WO2023124469A1

WO2023124469A1 - Device energy saving method and system, electronic device, and storage medium

Info

Publication number: WO2023124469A1
Application number: PCT/CN2022/127479
Authority: WO
Inventors: 刘梅红
Original assignee: 中兴通讯股份有限公司
Priority date: 2021-12-28
Filing date: 2022-10-25
Publication date: 2023-07-06
Also published as: CN116405990A

Abstract

Disclosed in embodiments of the present application are a device energy saving method and system, an electronic device, and a storage medium. The device energy saving method is applied to a base station, and comprises: obtaining historical service data of the base station; dynamically predicting an energy saving threshold of the base station according to the historical service data and a preset prediction model; obtaining the current service data of the base station; and controlling an energy saving operation of the base station according to the current service data and the energy saving threshold.

Description

Equipment energy saving method, system, electronic equipment and storage medium

related application

This application claims priority to a Chinese patent application with application number 202111681339.9 filed on December 28, 2021, the entire contents of which are incorporated herein by reference.

technical field

The embodiments of the present application relate to the communication field, and in particular, to a device energy saving method, system, electronic device, and storage medium.

Background technique

From 4G to 5G, mobile communication technology and product forms have undergone great changes. Wireless networks support higher speeds, lower latency, and greater connection density to meet the new challenges brought by the digital ecosystem to the network. Compared with 4G, 5G base stations support special requirements such as large bandwidth, multi-channel, and multi-antenna. The hardware processing capability is greatly improved, the equipment architecture and functions are more complex, and the power consumption is also greatly increased.

With the explosive growth of wireless network users, the investment scale of operators' equipment will also increase accordingly, and the energy saving of base stations is facing greater challenges. In the energy-saving method of traditional wireless networks, it is necessary to manually analyze massive data, and manually set unified shutdown parameters and energy-saving period configurations.

However, when the traffic volume becomes large, the service is damaged due to unreasonable shutdown parameter settings; and when the traffic volume becomes small, the energy-saving effect is not good due to unreasonable shutdown parameter settings.

Contents of the invention

The main purpose of the embodiment of the present application is to provide a device energy saving method, system, electronic device and storage medium, which can improve the flexibility of automatic energy saving of the device.

In order to achieve the above purpose, an embodiment of the present application provides a method for energy saving of equipment, which is applied to a base station, including: obtaining historical service data of the base station; dynamically predicting the energy saving threshold of the base station according to the historical service data and a preset prediction model; obtaining the energy saving threshold of the base station The current service data; according to the current service data and the energy-saving threshold, control the energy-saving operation of the base station.

In order to achieve the above purpose, the embodiment of the present application also provides a device energy saving method, which is applied to the management server, including: obtaining the reported business data of the base station; training the prediction model according to the reported business data; downloading the trained prediction model to sent to the base station; wherein, the prediction model is used to predict the energy-saving threshold of the base station, so that the base station controls the energy-saving operation according to the energy-saving threshold.

In order to achieve the above purpose, an embodiment of the present application also provides a device energy saving system, including a base station and a management server; wherein, the base station is used to obtain historical service data of the base station, and dynamically predict The energy-saving threshold of the base station obtains the current service data of the base station, and controls the energy-saving operation of the base station according to the current service data and the energy-saving threshold; the management server is used to obtain the reported service data of the base station, and train the prediction model according to the reported service data. The trained prediction model is sent to the base station, wherein the prediction model is used to predict the energy-saving threshold of the base station, so that the base station can control the energy-saving operation according to the energy-saving threshold.

To achieve the above purpose, an embodiment of the present application also provides an electronic device, including: at least one processor; and a memory connected to the at least one processor in communication; wherein, the memory stores information that can be executed by the at least one processor. Instructions, the instructions are executed by at least one processor, so that the at least one processor can execute the above-mentioned device energy saving method.

In order to achieve the above object, the embodiment of the present application also provides a computer-readable storage medium storing a computer program, and implementing the above device energy saving method when the computer program is executed by a processor.

The equipment energy-saving method proposed in this application obtains the historical service data of the base station, dynamically predicts the energy-saving threshold of the base station according to the historical service data and the preset prediction model, obtains the current service data of the base station, and controls the For the energy-saving operation of the base station, since the energy-saving threshold is dynamically predicted by the historical service data prediction model of the base station, the energy-saving threshold is in line with the service flow characteristics of the base station. Referring to this energy-saving threshold for energy-saving operation control can make the energy saving of the base station The behavior is more in line with the business characteristics and needs of the base station, and avoids the service performance degradation or poor energy saving effect of the base station due to energy saving, so the flexibility of automatic energy saving of equipment can be improved.

Description of drawings

FIG. 1 is a schematic flowchart of applying a device energy saving method provided by an embodiment of the present application to a base station;

FIG. 2 is a schematic diagram of device deployment of a device energy saving method provided by an embodiment of the present application;

Fig. 3 is a schematic diagram of deployment of a prediction model provided by an embodiment of the present application;

Fig. 4 is a schematic diagram of a flow chart of predictive model activation provided by an embodiment of the present application;

FIG. 5 is a schematic diagram of a forecasting model upgrade process provided by an embodiment of the present application;

FIG. 6 is a schematic flowchart of applying a device energy saving method provided by an embodiment of the present application to a management server;

Fig. 7 is a schematic diagram of an equipment energy-saving system provided by an embodiment of the present application;

Fig. 8 is a schematic structural diagram of an electronic device provided by an embodiment of the present application.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the embodiments of the present application will be described in detail below with reference to the accompanying drawings. However, those of ordinary skill in the art can understand that in each embodiment of the application, many technical details are provided for readers to better understand the application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in this application can also be realized. The division of the following embodiments is for the convenience of description, and should not constitute any limitation to the specific implementation of the present application, and the embodiments can be combined and referred to each other on the premise of no contradiction.

The embodiment of the present application relates to a method for energy saving of equipment, as shown in FIG. 1 , including the following steps:

Step 101, obtaining historical service data of the base station;

Step 102, dynamically predicting the energy-saving threshold of the base station according to historical service data and a preset prediction model;

Step 103, acquiring current service data of the base station;

Step 104: Control the energy-saving operation of the base station according to the current service data and the energy-saving threshold.

The equipment energy-saving method of this application is applied to base station equipment. In the entire base station system, it is mainly composed of wireless equipment, transmission equipment and power supply system; wherein, the wireless equipment includes an indoor baseband processing unit (Building Base band Unit, referred to as "BBU") and a remote radio unit (Remote Radio Unit, referred to as "RRU"), which respectively complete the uplink and downlink processing of baseband signals and radio frequency signals. In the deployment of distributed base stations, the BBU and RRU are deployed separately; the BBU is deployed in the equipment room, and the RRU is deployed on the tower, and the BBU and RRU are connected to each other through optical fibers. The Operation and Maintenance Center (OMC) manages the server to realize the unified management of wireless network devices.

According to statistics, the cost of energy consumption of base stations accounts for about 16% of network operating expenses (Operating Expense, referred to as "OPEX"), and there is a further increase trend. Therefore, effectively reducing base station energy consumption and improving network energy efficiency has become an important concern of global operators.

If the energy consumption of the base station is not dynamically adjusted according to the change of the traffic volume, the energy consumption of the wireless equipment will be wasted. In traditional wireless network energy consumption control methods, it is necessary to manually analyze massive data, including network configuration data, site coverage data, multi-frequency multi-standard network identification, etc., and often manually set unified shutdown parameters and energy-saving time configuration. In order to reduce operating costs, there is no difference in the configuration of energy-saving parameters, and it is impossible to automatically find the best configuration parameters according to the scene; when the traffic volume increases, the business is damaged due to unreasonable parameter settings; The setting is unreasonable and the energy saving effect is not good. In order to reduce the cost of energy-saving operation and maintenance, finding intelligent energy-saving strategies to save energy and reduce emissions for operators, and reduce network operation and maintenance costs has become one of the important concerns of major network equipment providers.

There are also some studies in the industry that combine machine learning models with base station energy saving. The energy-saving goal is to find the energy-saving time period, which itself is related to time. The application of time series model in telecommunications mostly adopts the traditional time series model, which was produced in the era of data scarcity, the accuracy and prediction effect are not good, and the engineering implementation is prone to excessive deviation. With the advancement of deep learning technology and the improvement of data processing capabilities and the growth of data, models with stronger generalization capabilities can be designed. However, the computing resources of the base station are limited, and the model design needs to find a balance between complexity and computing power; if the model is simple, the deviation will be too large, and the energy-saving prediction effect will not be good; if the model is complex, due to the large amount of calculation, result in additional power consumption costs. Moreover, the model interpretation of the machine learning model is not strong, and the prediction dimension is single.

In the traditional wireless network energy consumption control method, it is necessary to manually analyze a large amount of massive data, including network configuration data, site coverage data, multi-frequency multi-standard network identification, etc., and manually set the configuration parameters of unified shutdown: energy-saving switch, PRB utilization rate Threshold (uplink and downlink), user access threshold; and energy-saving period configuration: energy-saving period on weekdays, and energy-saving period on weekends, specific to the hour.

However, static configuration is based on a large amount of operation and maintenance experience, and often adopts the most conservative energy-saving strategy, and there is a bottleneck in the energy-saving effect. In addition, static configuration cannot predict the user access volume in emergencies and holidays. In this scenario, the energy-saving switch often needs to be manually turned off, resulting in high manual operation and maintenance costs.

In this application, by obtaining the historical business data of the base station, according to the historical business data and the preset prediction model, dynamically predict the energy saving threshold of the base station, obtain the current business data of the base station, and control the energy saving of the base station according to the current business data and the energy saving threshold Operation, since the energy-saving threshold is dynamically predicted by the historical service data prediction model of the base station, so the energy-saving threshold is in line with the traffic characteristics of the base station. Referring to this energy-saving threshold for energy-saving operation control can make the energy-saving behavior of the base station more in line with The business characteristics and requirements of the base station can avoid the service performance degradation or poor energy saving effect of the base station due to energy saving, so the flexibility of automatic energy saving of equipment can be improved.

The implementation details of the device energy-saving method in this embodiment are described in detail below, and the following content is only implementation details provided for easy understanding, and is not necessary for implementing this solution.

In step 101, the base station acquires historical service data of the base station.

Wherein, the base station periodically collects service data of the device at time intervals, or collects service data centrally at a time point. For massive business data, it is necessary to perform feature selection on the data. These features include: base station ID, configuration information such as time, serving cell, and neighbor cell relationship, as well as KPI data such as user access volume, and data such as emergencies and holiday events. Specifically, the energy consumption data of a base station can be expressed as the following triplet information: {site_id, date, fetures}

Among them, site_id represents the base station identification, date represents the energy consumption data sampling time stamp, and features represents the sampling feature vector of the base station energy consumption data, namely: features={serving cell, neighbor cell relationship, user access volume, historical physical resource block (Physical Resource Block, referred to as "PRB") utilization rate, holiday events, ...}

In step 102, the base station dynamically predicts the energy saving threshold of the base station according to historical service data and a preset prediction model.

The base station is preset with a pre-trained prediction model issued by the management server, and the base station inputs historical service data into the prediction model to obtain the predicted value of the energy saving threshold output by the prediction model.

In one example, the prediction model is established through any one of the following dimensions, or any combination of dimensions: change trends, cycle seasons, holidays, and major events.

In this embodiment, since the prediction model can be established according to any one or any combination of dimensions of change trends, periodic seasons, holidays, and major events, the energy-saving threshold predicted by the prediction model is in line with the change trend of base station services, periodic seasons The business characteristics of any one or any combination of dimensions, holidays, and major events can improve the accuracy of energy-saving thresholds and enable devices to have better energy-saving effects.

In one example, the energy-saving threshold of the base station within a preset period is predicted according to the historical service data and the preset prediction model, so as to dynamically predict the energy-saving threshold of the base station according to the historical service data and the preset prediction model. Wherein, the preset cycle duration may be one day, one week, one month and so on.

In this embodiment, since the base station only predicts the energy-saving threshold within the preset period each time, the base station needs to perform a prediction of the energy-saving threshold in each preset period, that is, the base station will divide the working time of the device into Multiple cycles, and the prediction of the energy-saving threshold is performed in each cycle, which increases the number of predictions and reduces the number of energy-saving thresholds calculated for each prediction, thereby reducing the calculation pressure of the device.

In an example, predicting the energy saving threshold of the base station within the preset period includes: predicting the energy saving threshold of the base station corresponding to each time period within the preset period.

In this embodiment, since the preset cycle is divided into multiple time periods, and each time period has a corresponding energy saving threshold, the calculated energy saving threshold can fit the service characteristics of each time period respectively, and the energy saving threshold can be made real-time. Therefore, the base station will use different energy-saving thresholds for energy-saving control in each time period within the preset period, so as to have a better energy-saving effect.

In step 103, the base station acquires current service data of the base station. Wherein, the base station detects and acquires the service data of the device in real time after acquiring the energy-saving threshold prediction value output by the prediction model.

In step 104, the base station controls the energy-saving operation of the base station according to the current service data and the energy-saving threshold. The base station compares the current service data with the energy-saving threshold, and controls the base station to enter the energy-saving state when the current service data is greater than the energy-saving threshold, and controls the base station to exit the energy-saving state when the current service data is greater than or equal to the energy-saving threshold.

In an example, the energy saving threshold may be a resource utilization threshold of a physical resource block PRB and a threshold of the number of access users;

The base station obtains the current service data of the base station in the following manner: obtains the current PRB resource utilization rate and the number of current access users of the base station.

The base station controls the energy-saving behavior of the base station according to the current service data and the energy-saving threshold in the following way: when the current PRB resource utilization rate is less than the PRB resource utilization rate threshold, and the current number of access users is less than the access user number threshold, control the base station Enter the energy-saving state; when the current PRB resource utilization rate is greater than or equal to the PRB resource utilization rate threshold, or the current number of access users is greater than or equal to the access user number threshold, control the base station to exit the energy-saving state.

In this embodiment, since the energy saving threshold includes the PRB resource utilization threshold of the physical resource block and the access user number threshold, the current PRB resource utilization rate is less than the PRB resource utilization rate threshold, and the current access user number is less than the access user number In the case of the threshold, the base station is controlled to enter the energy-saving state. When the current PRB resource utilization rate is greater than or equal to the PRB resource utilization rate threshold, or the current number of access users is greater than or equal to the threshold of the number of access users, the base station is controlled to exit the energy-saving state. Realize the control of the energy-saving behavior of the base station, and improve the energy-saving performance of the base station equipment.

Further, the PRB resource utilization threshold includes: uplink PRB resource utilization and downlink PRB resource utilization.

When the current PRB resource utilization rate is less than the PRB resource utilization rate threshold, and the current number of access users is less than the access user number threshold, control the base station to perform an energy-saving state, including any one or any combination of the following:

When the current moment is the downlink, and the current downlink PRB resource utilization rate is less than the downlink PRB resource utilization rate threshold, and the current number of access users is less than the access user number threshold, the symbol is turned off;

When the current uplink PRB resource utilization rate is less than the uplink PRB resource utilization rate threshold, and the current number of access users is less than the access user number threshold, turn off the power amplifier of the uplink channel;

When the current downlink PRB resource utilization rate is less than the downlink PRB resource utilization rate threshold, and the current number of access users is less than the access user number threshold, turn off the power amplifier of the downlink channel;

When the current PRB resource utilization rate of the current carrier is less than the PRB resource utilization rate threshold and the current number of access users is less than the access user number threshold, the power amplifier of the current carrier is turned off.

When the base station enters the energy-saving state, the OMC can monitor the network status of the base station in real time. When the traffic volume of the base station is higher than the corresponding threshold within a certain time period, the energy-saving state is turned off. In addition, OMC monitors network KPIs in real time, that is, network operation indicators. For example, by observing data in dimensions such as access, handover, call drop, and rate, the performance data declines. The decline not only depends on the trend but also depends on the magnitude and type, and judges whether to cover loopholes, etc. , that is, to analyze the reasons for the decline based on business experience to see whether to turn off the main energy-saving switch. When the network performance declines at the site, turn off the energy-saving switch of the energy-saving base station and restore the power amplifier of the base station.

In one example, the equipment energy-saving method is based on the intelligent energy-saving method of machine learning, which fully mines the rules of network operation and maintenance data, automatically finds the control threshold of energy-saving period through the machine learning model, and fully considers the trend prediction of the control threshold and seasonality. The periodic control threshold prediction and the burst control threshold prediction of holidays make the threshold control match the real business volume and achieve the best energy-saving control strategy. The deployment of the base station and the management server is shown in FIG. 2 .

In order to realize the intelligent energy saving of wireless devices, it is necessary to implement a data distributed computing platform between traditional network elements and OMCs to realize effective collection, processing and computing of data. In addition, OMC should provide a general machine learning training platform to complete machine learning training calculations, machine model deployment and upgrade functions.

The OMC deploys the trained model on the BBU. The BBU can support the inference calculation of the model through proprietary hardware, complete the prediction of the model at runtime, and output the prediction of the energy saving threshold, including the energy saving of the uplink and downlink PRB utilization and the number of user accesses. threshold. According to the energy-saving threshold predicted by the model, the BBU starts the energy-saving assessment, finally completes the formulation of the energy-saving strategy, and notifies the RRU to start the control of the corresponding energy-saving behavior.

This application combines the current machine learning and wireless network intelligent operation and maintenance methods, and proposes a base station intelligent energy-saving workflow method to ensure the process, automation and engineering of intelligent operation and maintenance. The entire workflow follows the following stages:

1. Model estimation stage: According to energy-saving business scenarios and base station configuration models, implement energy-saving scheme estimation and scheme design; collect base station operation and maintenance data based on this, complete data labeling, and divide the data set into training set and test set.

2. Model design stage: design a time series-based prediction model, and complete the model design process of uplink and downlink PRB utilization and the number of user accesses. In the model design phase, the training parameters and hyperparameters are determined.

3. Model verification stage: In the experimental network environment, load the model and complete the evaluation of energy-saving effect and plan correction; when the network performance is not good, it is necessary to intervene in manual analysis and model upgrade.

4. Model tuning stage: establish an end-to-end feedback mechanism and visualization environment, continuously collect network KPI data and energy-saving effect data, continuously improve model accuracy, and improve energy-saving effect.

Wherein, the control threshold includes a PRB utilization threshold (uplink and downlink), and a user access threshold. Through the machine learning model, the control threshold value of the current base station is automatically predicted, and automatically configured to the energy-saving strategy controller. The energy-saving strategy controller samples the current PRB utilization rate and user access volume; if it is less than the control threshold, it starts energy saving; if it is higher than the threshold, it turns off energy saving.

For the PRB utilization threshold (uplink and downlink), there may be differences in the machine learning models for user access threshold prediction, but it is challenging to find a balance between the model generalization capability and the amount of computation. This patent proposes trend forecasting modeling, seasonal modeling and holiday modeling, which are suitable for implementing model reasoning calculations on computing devices with limited calculations.

As shown in Figure 3, after the model training is completed, it can be deployed on a general-purpose CPU to complete inference calculations. However, in order to implement real-time reasoning calculations on the base station and improve computing power, this patent proposes that the energy-saving model can be deployed on a dedicated AI chip to implement a hardware acceleration method for energy-saving model reasoning. The energy-saving model of machine learning is deployed on the base station through the OMC, and the model is loaded on the dedicated AI inference chip through the inference engine to realize heterogeneous calculation of the energy-saving model.

In order to improve the real-time performance of model reasoning and calculation, it is necessary to complete model optimization before deployment, including optimization technologies such as model compression, model quantization, operator fusion, and heterogeneous splitting. Even, in order to make full use of the computing power of dedicated chips, consider hardware-related model optimization techniques, improve memory transmission bandwidth and concurrency of operator calculations, and achieve hardware acceleration for energy-saving model reasoning.

In this embodiment, the modeling of the prediction model is performed based on time series, and data is collected periodically at time intervals, or collected at one point in time. For massive business data, it is necessary to perform feature selection on the data. These features include: base station ID, configuration information such as time, serving cell, and neighbor cell relationship, as well as KPI data such as user access volume, and data such as emergencies and holiday events. Specifically, the energy consumption data of a base station can be expressed as the following triplet information:

{site_id, date, fetures}

Among them, site_id indicates the base station identification, date indicates the sampling timestamp of energy consumption data, and features indicates the sampling feature vector of energy consumption data of the base station, namely: features={serving cell, neighbor cell relationship, user access volume, historical PRB utilization rate, holidays event,...}.

Since time series has trend changes, periodicity, outliers and holiday effects, time series decomposition can be used to decompose them into trends, seasonality and holidays:

y(t)=trend(t)+peroid(t)+holiday(t)+ε(t)

Among them, y(t) represents the uplink and downlink PRB utilization rate at time t, and the function of the number of user accesses at time t; trend(t) represents the utilization rate of uplink and downlink PRBs at time t, and the number of user accesses at time t; peroid( t) represents the uplink and downlink PRB utilization rate at time t, and the seasonal variation function of the number of user accesses at time t; holiday(t) represents the utilization rate of uplink and downlink PRBs at time t, and the holiday function of the number of user accesses at time t; ε(t) represents the data noise or error term, which is assumed to follow a normal distribution.

In view of the constraints of the computing resources of the base station, the utilization rate of uplink and downlink PRBs, and the time series model for predicting the number of user accesses, the calculation is completed by using generalized linear combination, which has the characteristics of real-time, automation, flexibility, and low operation and maintenance costs. Generally, in order to obtain more efficient computing power, a dedicated chip can be designed for inference calculation of the model to achieve more efficient concurrent capabilities.

Specifically, any component time series function satisfies the following expression in the form of generalized linear combination.

y(t)＝w ₀ +w ₁ y ₁ (t)+w ₂ y ₂ (t)+w _n y _n (t)

Wherein, y _i (t) is a nonlinear function, i=1, 2, . . . , n, and n represents the length of the feature vector.

In order to better describe the PRB utilization rate and the trend (increase or decrease) of the number of user accesses over time, this embodiment uses a hyperbolic tangent saturation function to represent the trend change:

Among them, c represents the saturation capacity, which represents the maximum number of users of the base station in the service; k represents the growth rate, and m represents the offset parameter. These two parameters are super parameters in the model, and the values are determined based on engineering experience. The hyperbolic tangent saturation function is conducive to faster gradient updates, avoiding the problem of gradient disappearance, and making model training better and faster convergence.

In order to better describe the PRB utilization rate and the periodic law of the number of user accesses changing with the day, week, month, and year, this embodiment introduces an N-order Fourier function to represent a periodic seasonal model.

in,

are non-linear functions. p represents a cycle, when p represents a year, p=365; when p=month, p=30; when p represents a week, p=7; when p represents a day, p=1. The energy-saving time control of the base station often uses a cycle of one week, and the value of p here is 7. Secondly, N is a super parameter, which depends on engineering experience to determine the value; the larger N is, the stronger its fitting ability is, and if it is too large, it will easily lead to overfitting; the smaller N is, the weaker its fitting ability will be, and it will easily lead to Underfitting; therefore, the value of N is also a super parameter, which depends on the value of engineering experience.

The training parameter can be represented as a vector, and the model training is performed based on the data, and the value is determined when the model converges.

[(w ₁ ,u ₁ ),(w ₂ ,u ₂ ),...,(w _N ,u _N )] ^T

And it is assumed that the value of the vector satisfies the normal distribution Normal(0,δ ² ) when it is initialized, and implements random initialization when initializing the model training. Among them, the larger δ, the greater the periodic effect; the smaller δ, the smaller the periodic effect. Therefore, this parameter is a super parameter, and its value depends on engineering experience.

In order to better describe the impact of holidays or some major events on the PRB utilization rate and the time series of user access numbers, and such impacts often do not follow a fixed or periodic pattern. This patent proposes an effect modeling method based on holidays and major events, so as to realize the PRB utilization rate and the holiday and major event modeling method of the number of user accesses.

Assuming that there are L holidays, for any H _i , representing the data set of holiday i, the regression vector of holidays can be constructed.

Among them, I is an indicator function, that is, I(true) is 1, and I(false) is 0, namely:

Based on the regression vector of holidays, the objective function of PRB utilization and its user access numbers holidays and major events can be constructed:

Among them, α represents the effect factor vector [α ₁ ,α ₂ ,...,α _L ] of holidays, which satisfies the normal distribution Normal(0,v ² ). Among them, the larger v is, the greater the holiday effect is; the smaller v is, the smaller the holiday effect is. Therefore, this parameter is a super parameter, and its value depends on engineering experience.

In this embodiment, the OMC is based on the PRB utilization rate of the time series and its user access number prediction model. The OMC can collect data for each site and complete the model training; and then distribute the model to each base station through the OMC to complete the deployment of the model. , realize the independent decision-making of the energy-saving strategy of the base station, and finally realize the control method of one station, one policy, and realize the best energy-saving effect of the station. Among them, the activation process of the prediction model between the OMC and the base station is shown in Figure 4, and the upgrade process of the prediction model is shown in Figure 5.

When the capacity of the model is improved, the upgrade and replacement of the model can be completed through OMC. When the base station is running, according to the version policy, the existing model is uninstalled and the new model is replaced, so as to realize the smooth transition of the business. In order to provide dedicated computing capability, the base station can deploy a dedicated computing unit to complete the hardware acceleration capability of the model reasoning runtime. During inference runtime, hardware-independent optimization and hardware-related optimization of the model can be completed, and finally the ultimate performance and power consumption efficiency ratio of inference runtime can be achieved.

Therefore, the OMC only needs to control the main switch for site energy saving, and complete the O&M work such as loading and upgrading the site model, which greatly reduces the cost of network operation. The energy-saving prediction model deployed in the base station divides the energy-saving periods into working days and holidays, and then predicts the energy-saving periods t ₁ , t ₂ ,...,t _n of each day, where t _i represents the energy-saving time period with a cycle of 1 hour, And n≤24.

In addition, the energy saving prediction model of the base station will predict the uplink and downlink PRB utilization threshold of t _i and the threshold of the number of user accesses, so as to support the base station to complete the customization of the energy saving strategy. Generally, energy saving is enabled when the following conditions are met:

a) The energy-saving switch is turned on;

b) The current uplink PRB utilization rate is less than the uplink PRB utilization rate threshold;

c) The current downlink PRB utilization rate is less than the uplink PRB utilization rate threshold;

d) The current user access volume is less than the user access volume threshold.

When the base station enters the energy-saving state, it monitors the network monitoring state in real time; when the traffic volume of the base station is higher than the corresponding threshold within a certain time period, the energy-saving state is turned off. In addition, the OMC monitors network KPIs in real time. When the network performance declines at the site, it turns off the energy-saving switch of the energy-saving base station and restores the power amplifier of the base station.

The KPI detection part is also relatively complicated. You can observe the decline in performance data by observing the access, handover, call drop, and rate data. The decline depends not only on the trend but also on the magnitude and type, and whether to determine whether to cover loopholes. That is, according to the business Analyze the reasons for the decline based on experience to see if the main energy-saving switch is turned off.

Specifically, the energy-saving state of the base station includes three functions of turning off: symbol turning off, channel turning off and carrier turning off.

When the OMC turns on the symbol off switch, the BBU loads the energy-saving prediction model, and completes the prediction of the energy-saving time period, and the downlink PRB utilization threshold for each energy-saving time period, and the user access number threshold, and then completes the RRU configuration; After the RRU receives the BBU configuration, when the time slot where the symbol is turned off is the downlink, and the current downlink PRB utilization rate is lower than the downlink PRB utilization threshold, and the number of cell access users is less than the threshold, the symbol is turned off.

When the OMC turns on the channel switch, the BBU loads the energy-saving prediction model, and completes the prediction of the energy-saving time period, as well as the uplink and downlink PRB utilization threshold for each energy-saving time period, and the user access number threshold, and then completes the RRU configuration; RRU After receiving the BBU configuration, if the current PRB utilization rate of the uplink and downlink channels is lower than the corresponding threshold, and the number of access users in the cell is less than the threshold, the power amplifier of the corresponding channel is turned off.

When the OMC turns on the channel switch, the BBU loads the energy-saving prediction model, and completes the prediction of the energy-saving time period, as well as the uplink and downlink PRB utilization threshold for each energy-saving time period, and the user access number threshold, and then completes the RRU configuration; RRU After receiving the BBU configuration, if the PRB utilization rate of the current carrier is lower than the corresponding threshold and the number of users accessing the cell is smaller than the threshold, the power amplifier of the corresponding carrier is turned off.

The embodiment of the present application relates to a device energy saving method, as shown in FIG. 6, applied to a management server, including the following steps:

Step 601, obtaining the reported service data of the base station;

Step 602, train the prediction model according to the reported business data;

Step 603, sending the trained prediction model to the base station; wherein, the prediction model is used to predict the energy-saving threshold of the base station, so that the base station can control the energy-saving operation according to the energy-saving threshold.

In an example, after sending the trained prediction model to the base station, the method further includes: continuously acquiring the feedback service data and energy-saving effect data of the base station; when the energy-saving effect data of the base station is lower than a preset threshold, Feedback business data, train the prediction model again; deliver the retrained prediction model to the base station.

The step division of the above various methods is only for the sake of clarity of description. During implementation, it can be combined into one step or some steps can be split and decomposed into multiple steps. As long as they include the same logical relationship, they are all within the scope of protection of this patent. ; Adding insignificant modifications or introducing insignificant designs to the algorithm or process, but not changing the core design of the algorithm and process are all within the scope of protection of this patent.

The embodiment of the present application also relates to a device energy saving system, as shown in FIG. 7 , including a base station 701 and a management server 702;

Among them, the base station 701 is used to obtain the historical service data of the base station, dynamically predict the energy saving threshold of the base station according to the historical service data and the preset prediction model, obtain the current service data of the base station, and control the energy saving threshold of the base station according to the current service data and the energy saving threshold. energy saving operation;

The management server 702 is used to obtain the reported business data of the base station, train the prediction model according to the reported business data, and send the trained prediction model to the base station, wherein the prediction model is used to predict the energy saving threshold of the base station, for The base station controls the energy-saving operation according to the energy-saving threshold.

In an example, dynamically predicting the energy-saving threshold of the base station according to the historical service data and the preset prediction model includes: predicting the energy-saving threshold of the base station within a preset period according to the historical service data and the preset prediction model.

In an example, the energy-saving threshold includes: the PRB resource utilization threshold of the physical resource block and the access user number threshold; obtaining the current service data of the base station includes: obtaining the current PRB resource utilization rate of the base station and the current access user number; according to The current service data and the energy-saving threshold control the energy-saving behavior of the base station, including: when the current PRB resource utilization rate is less than the PRB resource utilization rate threshold, and the current number of access users is less than the access user number threshold, control the base station to enter the energy-saving state; When the current PRB resource utilization rate is greater than or equal to the PRB resource utilization rate threshold, or the current number of access users is greater than or equal to the access user number threshold, the base station is controlled to exit the energy-saving state.

In an example, the PRB resource utilization threshold includes: uplink PRB resource utilization and downlink PRB resource utilization; when the current PRB resource utilization is less than the PRB resource utilization threshold, and the current number of access users is less than the access user number threshold In the case of control, the base station is controlled to be in an energy-saving state, including any one or any combination of the following: at the current moment, it is downlink, and the current downlink PRB resource utilization rate is less than the downlink PRB resource utilization rate threshold, and the current number of access users is less than the access In the case of the threshold of the number of incoming users, the symbol is turned off; when the current uplink PRB resource utilization is less than the uplink PRB resource utilization threshold, and the current number of accessing users is less than the threshold of the number of accessing users, the power amplifier of the uplink channel is turned off; When the current downlink PRB resource utilization rate is less than the downlink PRB resource utilization rate threshold, and the current number of access users is less than the access user number threshold, turn off the power amplifier of the downlink channel; the current PRB resource utilization rate of the current carrier is less than the PRB resource The utilization threshold, and when the number of current access users is less than the threshold of the number of access users, turn off the power amplifier of the current carrier.

The embodiment of the present application also relates to an electronic device, as shown in FIG. 8 , including: at least one processor 801; a memory 802 communicatively connected to the at least one processor; The executed instructions are used by at least one processor 801 to execute the device energy saving method in any of the foregoing embodiments.

Wherein, the memory 802 and the processor 801 are connected by a bus, and the bus may include any number of interconnected buses and bridges, and the bus connects one or more processors 801 and various circuits of the memory 802 together. The bus may also connect together various other circuits such as peripherals, voltage regulators, and power management circuits, all of which are well known in the art and therefore will not be further described herein. The bus interface provides an interface between the bus and the transceivers. A transceiver may be a single element or multiple elements, such as multiple receivers and transmitters, providing means for communicating with various other devices over a transmission medium. The information processed by the processor 801 is transmitted on the wireless medium through the antenna, further, the antenna also receives the information and transmits the information to the processor 801 .

The processor 801 is responsible for managing the bus and general processing, and may also provide various functions including timing, peripheral interface, voltage regulation, power management and other control functions. Instead, memory 802 may be used to store information used by the processor in performing operations.

Embodiments of the present application relate to a computer-readable storage medium storing a computer program. The above method embodiments are implemented when the computer program is executed by the processor.

That is, those skilled in the art can understand that all or part of the steps in the method of the above-mentioned embodiments can be completed by instructing related hardware through a program, the program is stored in a storage medium, and includes several instructions to make a device ( It may be a single-chip microcomputer, a chip, etc.) or a processor (processor) to execute all or part of the steps of the methods in the various embodiments of the present application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disc, etc., which can store program codes. .

Claims

A device energy saving method applied to a base station, comprising:

Obtain historical service data of the base station;

dynamically predicting the energy-saving threshold of the base station according to the historical service data and a preset prediction model;

Acquiring current service data of the base station;

Control the energy saving operation of the base station according to the current service data and the energy saving threshold.
The equipment energy-saving method according to claim 1, wherein the prediction model is established by any one of the following dimensions, or any combination of dimensions:

Changes in trends, cycle seasons, holidays and major events.
The device energy-saving method according to claim 1, wherein the dynamically predicting the energy-saving threshold of the base station according to the historical service data and a preset prediction model includes:

Predict the energy saving threshold of the base station within a preset period according to the historical service data and the preset prediction model.
The device energy saving method according to claim 3, wherein the predicting the energy saving threshold of the base station within a preset period includes:

Predicting the energy saving threshold corresponding to each time period of the base station within the preset cycle duration.
The device energy saving method according to any one of claims 1 to 4, wherein the energy saving threshold includes:

Physical resource block PRB resource utilization threshold and access user number threshold;

The acquisition of the current service data of the base station includes:

Obtain the current PRB resource utilization rate and the current number of access users of the base station;

The controlling the energy-saving behavior of the base station according to the current service data and the energy-saving threshold includes:

When the current PRB resource utilization rate is less than the PRB resource utilization rate threshold and the current number of access users is less than the access user number threshold, control the base station to enter an energy-saving state;

When the current PRB resource utilization rate is greater than or equal to the PRB resource utilization rate threshold, or the current number of access users is greater than or equal to the access user number threshold, control the base station to exit the energy saving state.
The device energy saving method according to claim 5, wherein the PRB resource utilization threshold includes: uplink PRB resource utilization and downlink PRB resource utilization;

In the case where the current PRB resource utilization rate is less than the PRB resource utilization rate threshold and the current number of access users is less than the access user number threshold, controlling the base station to perform an energy-saving state includes any of the following One or any combination:

When the current moment is downlink, and the current downlink PRB resource utilization rate is less than the downlink PRB resource utilization rate threshold, and the current access user number is less than the access user number threshold, start symbol off;

When the current uplink PRB resource utilization rate is less than the uplink PRB resource utilization rate threshold, and the current number of access users is less than the access user number threshold, turn off the power amplifier of the uplink channel;

When the current downlink PRB resource utilization rate is less than the downlink PRB resource utilization rate threshold, and the current number of access users is less than the access user number threshold, turn off the power amplifier of the downlink channel;

When the current PRB resource utilization rate of the current carrier is less than the PRB resource utilization rate threshold and the current number of access users is less than the access user number threshold, turn off the power amplifier of the current carrier.
A device energy-saving method applied to a management server, comprising:

Obtain the reported business data of the base station;

Training the prediction model according to the reported business data;

sending the trained prediction model to the base station;

Wherein, the prediction model is used to predict the energy-saving threshold of the base station, so that the base station can control the energy-saving operation according to the energy-saving threshold.
The device energy-saving method according to claim 7, wherein, after sending the trained prediction model to the base station, the method further comprises:

Continuously acquire the feedback service data and energy-saving effect data of the base station;

When the energy-saving effect data of the base station is lower than a preset threshold, retrain the prediction model according to the feedback service data;

Sending the retrained prediction model to the base station.
A device energy-saving system, including a base station and a management server;

Wherein, the base station is used to acquire the historical service data of the base station, dynamically predict the energy-saving threshold of the base station according to the historical service data and a preset prediction model, obtain the current service data of the base station, and The current service data and the energy-saving threshold are used to control the energy-saving operation of the base station;

The management server is configured to acquire the reported service data of the base station, train the prediction model according to the reported service data, and deliver the trained prediction model to the base station, wherein the prediction model is used for Predicting the energy-saving threshold of the base station for the base station to control energy-saving operations according to the energy-saving threshold.
An electronic device comprising:

at least one processor;

memory communicatively coupled to the at least one processor;

The memory is stored with instructions executable by the at least one processor, the instructions are executed by the at least one processor, so that the at least one processor can perform any one of claims 1 to 6 The equipment energy-saving method, or the equipment energy-saving method according to claim 7 or 8.
A computer-readable storage medium storing a computer program, wherein, when the computer program is executed by a processor, the device energy-saving method according to any one of claims 1 to 6 is implemented, or, as described in claim 7 or 8 Energy-saving methods for equipment described above.