the conditions in fewer training rounds, reducing training time and resource consumption; (2) It realizes the dynamic adaptability of the strategy. Through training on different historical service combinations, the model can learn and understand the changing patterns of service demand and generate strategies that can quickly adapt to new service combinations, improving the flexibility and response speed of the system; (3) It ensures the SLA level. By setting the constrained model, it ensures that the strategy generated by the model can meet the SLA requirements of all services, including key indicators such as service rate, latency, and packet reliability, thereby improving user satisfaction and network service quality.

[0091] In one optional embodiment, the intelligent processing system first processes the i-th historical service combination according to the first strategy to obtain the processing result, wherein the processing result includes at least the performance index of completing the i-th historical service combination, and then determines whether the processing result conforms to the constraint model. If the processing result conforms to the constraint model, the third parameter combination is determined to be the first parameter combination of the initial model.

[0092] Optionally, the intelligent processing system processes the i-th historical service combination based on the first strategy, simulating the entire process of service request arrival, resource allocation, data transmission, and service completion. This process may involve allocating service requests to the radio resources of different base stations, adjusting transmission power, allocating channel bandwidth, and controlling service time variance. After processing, the system records and analyzes the results, including performance indicators such as service completion time, rate, and packet reliability, as well as whether any service interruptions or timeouts occurred. The intelligent processing system then compares the processing results obtained in the previous step with the constraint model, checking whether the results meet the various indicators and constraints specified in the constraint model. If the processing results meet the constraint model, it indicates that the first strategy performs well in processing the i-th historical service combination, achieving the expected service quality and system performance goals. Once the system confirms that the processing results conform to the constraint model, it immediately performs a model parameter update operation. This typically involves assigning the third parameter combination to the parameters of the initial model, making it the first parameter combination, thereby updating the model state and preparing for the processing of subsequent service combinations and strategy generation. The updated model will become the second model, used to handle more complex service combinations to achieve more efficient and intelligent load balancing.

[0093] Optionally, the overall framework of the target model comprises two parts: inner-layer training and outer-layer training. The inner-layer training incorporates reinforcement learning algorithms to train a highly adaptable load balancing mechanism for different business combination scenarios. Through reinforcement learning, the inner-layer model can dynamically adjust its strategy based on specific business combinations, thereby achieving efficient load resource allocation. In the outer-layer training, a meta-learning algorithm guides the inner-layer training. The core function of meta-learning is to quickly adapt to different service combination scenarios by adjusting only the model's hyperparameters while maintaining the model type and structure. This method improves the model's convergence speed in new scenarios, enabling the model to complete optimization in a short time. This framework combines the advantages of reinforcement learning and meta-learning, ensuring the flexibility of the inner-layer model in specific tasks while achieving efficient inter-task transfer through outer-layer meta-learning, providing strong support for complex and ever-changing service combination scenarios.

[0094] Optionally, since services have multiple SLA levels, this embodiment treats different services as multiple agents and proposes a reinforcement learning-based algorithm to train the load balancing model. Reinforcement learning tasks are typically described using Markov Decision Processes (MDPs). An MDP consists of tuples (S, A, p, R, γ), where S is the state, A is the action, p is the state transition probability, R is the reward, and γ is the discount factor. Given a policy π, the cumulative reward G is initially defined. t As shown in formula (12):

[0095]

[0096] Among them, G t R is the cumulative reward after taking an action at time step t, that is, the sum of rewards the agent will receive over all future time steps; t+1 Indicates the next time step t+1 after the action is taken, R t+2 This indicates the second time step t+2, where γ is a discount factor between 0 and 1; It is an infinite series expression that summarizes the reward R that the agent receives at each future time step t+k+1, starting from time step t. t+k+1 And each item is discounted by the corresponding discount factor raised to the power γ. k Weighted.

[0097] Optionally, the value of the current state can be defined using the mathematical expectation of the cumulative revenue, called the value function. Using v... π Let (s) represent the value function of state s under policy π, as shown in formula (13):

[0098]

[0099] Among them, S t =s represents the condition that the current state is s. Value function v π (s) for state s is the sum of expected future returns in this state.

[0100] Optionally, Soft Actor-Critic (SAC) is a reinforcement learning algorithm that combines policy gradient and value function estimation. It is suitable for solving problems in both discrete and continuous action spaces and is an offline policy algorithm. SAC implements the decision-making and learning process through two main components: an actor and a critic. The actor is responsible for generating actions, while the critic evaluates the value of these actions. SAC incorporates the idea of ​​maximum entropy reinforcement learning, aiming to increase policy entropy (i.e., exploratory nature) while maximizing cumulative reward, thus encouraging the agent to be more random in choosing actions and more easily adjust to perturbations.

[0101] Optionally, in the reinforcement learning algorithm used in this embodiment, the agent represents each service, and its state set, action set, reward, and environment are defined as follows. The state space S of a single agent contains the following information: the latency of the current service, the data packet size of the current service, the number of data packets to be transmitted at the current moment, and the service SLA level. The states of all agents can be combined into a global state vector S. globalThis is used to describe the dynamic changes of the entire system. Traffic prediction results based on ensemble learning will be incorporated into the state space, affecting the number of data packets.

[0102] Optionally, the action 'a' of a single agent represents whether the service is transmitting data at the current moment. This can be expressed by the formula shown in (14):

[0103]

[0104] Optionally, since there are N services in total, and each service has two options, the size of the action space A is 2N.

[0105] Optionally, the design goal of the reward function R is to optimize the overall system load balancing performance while meeting the SLA requirements of different services. The specific definition is shown in formula (15):

[0106]

[0107] in, R represents the lowest service rate among all services, and σ(t) is the variance of the service times of various services. By maximizing R, we can minimize the overall system latency while ensuring the basic performance of low-priority services.

[0108] Optionally, the state of the environment changes dynamically over time and is affected by factors such as the request arrival rate of different services and the amount of data being transmitted for the current service. The core functions of the environment include: 1) receiving the action set A of the intelligent agent and updating the system state S. global 2) Calculate the reward R based on the new state; 3) Return to the state S at the next time step. t+1 and reward R t .

[0109] Alternatively, in meta-reinforcement learning, the agent strives to learn a policy with parameter θ that can solve a given distribution. Multiple tasks within it. Each task It is an MDP process, consisting of its input s i Output a i loss function L i The transfer function P i Reward function R i and round length H i Typically, meta-reinforcement learning methods involve two steps: a meta-training process and a meta-testing process. During the meta-training process, the agent trains on a set of meta-training tasks. Training is conducted on the platform, and during the meta-testing process, the agent performs a set of test tasks. The evaluation is performed on the same dataset. It is typically assumed that the training and test sets come from the same distribution. but It can be different Each task Both training and validation sets are available. For each task, the objective is to start from the parameters θ and utilize the training set. Learn parameters specific to this task. Make the validation set Loss L i Minimize. Alg(.) represents the algorithm used to update parameters for a specific task.

[0110] Optionally, this embodiment focuses on gradient-based meta-reinforcement learning methods, such as MAML (Model-Agnostic Meta-Learning). The meta-training phase of such methods can be viewed as a two-layer optimization problem, the goal of which is to learn the optimal meta-parameters θ such that:

[0111]

[0112] Formula (16) represents the search for a set of parameters. This set of parameters makes the function F(θ) reach its minimum value, and F(θ) is represented as shown in (17):

[0113]

[0114] Alternatively, in MAML, inner-layer optimization is addressed through one (or more) gradient descent steps:

[0115]

[0116] Where, θ i This represents a specific task. Model parameters; This is the formulaic expression of the gradient descent update rule, which illustrates how to adjust the parameter θ for a specific task from the meta-parameter θ. i θ is the initial parameter; β is the step size of the inner optimization. The loss function L i The gradient with respect to parameter θ points in the direction in which the loss function value grows the fastest.

[0117] Optionally, for the multi-objective load balancing problem in this embodiment, each task It is a specific weight vector ω iThe corresponding Markov decision process (meaning that each load balancing task, such as resource allocation under different business combinations, can be regarded as an MDP with state S, action a, reward r, and state transition probability p). The learning problem represented by formula (18) is solved by alternately executing two optimization steps: 1) Task adaptation (inner optimization): Starting from the meta-policy parameter θ, learn multiple task-specific policies (at this stage, the algorithm uses the pre-trained meta-policy parameter θ as the starting point, and adjusts the policy parameters through inner optimization according to the characteristics and feedback (reward) of each task to form a policy for each task). The specific strategy is equivalent to performing local policy optimization on each task to better serve the goal of that task; 2) Meta-adaptation (outer layer optimization): Adjusting the meta-parameters using trajectories sampled from the adapted policy (after inner layer optimization, for each task, a set of adapted policy parameters is obtained. Next, the algorithm will use these policies to sample a series of trajectories (state-action sequences) in the environment and update the meta-policy parameters θ through the feedback (such as rewards) of these trajectories. This process is achieved by calculating the policy gradient or other optimization methods, with the aim of enabling the meta-policy parameters to better initialize the learning of future tasks). These two steps are repeated a fixed number of meta-iterations N. meta Once training is complete, the resulting meta-policy can be used as initialization parameters to quickly learn the optimal solution for new tasks. This means that when encountering an unfamiliar load balancing task, the model does not need to be trained from scratch, but can start from the pre-trained meta-policy parameters and achieve good performance with only a small number of updates, which greatly saves the adaptation cost and time for new tasks.

[0118] Optionally, the main process of the meta-learning-based load balancing algorithm (target model) specifically employs the MAML framework to quickly learn load balancing strategies that adapt to different business scenarios. The following is a detailed explanation of each step:

[0119] Step 1 Input: This part defines the input parameters required when the algorithm starts, including: preference distribution p(ω), number of meta-iterations N. meta The number of tasks N in each meta-iteration taskThe number of trajectories sampled for each task is M; where the preference distribution represents the probability of occurrence of different service combinations or business scenarios. In this algorithm, it is used to sample weight vectors for training on different tasks; the number of meta-iterations indicates how many rounds of meta-learning iterations the algorithm will perform; the number of tasks in each meta-iteration indicates how many independent tasks the algorithm will process in each meta-iteration, which come from different business combinations or scenarios; and the number of trajectories sampled for each task indicates how many trajectory data generated by the agent performing actions will be sampled for each task to update policy parameters.

[0120] Step 2: Randomly initialize the meta-policy π θ At the start of the algorithm, the model's parameters θ need to be randomly initialized. Meta-policy π θ This refers to an initial policy function whose parameter θ will be optimized through a subsequent meta-learning process.

[0121] Step 3: Iterate through t = 0, 1, ..., N meta :

[0122] Step 3.1 Sample N preference vectors ω i p(ω): Sample N from the preference distribution task Each preference vector represents a characteristic of each task, such as service composition or business scenario.

[0123] Step 3.2 Iterate through each weight vector ω i :

[0124] Step 3.2.1 Based on π θ Sample M trajectories and denote them as Using the current meta-policy π θ In a simulated load-balanced environment, (M) actions are performed, generating (M) trajectory data. The trajectory data contains information about the agent's state, actions, rewards, and new states when performing actions in the environment, which will be used for parameter adjustment.

[0125] Step 3.2.2 Use Estimate Based on sampled trajectory data Calculate the gradient of the policy on the current task. Where L i It is the loss function for task i.

[0126] Step 3.2.3 Calculate parameters Using the calculated gradient, the policy parameters θ are updated with a learning rate β to obtain the specific parameters θ for task i. i .

[0127] Step 3.2.4 According to the strategy Collection trajectory Use the updated parameter θ i In the same business scenario, generate another set of verification trajectory data. This data will be used to evaluate the policy's performance on the task and to update it through meta-learning.

[0128] Step 3.3 Update Strategy After completing inner-layer training and validation for all tasks, the gradient information from all tasks is used to update the parameters θ of the meta-policy via the learning rate η of meta-learning. This is the key to meta-learning, which enables the model to learn shareable patterns from multiple tasks, thereby accelerating adaptation to new tasks.

[0129] Step 4 of the algorithm ends. Complete N meta After one iteration, the algorithm ends, and the meta-policy parameter θ should have been optimized to a level that can quickly adapt to new tasks.

[0130] As shown above, through practical strategy application and result verification, the intelligent processing system can ensure that it generates the optimal load balancing strategy that satisfies the constraint model. This method improves the practicality and performance reliability of the strategy. Through simulation and verification of historical service combinations, the system can accurately evaluate the effectiveness of the strategy, ensuring that its application in real-world scenarios does not violate Service Level Agreements (SLAs) and network resource usage constraints.

[0131] In one optional embodiment, the intelligent processing system allocates resources to the target service combination according to a load balancing strategy to obtain the allocation result, then executes the target service combination based on the allocation result to obtain the execution result, wherein the execution result includes at least the performance indicators of completing the target service combination, and then updates the target model based on the execution result.

[0132] Optionally, the intelligent processing system reads the load balancing strategy generated by the target model and allocates resources to each service in the target service combination according to the guiding principles in the strategy. For example, the system may prioritize allocating more wireless resources to high-priority services, while allocating resources to low-priority services only to meet their basic performance requirements. The allocation results will record in detail the type, quantity, and time of resources allocated to each service, as well as the expected service performance indicators. Then, the intelligent processing system executes the target service combination according to the allocation results, monitoring the performance during service execution in real time, including data transmission rate, latency, packet reliability, and service time variance. After service execution is completed, the system will collect the execution performance data of all services to form an execution result. This result will be used to evaluate the effectiveness of the load balancing strategy and fed back to the target model for learning and adjustment. Finally, the intelligent processing system analyzes the execution result, evaluates the difference between the actual service performance and the expected target, and then uses machine learning techniques such as backpropagation to update the parameters of the target model using the execution result. This update process may involve multiple iterations until the model can generate a strategy that is closer to the expected execution result in similar scenarios. The updated model is the latest target model, which can more accurately predict service demand and generate effective resource allocation strategies.

[0133] As can be seen from the above, the intelligent processing system achieves proactive load balancing of services within the base station through a closed-loop process of strategy generation, execution, and model update, thereby optimizing resource allocation efficiency and improving service quality and user experience.

[0134] This application also provides a base station load balancing device. It should be noted that the base station load balancing device of this application can be used to execute the base station load balancing method provided in this application. The base station load balancing device provided in this application is described below.

[0135] According to an embodiment of this application, an apparatus for implementing the above-described base station load balancing method is also provided. Figure 2 This is a schematic diagram of an optional base station load balancing device according to an embodiment of this application, such as... Figure 2 As shown, the device includes: a prediction unit 201, an adjustment unit 202, a receiving unit 203, and a generation unit 204.

[0136] Optionally, the prediction unit 201 is used to predict the distribution characteristics of network service demand for the target scenario within a target time period based on historical service data; the adjustment unit 202 is used to adjust the resource distribution of the base station according to the distribution characteristics; the receiving unit 203 is used to receive the target service combination within the target time period, wherein the target service combination includes M services and the priority of each service, where M is an integer greater than or equal to 1; the generation unit 204 is used to adjust the target model based on the resource distribution of the base station, the target service combination, and the target algorithm, and generate a load balancing strategy for the target service combination based on the adjusted target model, wherein the target algorithm is used to adjust the model parameters of the target model based on the target service combination without adjusting the structure of the target model, so that the target model generates a load balancing strategy that conforms to the constraint model, wherein the constraint model includes at least a strategy objective, wherein the strategy objective is used to characterize that the service rate of the service with a priority less than a preset value among the M services is greater than a preset rate, and the service time variance of the target service combination is less than a preset threshold.

[0137] Optionally, the generation unit 204 includes: a first setting subunit, a first determining subunit, and a second determining subunit. The first setting subunit is used to set an objective function, wherein the objective function is used to quantify the strategy objective; the first determining subunit is used to determine N constraints based on a first model, wherein N is an integer greater than or equal to 1, and the first model is used to characterize the performance indicators of M services; the second determining subunit is used to determine a constraint model based on the objective function and the N constraints.

[0138] Optionally, the generation unit 204 includes: a first acquisition subunit and a first processing subunit. The first acquisition subunit is used to acquire S historical service combinations, wherein each historical service combination includes multiple historical services, and each historical service includes a corresponding priority; the first processing subunit is used to perform multiple target operations on the initial model based on the S historical service combinations until the number of target operations is greater than or equal to the number of first iterations, thereby obtaining the target model, wherein each target operation is used to train the initial model based on each historical service combination.

[0139] Optionally, the first processing subunit includes: a first training module, a first update module, a first determination module, and a second determination module. The first training module is used to perform multiple target trainings on the initial model based on the i-th historical service combination until a first parameter combination of the initial model is obtained, where i is an integer greater than or equal to 1 and less than or equal to S, and the first parameter combination is used to characterize the model parameters of the initial model under the constrained model. The first update module is used to update the initial model based on the first parameter combination to obtain the first model. The first determination module is used to determine a second parameter combination based on the first model using a target algorithm, where the second parameter combination is a guiding parameter determined by the target algorithm and used to adjust the model parameters of the first model. The second determination module is used to determine the target model based on the second parameter combination and the first model.

[0140] Optionally, the first training module includes: a first adjustment submodule, a first update submodule, a first generation submodule, and a first determination submodule. The first adjustment submodule is used to adjust the initial parameters of the initial model based on the i-th historical service combination to obtain a third parameter combination; the first update submodule is used to update the initial model according to the third parameter combination to obtain a second model; the first generation submodule is used to train the i-th historical service combination based on the second model to generate a first policy; and the first determination submodule is used to determine whether the third parameter combination is the first parameter combination based on the first policy and the constraint model.

[0141] Optionally, the first determining submodule includes: a first processing component, a first judgment component, and a first determining component. The first processing component is used to process the i-th historical service combination according to a first strategy to obtain a processing result, wherein the processing result includes at least the performance indicators for completing the i-th historical service combination; the first judgment component is used to determine whether the processing result conforms to the constraint model; and the first determining component is used to determine that, if the processing result conforms to the constraint model, the third parameter combination is the first parameter combination of the initial model.

[0142] Optionally, the base station load balancing device further includes: an allocation unit, an execution unit, and an update unit. The allocation unit is used to allocate resources to the target service combination according to the load balancing strategy to obtain an allocation result; the execution unit is used to execute the target service combination based on the allocation result to obtain an execution result, wherein the execution result includes at least the performance indicators for completing the target service combination; and the update unit is used to update the target model according to the execution result.

[0143] According to another aspect of this application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the above-described base station load balancing method.

[0144] According to another aspect of this application, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to perform the base station load balancing method described above.

[0145] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0146] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0147] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0148] The units described as separate components may or may not be physically separate. 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 units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0149] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0150] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0151] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A base station load balancing method, characterized in that, include: Predict the distribution characteristics of network service demand for the target scenario within the target time period based on historical service data; Adjust the resource distribution of the base station according to the aforementioned distribution characteristics; Receive a target service combination within the target time period, wherein the target service combination includes M services and the priority of each service, where M is an integer greater than or equal to 1; Based on the resource distribution of the base station, the target model is adjusted according to the target service combination and the target algorithm, and a load balancing strategy for the target service combination is generated based on the adjusted target model. The target algorithm is used to adjust the model parameters of the target model based on the target service combination without adjusting the structure of the target model, so that the target model generates a load balancing strategy that conforms to the constraint model. The constraint model includes at least a strategy objective, which is used to characterize that the service rate of the service with a priority lower than a preset value among the M services is greater than a preset rate, and the service time variance of the target service combination is less than a preset threshold. The target model is obtained through the following steps: obtaining S historical service combinations, wherein each historical service combination includes multiple historical services, and each historical service includes a corresponding priority; performing multiple target operations on the initial model based on the S historical service combinations until the number of target operations is greater than or equal to the number of first iterations, thereby obtaining the target model, wherein each target operation is used to train the initial model based on each historical service combination, and the target operation includes the following steps: performing multiple target training operations on the initial model based on the i-th historical service combination until a first parameter combination of the initial model is obtained, wherein i is an integer greater than or equal to 1 and less than or equal to S, and the first parameter combination is used to characterize the model parameters of the initial model under the constraint model; updating the initial model based on the first parameter combination to obtain a first model; determining a second parameter combination based on the first model using the target algorithm, wherein the second parameter combination is a guiding parameter determined by the target algorithm and is used to adjust the model parameters of the first model; and determining the target model based on the second parameter combination and the first model.

2. The base station load balancing method according to claim 1, characterized in that, The constraint model is obtained through the following steps: Set an objective function, wherein the objective function is used to quantify the policy objective; N constraints are determined based on the first model, where N is an integer greater than or equal to 1, and the first model is used to characterize the performance metrics of the M services. The constraint model is determined based on the objective function and the N constraints.

3. The base station load balancing method according to claim 1, characterized in that, The target training includes the following steps: The initial parameters of the initial model are adjusted based on the i-th historical service combination to obtain the third parameter combination; The initial model is updated based on the combination of the third parameters to obtain the second model; The first strategy is generated by training the i-th historical service combination according to the second model; Based on the first strategy and the constraint model, determine whether the third parameter combination is the same as the first parameter combination.

4. The base station load balancing method according to claim 3, characterized in that, Determining whether the third parameter combination is the same as the first parameter combination based on the first strategy and the constraint model includes: The first strategy is used to process the i-th historical service combination to obtain a processing result, wherein the processing result includes at least the performance metrics for completing the i-th historical service combination. Determine whether the processing result conforms to the constraint model; Given the processing result and the constraint model, the third parameter combination is determined to be the first parameter combination of the initial model.

5. The base station load balancing method according to claim 1, characterized in that, After adjusting the target model based on the target service combination and generating a load balancing strategy for the target service combination based on the adjusted target model, the method further includes: Based on the load balancing strategy, resources are allocated to the target service combination to obtain the allocation result; Based on the allocation result, the target service combination is executed to obtain the execution result, wherein the execution result includes at least the performance indicators for completing the target service combination; Update the target model based on the execution results.

6. A base station load balancing device, characterized in that, include: The prediction unit is used to predict the distribution characteristics of network service demand for a target scenario within a target time period based on historical service data. An adjustment unit is used to adjust the resource distribution of the base station according to the distribution characteristics; A receiving unit is configured to receive a target service combination within the target time period, wherein the target service combination includes M services and the priority of each service, wherein M is an integer greater than or equal to 1; A generation unit is configured to adjust a target model based on the resource distribution of the base station, according to the target service combination and the target algorithm, and generate a load balancing strategy for the target service combination based on the adjusted target model. The target algorithm is configured to adjust the model parameters of the target model based on the target service combination without adjusting the structure of the target model, so that the target model generates a load balancing strategy that conforms to a constraint model. The constraint model includes at least a policy objective, which characterizes that the service rate of a service with a priority lower than a preset value among the M services is greater than a preset rate, and the service time variance of the target service combination is less than a preset threshold. The generation unit includes: a first acquisition subunit, configured to acquire S historical service combinations, wherein each historical service combination includes multiple historical services, and each historical service includes a corresponding priority; and a first processing subunit, configured to perform multiple target operations on an initial model based on the S historical service combinations until the number of target operations is greater than or equal to the number of first iterations, thereby obtaining the target model, wherein each target operation is used to train the initial model based on each historical service combination. The first processing subunit includes a first training module, configured to perform multiple target trainings on the initial model based on the i-th historical service combination until a first parameter combination of the initial model is obtained, where i is an integer greater than or equal to 1 and less than or equal to S, and the first parameter combination is used to characterize the model parameters of the initial model under the constraint model; a first update module, configured to update the initial model based on the first parameter combination to obtain a first model; a first determination module, configured to update the initial model based on the first parameter combination to obtain a first model; and a second parameter combination, configured to determine a second parameter combination based on the first model using the target algorithm, wherein the second parameter combination is a guiding parameter determined by the target algorithm and used to adjust the model parameters of the first model; and a second determination module, configured to determine the target model based on the second parameter combination and the first model.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, wherein, when the executable program is executed, it controls the device containing the computer-readable storage medium to perform the base station load balancing method according to any one of claims 1 to 5.

8. An electronic device, characterized in that, It includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to perform the base station load balancing method according to any one of claims 1 to 5.

Images