Managing the delivery of power to the radio head

By employing a local battery managed by machine learning to differentiate between average and peak power consumption, the method optimizes power transmission in RANs, reducing infrastructure needs and enhancing system robustness and sustainability.

JP2024543818AActive Publication Date: 2024-11-26TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
JP2024526003
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2021-11-09
Publication Date
2024-11-26
Estimated Expiration
2041-11-09

AI Technical Summary

Technical Problem

The challenge lies in managing the difference between average and peak power consumption in radio access networks (RAN) to reduce energy consumption and operating costs when adding new radio access technologies, as existing methods often lead to increased energy consumption and infrastructure costs.

Method used

A method utilizing a local battery in proximity to wireless heads, managed by a computing device with machine learning models, determines the difference in power demand statistics to optimize power transmission, reducing the need for additional power lines and fuses by supplying peak power from the local battery rather than the grid.

Benefits of technology

This approach reduces the number and length of cables and fuses, enhances system robustness against grid failures, and enables zero-emission RAN operations by using green energy sources, while optimizing battery charging and extending battery life.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024543818000001_ABST
    Figure 2024543818000001_ABST
Patent Text Reader

Abstract

A method is provided for managing the transmission of power from a local battery located proximate to at least one radio head, the method comprising: making a decision for transmitting power from the local battery to the at least one radio head in a future time window (1001). The decision is made by a machine learning model based on (i) a difference determination of output power data statistics including average power demand and peak power demand, and (ii) time- and location-dependent cost data of charging and / or discharging the local battery and grid utilization. The method further comprises outputting a decision for transmitting power from the local battery to the at least one radio head in a future time window (1003).
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present disclosure relates generally to methods, and related methods and apparatus, for managing the delivery of power to a radio head from a local battery located in close proximity to the radio head. [Background technology]

[0002] Radio Access Networks (RANs) account for a large portion of the energy consumption of mobile networks. Increasing the number of radio units in a RAN is a common approach to expand capacity and coverage to improve service quality. Adding radio units may include adding more radio units of the same type and / or radio units for different radio access technologies (RATs) and bands. However, such approaches increase energy consumption.

[0003] In some approaches, the site dimensioning (e.g. power cables, circuit breakers, power supply units (PSUs)) is done based on the maximum radio power consumption of each radio unit, which can significantly affect and increase the input rating of the site fuses. Increasing the rating of the site fuses can have a significant impact on operational costs for the operator.

[0004] Future and existing RAN features (e.g., microsleep transmit (Tx), low energy scheduler solutions (LESS), multiple-input multiple-output (MIMO) sleep modes, cell sleep modes, etc.) will increase the sleep time of radio units and reduce power consumption. Such techniques will substantially increase the difference between the average power consumption and the peak power used by an installed radio unit. Summary of the Invention

[0005] Currently, one or more of certain challenges exist: There is a need to manage and / or transmit power in a way that allows one to determine the difference between the average power and the peak power of a wireless head, while still adding new radio access technologies (RATs) in a sustainable manner.

[0006] Certain aspects of the present disclosure and their embodiments may provide solutions to these and other problems.

[0007] According to various embodiments, a method is provided for managing the transmission of power from a local battery located proximate to at least one radio head, the method comprising making a decision on the transmission of power from the local battery to the at least one radio head in a future time window, the decision being made by a machine learning model based on (i) a determination of a difference in output power data statistics including average and peak power demands in a set of past time windows covering a defined period for the at least one radio head, and (ii) time- and location-dependent cost data of charging and / or discharging the local battery and grid utilization in the future time window. The method further comprises outputting the decision on the transmission of power from the local battery to the at least one radio head in the future time window.

[0008] According to another embodiment, a computing device in a communication system for managing the transmission of power from a local battery located in proximity to at least one radio head is provided. The computing device includes at least one processor and at least one memory coupled to the at least one processor and storing program code executed by the at least one processor for performing operations of making a decision on the transmission of power from the local battery to the at least one radio head in a future time window. The decision is made by a machine learning model based on (i) a determination of a difference in output power data statistics including average power demand and peak power demand in a set of past time windows covering a defined period for the at least one radio head, and (ii) time- and location-dependent cost data of charging and / or discharging of the local battery and grid utilization in the future time window. The operation further includes outputting the decision on the transmission of power from the local battery to the at least one radio head in the future time window.

[0009] According to another embodiment, a computing device in a communication system for managing the transmission of power from a local battery located in proximity to at least one radio head is provided. The computing device is adapted to perform operations including making a decision on the transmission of power from the local battery to the at least one radio head in a future time window. The decision is made by a machine learning model based on (i) a determination of a difference in output power data statistics including average power demand and peak power demand in a set of past time windows covering a defined period for the at least one radio head, and (ii) time- and location-dependent cost data of charging and / or discharging of the local battery and grid utilization in the future time window. The operations further include outputting the decision on the transmission of power from the local battery to the at least one radio head in the future time window.

[0010] According to another embodiment, a computer program is provided, the computer program including program code executed by a processing circuit of a computing device in a communication system for managing the transmission of power from a local battery located proximate to at least one radio head to the at least one radio head. Execution of the program code causes the computing device to perform operations including making a decision on the transmission of power from the local battery to the at least one radio head in a future time window. The decision is made by a machine learning model based on (i) a determination of a difference in output power data statistics including average power demand and peak power demand in a set of past time windows covering a defined period for the at least one radio head, and (ii) time- and location-dependent cost data of charging and / or discharging the local battery and grid utilization in the future time window. The operation further includes outputting the decision on the transmission of power from the local battery to the at least one radio head in the future time window.

[0011] According to another embodiment, a computer program product is provided that includes a non-transitory storage medium including a program code executed by a processing circuit of a computing device in a communication system for managing the transmission of power from a local battery located proximate to at least one radio head to the at least one radio head. Execution of the program code causes the computing device to perform operations including making a decision on the transmission of power from the local battery to the at least one radio head in a future time window. The decision is made by a machine learning model based on (i) a determination of a difference in output power data statistics including average power demand and peak power demand in a set of past time windows covering a defined period for the at least one radio head, and (ii) time- and location-dependent cost data of charging and / or discharging the local battery and grid utilization in the future time window. The operation further includes outputting the decision on the transmission of power from the local battery to the at least one radio head in the future time window.

[0012] Particular embodiments may provide one or more of the following technical advantages. As described above, the method of some embodiments may manage the transmission of power from a local battery located in proximity to at least one radio head to at least one radio head. The method includes making a decision on the transmission of power from the local battery to at least one radio head in a future time window. The decision is made by (i) determining a difference in output power data statistics including average power demand and peak power demand in a set of past time windows covering a defined period for the at least one radio head, and (ii) outputting a decision on the supply of power from the local battery to the at least one radio head for the future time window based on a machine learning model based on charging and / or discharging the local battery and time- and location-dependent cost data of grid utilization for the future time window. As a result, technical advantages provided by the method may include energy cost savings based on charging and / or discharging the local battery against peak power demand. Furthermore, peak shaving at the radio head may be enhanced based on the management of the local battery using ML (machine learning). Such management may not only provide and / or optimize local battery charging / discharging policies, but may also extend the life of the local battery due to the determination and output of decisions. [Brief description of the drawings]

[0013] The accompanying drawings, which are included to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this application, illustrate certain non-limiting embodiments of the inventive concepts.

[0014] [Figure 1] FIG. 1 is a schematic diagram showing a conventional RAN site deployment.

[0015] [Diagram 2]FIG. 2 is a schematic diagram showing the addition of a radio head to an existing conventional RAN site.

[0016] [Diagram 3] 3 is a schematic diagram illustrating an additional radio head of the RAN site of FIG. 2 in accordance with some embodiments of the present disclosure.

[0017] [Figure 4] Figure 1 shows a plot of the input power to four radio heads at a two-sector radio site over the course of a day.

[0018] [Diagram 5] 5 is the plot of FIG. 4 annotated in accordance with some embodiments of the present disclosure.

[0019] [Figure 6] FIG. 2 is a schematic diagram illustrating the state of charge of a local battery according to some embodiments of the present disclosure.

[0020] [Figure 7] 1 is a sequence diagram illustrating operations according to some embodiments of the present disclosure.

[0021] [Figure 8] 1 is a sequence diagram illustrating operations according to some embodiments of the present disclosure.

[0022] [Figure 9] 1 is a block diagram illustrating a computing device according to some embodiments of the present disclosure.

[0023] [Figure 10] and [Figure 11] 1 is a flowchart illustrating the operation of a computing device node according to some embodiments of the present disclosure.

[0024] [Figure 12]1 is a block diagram of a communication system in which a communication system according to some embodiments of the present disclosure may be implemented.

[0025] [Figure 13] 1 is a block diagram of a network node according to some embodiments of the present disclosure.

[0026] [Figure 14] 1 is a block diagram of a host computer in communication with user equipment according to some embodiments of the present disclosure.

[0027] [Figure 15] 1 is a block diagram of a virtualization environment according to some embodiments of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0028] The concept of the present disclosure will now be described more fully with reference to the accompanying drawings, which show examples of embodiments of the concept of the present disclosure. However, the concept of the present disclosure can be embodied in many different forms and should not be construed as being limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the concept of the present disclosure to those skilled in the art. It should also be noted that these embodiments are not mutually exclusive. Elements of one embodiment may be implicitly assumed to be present / used in another embodiment.

[0029] The following description presents various embodiments of the disclosed subject matter. These embodiments are presented as instructional examples and are not to be construed as limiting the scope of the disclosed subject matter. For example, certain details of the described embodiments may be modified, omitted, or expanded without departing from the scope of the described subject matter.

[0030] Figure 1 is a schematic diagram illustrating a conventional RAN site deployment. As shown in Figure 1, a power source 101 is connected to three 4G radio heads (also called radio units) 103a, 103b, 103c via power supply line 105. The power source 101 also includes a backup battery (labeled "B" in Figure 1).

[0031] Figure 2 is a schematic diagram illustrating the addition of radio heads to an existing conventional RAN site (e.g., the RAN site of Figure 1). As shown in Figure 2, three additional radio heads 204a, 204b, 204c are added to the RAN site of Figure 1 and connected to the power source 101 via three additional power supply lines 105, so that there are six power lines extending from the power source 101 to each of the radio heads 103a, 103b, 103c, 204a, 204b, and 204c, respectively.

[0032] "Peak shaving" is a term that refers to the cutting of power at a utility when demand from customers is high. Several publications discuss architectures that control peak shaving for utilities and / or power transmission lines.

[0033] Currently, one or more challenges of some kind exist.

[0034] Architectures to control peak shaving for utilities and / or power transmission lines behave differently than power systems with integrated local backup batteries at the site for wireless communications, for example as shown in Figures 1 and 2. Wireless communications add non-trivial challenges and new variables with respect to time-varying wireless network loads and targeted key performance indicators (KPIs).

[0035] Existing approaches for site equipment include a backup battery (e.g. B in Figures 1 and 2) used as a backup in case of power outage. The site backup battery B is typically dimensioned according to the site's local regulations (e.g. time) and the average power consumption requirement of the site. As a result, the site dimensioning (e.g. power cables, circuit breakers, PSUs) is often done based on the maximum radio power consumption of each radio unit, which has a very large impact on and increases the input rating of the site fuses. However, increasing the site fuse rating can have a significant negative impact on operational costs for the operator.

[0036] In addition, the approach of adding additional radio units to existing RAN sites (e.g., as shown in Figure 2) lacks the determination of the difference between the average and peak power demands for the radio units, and instead adds more equipment (e.g., more power lines, fuses, etc.) to provide the power demand for the additional radio heads (other than backup battery power) from the grid via a grid-tied source. However, such a determination of the difference between the average and peak power is important to optimize / improve the design and operation of local batteries to make the process of adding new radio access technologies (RATs) more sustainable while reducing the total carbon footprint of wireless network operations.

[0037] There is a need to improve site equipment and methods for transmitting power in a manner that allows for the determination of the difference between the average power and peak power of the radio heads while adding new RATs in a more sustainable manner (e.g., so that peak power needs are supplied from local batteries rather than from the grid).

[0038] Certain aspects of the present disclosure and their embodiments may provide solutions to these and other problems. Certain embodiments of the present disclosure include a method that uses machine learning (ML) to determine the difference between the average power usage and the peak power usage of different radio heads. A local battery provides auxiliary power to handle the peak power consumption of the radio heads. The local battery is located in close proximity to the radio heads. A power line from a site power source provides average power to the radio heads.

[0039] Certain embodiments may provide one or more of the following additional technical advantages: As described above, according to some embodiments, a method for managing the transmission of power from a local battery located in proximity to at least one radio head is provided. The method includes making a decision on the transmission of power from the local battery to the at least one radio head in a future time window. The decision is performed by (i) determining a difference in output power data statistics including average power demand and peak power demand in a set of past time windows covering a defined period for the at least one radio head, and (ii) a machine learning model based on time- and location-dependent cost data of charging and / or discharging the local battery and grid utilization for the future time window, and outputting a decision on the supply of power from the local battery to the at least one radio head for the future time window.

[0040] As a result of locating a local battery in close proximity to a radio head, making the determination, and outputting the determination on the transmission of power from the local battery, a reduction in the number or length of power lines (also referred to herein as "cables") required and a reduction in the number of fuses may be provided, and the expansion of a site with additional radio heads may be easily accomplished with fewer or shorter lengths of power lines. For example, when a radio head is added to an existing conventional RAN site (e.g., the RAN site of FIG. 1), based on the employment of a local battery in close proximity to the radio head (e.g., as shown in FIG. 3, further described herein), the number of power lines from the power source (e.g., power source 101) to the local battery (e.g., local battery 301) may remain the same (e.g., three power lines 105). This is in contrast to existing approaches for expanding a site to have additional radio heads, where new power lines need to be installed for the entire distance from the power source to the new radio head. In the example of FIG. 2, three additional power lines are added such that there are six power lines covering the entire distance between the power source (e.g., power source 101) and the radio heads (e.g., radio heads 103a-103b and 104a-104c as shown in FIG. 2). Furthermore, based on the adoption of local battery management for wireless peak power, the operator of the wireless system can provide robustness against grid disturbances since the local battery is an active power source (e.g., not simply a backup power source). Further technical advantages of including a local battery in close proximity to the radio head(s) to manage the transmission of power to the radio head(s) may include the use of the grid to power the radio heads and / or the use of green energy sources to charge the local battery (e.g., to charge the backup battery located at the grid power source), which may contribute to a zero-emission RAN, as well as increased network robustness against grid disturbances.

[0041] Figure 3 is a schematic diagram showing additional radio heads 204a, 204b, 204c of Figure 2, but with a RAN site configured according to some embodiments of the present disclosure. The architecture of Figure 3 can realize peak shaving of local radio heads based on installing a local battery 301. According to certain embodiments of the present disclosure, a method for management of peak power consumption of at least one radio head 103a-103c, 204a-204c from the local battery 301 includes differentiation (determining the difference) between average power and radio peak power and managing the transmission of power from two different power sources 101, 301 during normal network operation.

[0042] In FIG. 2, backup battery B is a backup power source, and additional radio heads 204a-204c are added to a site that includes a different RAT, e.g., LTE-enhanced radio head 204c. In contrast, FIG. 3 illustrates a site design that adds a local battery 301, which achieves peak shaving while further adding additional radio heads 204a-204c, while reducing the number of cables and fuses required. According to the exemplary embodiment of FIG. 3, the number of cables 105 from the power source 101 is reduced compared to FIG. 2, based on including the local battery 301 as an active power source (i.e., not a battery backup). In FIG. 2, the addition of radio heads 204a-204c also adds three power cables 105 that cover the entire distance between the power source 101 and the radio heads 204a-204c, such that there are six power lines 105 that cover the entire distance between the power source 101 and the radio heads 103a-103c and 204a-204c, respectively. In contrast, in Figure 3, adding wireless heads 204a-204c does not add power lines 105 between the power source 101 and the local battery 101. Thus, in contrast to Figure 2, the number and / or length of power lines 105 is reduced in Figure 3. Furthermore, in Figure 3, power injected from the grid through the power source 101 can be reduced because the local battery 301 is an active source of power for peak power demand.

[0043] Figure 4 is a plot of the input power to four radio heads (radio heads 1, 2, 3, 4 as shown in the legend) in a two-sector radio site over a defined period of approximately 30 seconds. The solid and dashed lines represent radios 1 and 2 in the first sector. The dotted and dashed lines represent radios 3 and 4 in the second sector. As can be seen from the plot illustrated in Figure 4, the radio heads have average power in watts and power peaks over the period.

[0044] 5 illustrates the plot of FIG. 4 annotated with horizontal lines 501, 503, 505, 507 illustrating the determination of the difference between the average and peak powers of the wireless heads over a period of time in accordance with some embodiments of the present disclosure. It can be seen from the illustrated plot that, according to the exemplary embodiment of FIG. 5, given the separation between the average powers shown by lines 503, 507 and the peak powers shown by lines 501, 505, respectively, with proper dimensioning of the power lines 105, the average powers 503, 507 of the wireless units can be carried and the peak powers 501, 505 can be controlled from the local battery 301.

[0045] 4 and 5 are described with reference to four radio heads and a defined period of approximately 30 seconds, embodiments of the present disclosure are not so limited. Alternatively, any quantity of one or more radio heads may be included, and the defined period may be any period defined in seconds, minutes, hours, days, weeks, etc. Additionally, while two peak power settings and two average power settings are shown for the plot shown in FIG. 5, embodiments of the present disclosure are not so limited. Alternatively, a peak power setting may be identified / determined (as further described herein) for each radio head and / or for one or more radio heads. Additionally, an average power setting may be identified / determined (as further described herein) for each radio head and / or for one or more radio heads. Additionally, although some embodiments are described herein with reference to average power supplied by the power grid and peak power supplied by a local battery, embodiments of the present disclosure are not so limited. Alternatively, an amount of power greater than or less than the peak power may be supplied by the local battery, and the remaining power demand of the radio head may be supplied from the power grid.

[0046] According to certain embodiments of the present disclosure, the ML-based method can manage the charging and discharging policies of the local battery 301 for peak shaving based on the time variation of wireless power peak demand as well as the grid energy price and carbon footprint profile for the time period.

[0047] According to some embodiments, the following observations are taken into account: - There may be significant variations in the input and output power of radio heads. Some radio heads are very active at high output power, while others rarely reach high output power. ●There is a high correlation between the peak power of the radio head relative to the input power from the site power supply (and grid) and the output power of the radio head (which peaks at peak traffic loads). For active time windows during which the radio head is substantially inactive, a differentiation is required between the average radio power and the peak radio power (see e.g. FIG. 4). Such a differentiation can support efficient operation of the local battery 301. ● The output power of a single radio head is highly correlated with time of day (e.g., over the course of an overnight period).

[0048] According to some embodiments, these observations are used to build ML methods for managing a local battery (eg, battery 301).

[0049] According to some embodiments, a defined time period (e.g., a day) is divided into a set of time windows (e.g., hourly) and decision variables are generated for future time windows. The method uses data collected or accessed over time derived from the power consumption of wireless heads (including wireless traffic variations) to train an ML model (e.g., an RL agent).

[0050] Training of the RL agent includes at least the following inputs: (1) local battery capacity, (2) average power consumption of all wireless heads in K past time windows among the set of time windows for a given integer K>0, (3) peak power consumption of all wireless heads in K past time windows among the set of time windows, and (4) charging cost of the local battery and cost of using the grid in the current or future time window. The RL agent optimizes a charging state decision including when to charge the local battery for the start of the current / future time window (e.g., each time window), when and how much to discharge the local battery for the wireless head, and does not transmit power from the local battery to the wireless head (e.g., if the charging state decision is to determine that the amount of discharge from the local battery for the wireless head is zero).

[0051] FIG. 6 is a schematic diagram illustrating a state of charge (SOC) of a local battery according to some embodiments of the present disclosure. As shown in FIG. 6, the SOC level of the local battery ranges from 0% to 100%. The ML model (e.g., RL model) defines / sets the SOC of the local battery for the predicted radio peak power in the future time window(s) (e.g., at each future time window). The SOC level can vary between 0%-100% and is defined at the SOC level depending on the predicted peak power. According to the example embodiment of FIG. 6, the SOC level may be set for the future time window at the SOC level indicated by the three arrows. Although the example of FIG. 6 shows three SOC levels, the present disclosure is not so limited and the SOC level may be defined at any non-zero SOC level depending on the predicted peak power.

[0052] According to some embodiments, the SOC decision (also referred to herein as “decision”) satisfies the following four constraints on the time window: C1. Constraint 1: The local battery discharge level to the radio head for any time window plus the input power (from the power cable) should be equal to the input power required for that radio head at that time. These time- and location-stamped input powers are inputs / constants to the ML model given in the dataset. C2. Constraint 2: The local battery discharge in any time window should not exceed the local battery level at the start of that time window plus the charge profile limit for that time window. C3. Constraint 3: The current local battery level is equal to the sum of the previous local battery level and the charge in this time window. C4. Constraint 4: The local battery level in any time window must not exceed the local battery capacity.

[0053] The objective of the approach is to minimize the total cost, including the cost of charging the local battery and the cost of using the grid, for future time window(s) by allowing the average power to the radio heads to be transmitted from the site power system based on the adoption of the above constraints.

[0054] The ML model is now further described. According to some embodiments, the ML model is an episodic RL agent, where each episode is of a defined duration (e.g., one day). According to some embodiments, the RL agent is trained in the cloud and then installed at the battery management service location (e.g., deployed on the site's local battery).

[0055] According to some embodiments, the RL agent uses a dataset of power consumption of wireless heads over a defined period of time and determines the difference between wireless average power and wireless peak power for one episode or more episodes (e.g., several days).

[0056] According to some embodiments, the actions include determining charging and discharging at least one local battery for the radio head.

[0057] According to some embodiments, the status includes the current local battery level, the current output power of all radio heads, and the current local battery charging cost.

[0058] According to some embodiments, the reward of the RL agent for every state-action pair is to minimize the total cost of input power in the next time window in the set of time windows. The total cost is a weighted sum of the power from the grid and the power from the local battery, weighted based on their own costs. According to some embodiments, the total cost includes, among others, a monetary cost of energy or a carbon footprint index.

[0059] An exemplary embodiment will now be described with reference to Figure 7. Although the exemplary embodiment of Figure 7 is described in the context of Figure 3, which shows one local battery and six radio heads, various embodiments of the present disclosure are not so limited and may include a different number of local batteries and different numbers or types of radio heads. Additionally, although certain embodiments are described with reference to one local battery for ease of explanation, various embodiments of the present disclosure are not so limited and may include any number of local batteries.

[0060] 7, according to an exemplary embodiment, a total of R wireless heads (e.g., wireless heads 103a-103c and 204a-204c), indexed by i∈{1, 2,...,R} (R=6 according to this exemplary embodiment), are connected to a local battery 301 with a charge of C, as shown by connection 701, and further connected to a power grid via a power source 101, as shown by connection 703. In operation 705, a computing device (e.g., computing device 900 described further herein) divides each episode (i.e., a defined period for decision-making, such as every day) into a set of W time windows of predefined duration T, and calculates the time windows B. w denotes the level of the local battery 301 at the beginning of a future time window w from the set of time windows. As shown in operation 709, in every window w, the i-th wireless head of the plurality of wireless heads (e.g., a wireless head as shown for wireless head 103a in FIG. 7) will have a maximum power p iw max , average power p iw avg , and instantaneous power p iw (t). For each time window w, the local battery 301 has d iw b0. As shown in box 711, in every time window w, the charging of the local battery 301 is performed according to a predefined charging profile h w b (t) and cost profile c w b By following (t), the total charge amount during this time window is

number

number

number

[0061] Continuing to refer to FIG. 7, the following constraints are included for each time window in the set of time windows: First constraint: Input power to the radio head: d iw b +d iw g =p iw (t). Second constraint: Local battery level (this includes constraints 2-4 above): B w +H w b -Td iw b ∈[0,C].

[0062] The operating cost (e.g., electricity cost) in any time window is C_w^b+C_w^g, which is f(d 11 b ,d 12 b ,...,d RW b )=Σ w∈W C w b + Σ i∈R C iw g for the entire episode. Thus, the optimization problem, where f is the reward function, is formulated as follows: Subject to constraints C1 and C2, minimize f(d 11 b ,d 12 b ,...,d RW b )

[0063] The reward function f is therefore given by f(d) for the entire episode constrained by C1 (i.e., in any time window, the local battery discharge level to the radio head plus the power input (from the power cable) should be equal to the input power required by the radio head at that time) and C2 (i.e., the local battery discharge in any time window must not exceed the local battery level at the start of this time window plus the charge profile limit for that time window). 11 b ,d 12 b ,...,d RW b )=Σ w∈W C w b + Σ i∈R C iw g This is to minimize the cost.

[0064] As shown in operation 715, the RL agent of the computing device 900 solves the optimization problem. According to some embodiments, the RL agent calculates the power consumption {p iw (t)}, as well as charging the local battery 301 via the power source 101 and powering the grid c w g The RL agent has access to a cost profile that uses (t) and the dataset to simulate the environment.

[0065] FIG. 8 is a sequence diagram illustrating operations for managing peak power demand from a local battery to one or more radio heads according to some embodiments of the present disclosure. According to the exemplary embodiment of FIG. 8, the ML model is an RL agent 803. The environment 801 is (i) a simulator based on data collected during training of key performance indicators (KPIs) and power data, or (ii) an actual network where the KPIs and power data are measured after the RL agent 803 and the local battery are installed. The KPIs include, but are not limited to, local battery and cost data including a capacity of the local battery, a first cost for charging the local battery in a future time window, and a second cost for using power from the grid for an average power demand of at least one radio head in a future time window. The power data includes, but is not limited to, (i) an average power demand of at least one radio head in a set of past time windows, (ii) a peak power demand of at least one radio head in a set of past time windows, and (iii) a traffic load of at least one radio head in a set of past time windows.

[0066] 8, at operation 807, the RL agent 803 gets input data from the environment 801, including KPIs and input power consumption. At operation 809, the RL agent 803 maps the input to an environment state and sends the state to the reward computing device 805. The RL agent 803 computes the next action to be taken and sends the action to the reward computing device 805 at operation 811. The reward computing device 805 computes a reward at operation 813. At operation 815, the RL agent 803 receives the reward and then improves the RL model and actions based on the reward at operation 817.

[0067] Continuing to refer to FIG. 8, according to some embodiments, during training, the loop of operations 807-819 continues to determine a local battery management policy (eg, an optimal battery management policy).

[0068] According to some embodiments, after determining the local battery management policy, the trained RL agent 803 is deployed in the real network. According to some embodiments, the RL agent 803 is deployed at every site (e.g., battery management service location) or in the cloud. The deployed RL agent 803 can access data from the network and monitor KPIs for potential retraining / tuning of the RL agent. Operations 807-819 of the loop are executed in the real network to determine the local battery management policy.

[0069] According to some embodiments, the at least one local battery has a set of capabilities for making the at least one local battery an auxiliary power source, including determining and executing active charging and discharging to minimize some cost function (e.g., for an operator).

[0070] According to some embodiments, at least one local battery is installed near the radio head, and the at least one local battery is responsible for only the peak power consumption. The power cable coming from the site power system for the radio head (e.g., a power source 60 m away from the radio head on site) is set to only deliver average power to the radio head. The at least one local battery has a local calculation or SC decision to be responsible for the peak power.

[0071] According to some embodiments, the method determines the difference between the average power and the radio peak power consumption of several radio heads in a state space, and a decision is provided that enables the local battery to transmit only the peak power to the radio heads.

[0072] According to some embodiments, the method calculates daily traffic variations and power demands and enables operation of at least one local battery (eg, charging or discharging for peak power demands).

[0073] FIG. 9 is a block diagram illustrating elements of a computing device 900 (also referred to as a server, cloud-based server, edge server, radio access network node, base station, radio base station, eNodeB / eNB, gNodeB / gNB, or any other functional physical or virtual network node in which a machine learning model may be implemented, as disclosed herein) of a communication network (e.g., communication network QQ100 configured to provide communication as described below with respect to FIG. 12) according to an embodiment of the concepts of the present disclosure. The computing device 900 may be provided as described below with respect to, for example, network nodes QQ110A, QQ110B of FIG. 12, network node QQ300 of FIG. 13, hardware QQ504, and / or virtual machines QQ508A, QQ508B of FIG. 15, all of which should be considered interchangeable in the examples and embodiments described herein, unless otherwise noted, and this is within the intended scope of the present disclosure. As shown, the computing device may include a computer 901 communicatively connected to an ML model 911. The ML model 911 may be an episodic RL agent. The ML model 911 may also include a reward computing unit (e.g., reward computing unit 705). The computer 901 may include a transceiver circuit (e.g., corresponding to a portion of the RF transceiver circuit QQ312 and the wireless front-end circuit QQ318 of FIG. 13, also referred to as a transceiver) including a transmitter and a receiver configured to provide uplink and downlink wireless communication with a mobile terminal. The computer may include a network interface circuit 907 (e.g., corresponding to a portion of the communication interface QQ306 of FIG. 13, also referred to as a network interface) configured to provide communication with other devices or nodes of a communication network (e.g., other network nodes, communication devices, and / or data repositories).The computing device may also include a processing circuit 903 (e.g., also referred to as a processor, corresponding to processing circuit QQ302 of FIG. 13) coupled to the transceiver circuit, and a memory circuit 905 (e.g., also referred to as a memory, corresponding to memory QQ304 of FIG. 13) coupled to the processing circuit. The memory circuit 905 may include computer readable program code that, when executed by the processing circuit 903, causes the processing circuit to perform operations according to embodiments disclosed herein. According to other embodiments, the processing circuit 903 may be defined to include a memory, such that a separate memory circuit is not required.

[0074] As described herein, operations of the computing device may be performed by the processing circuitry 903, the network interface 907, and / or the transceiver. For example, the processing circuitry 903 may control the transceiver to transmit downlink communications to one or more mobile terminals UE over the radio interface through the transceiver, and / or receive uplink communications from one or more communication devices over the radio interface through the transceiver. Similarly, the processing circuitry 903 may control the network interface 907 to transmit communications to one or more other devices or network nodes over the network interface 907, and / or receive communications over the network interface from one or more network nodes, radio heads, local batteries, communication devices, etc. Furthermore, modules may be stored in the memory 905, and these modules may provide instructions such that, when the instructions of the modules are executed by the processing circuitry 903, the processing circuitry 903 performs respective operations (e.g., operations described below with respect to exemplary embodiments related to the computing device). According to some embodiments, the computing device 900 and / or its elements / functionality may be embodied as a virtual node / nodes and / or virtual machines / machines, for example, as described with respect to FIG. 15 .

[0075] According to some other embodiments, the computing device may be implemented as a core network node without a transceiver. According to such embodiments, a transmission to a communication device, network node, radio head, local battery, etc. may be initiated by the computing device 900 such that the transmission to the communication device, network node, etc. is provided through the computing device 900 that includes a transceiver (e.g., through a base station or RAN node). According to embodiments in which the computing device is a RAN node that includes a transceiver, initiating the transmission may include transmitting through the transceiver.

[0076] The operation of a computing device (e.g., a computing device including an ML model 911) (implemented using the structure of FIG. 9) will now be discussed with reference to the flowcharts of FIGS. 10 and 11 according to some embodiments of the present disclosure. In the following description, the computing device may be any of the computing device 900, the network nodes QQ110A, QQ110B, QQ300, QQ606, the hardware QQ504, or the virtual machines QQ508A, QQ508B, but the computing device 900 shall be used to describe the functionality of the operation of the computing device. For example, modules may be stored in the memory 905 of FIG. 9, and these modules may provide instructions such that the processing circuitry 903 performs the respective operations of the flowcharts when the instructions of the modules are executed by the processing circuitry 903 of the respective computing device.

[0077] Referring to FIG. 10, a method is provided for managing the transmission of power from a local battery located in proximity to at least one radio head to at least one radio head, the method being performed by a computing device (900) in a communication system. The method includes making a decision (1001) for a future time window on the transmission of power from the local battery to at least one radio head. The decision is made by a machine learning model based on (i) a determination of a difference in output power data statistics including average power demand and peak power demand in a set of past time windows covering a defined period for the at least one radio head, and (ii) time- and location-dependent cost data of charging and / or discharging the local battery and grid utilization in the future time window. The method further includes outputting (1003) a decision on the transmission of power from the local battery to at least one radio head in the future time window.

[0078] According to some embodiments, the output power data statistics include at least (i) an average power demand of the at least one radio head over a set of past time windows, and (ii) a peak power demand of the at least one radio head over a set of past time windows.

[0079] According to some embodiments, the time-location dependent cost data includes a first cost for charging a local battery in a future time window and a second cost for using power from the grid for an average power demand of the at least one wireless head in the future time window.

[0080] According to some embodiments, the decision includes one of the following for a future time window (i) where power is transferred from the local battery to the at least one wireless head during the future time window, the future time window including a period of peak power consumption in the at least one wireless head, (ii) charging the local battery during the future time window, and (iii) not transferring power from the local battery to the at least one wireless head during the future time window.

[0081] According to some embodiments, the determination is made by a machine learning model based on (i) inputting the power data and the battery and cost data into the machine learning model, and (ii) determining a minimized total cost for transmitting power to the at least one wireless head for a future time window, where the future time window is constrained by multiple constraints.

[0082] According to some embodiments, the constraints include (i) a first constraint set as an input power consistency, where the input power to the radio head is equal to the discharge level of the local battery to the radio head plus the power from the grid, and (ii) a second constraint, where the discharge is set such that the amount of discharge from the local battery is less than or equal to the charge level of the local battery at the start of the future time window plus the charge profile limit of the future time window.

[0083] According to some embodiments, the charge profile limits for the future time windows include further constraints, including (i) a third constraint that the current charge level of the local battery is set such that the current charge level is equal to the previous (immediately preceding) charge level of the local battery plus the charge minus the discharge of the local battery in the current time window, and (ii) a fourth constraint that the charge level is set such that the charge level of the local battery in the set of time windows is less than or equal to the charge of the local battery.

[0084] According to some embodiments, the machine learning model receives reward feedback for state and action pairs. The states include a current charge level of a local battery, a current input power of at least one wireless head, and a current cost to charge the local battery. The action in the state and action pair includes a decision.

[0085] 11, according to some embodiments, the at least one radio head includes a plurality of radio heads. The method further includes dividing the defined period into a set of time windows (1101). The method further includes making a decision (1103) for each time window in the set of time windows, for each radio head.

[0086] Referring again to FIG. 10, according to some embodiments, outputting the decision (903) includes outputting a decision for controlling transmission of power to the at least one wireless head based on the decision.

[0087] According to some embodiments, the power data is offline data, and the determining (1001) and outputting (1003) is performed during training of a machine learning model that uses the offline data.

[0088] According to some embodiments, the power data is online data, the machine learning model is installed in the communication system, and the determination (1001) and output (1003) are performed by the installed machine learning model.

[0089] According to some embodiments, the computing device is located in one of a location proximate to a local battery and a cloud-based location.

[0090] Various operations from the flowchart of FIG. 11 may be optional with respect to some embodiments of the method performed by a computing device.

[0091] FIG. 12 illustrates an example of a communication network QQ100 in which a communication system according to some embodiments may be implemented.

[0092] In the illustrated example, the communication network QQ100 includes a telecommunications network QQ102 including an access network QQ104, such as a radio access network (RAN), and a core network QQ106 including one or more core network nodes QQ108. The access network QQ104 includes one or more access network nodes, such as network nodes QQ110a and QQ110b (one or more of which may be generally referred to as network node QQ110), or any other similar third generation partnership project (3GPP®) access node or non-3GPP® access point. The network node QQ110 facilitates direct or indirect connection of user equipment (UE), such as by connecting UE QQ112a, QQ112b, QQ112c, and QQ112d (one or more of which may be generally referred to as UE QQ112) to the core network QQ106 via one or more wireless connections. The network nodes QQ110A, QQ110B may be cloud-implemented network nodes (e.g., servers) or may be located in the cloud or edge-implemented network nodes (e.g., servers). The network nodes QQ110A, QQ110B facilitate direct or indirect connection of communication devices QQ112a, QQ112b, QQ112c, and QQ112d (one or more of which may be generally referred to as communication devices / UEs QQ112) to the communication network QQ100 via one or more wireless connections.

[0093] Exemplary wireless communication over wireless connections includes transmitting and / or receiving radio signals using electromagnetic waves, radio waves, infrared, and / or other types of signals suitable for conveying information without the use of wires, cables, or other conductive materials. Furthermore, according to various embodiments, communication network QQ100 may include any number of wired or wireless networks, network nodes, communication devices, and / or any other components or systems that may facilitate or participate in the communication of data and / or signals, whether via wired or wireless connections. Communication network QQ100 may include and / or interface with any type of communication, telecommunication, data, cellular, wireless networks, and / or other similar types of systems, including but not limited to 5G and / or 6G networks.

[0094] In the illustrated example, the core network QQ106 connects the network node QQ110 to one or more hosts, such as the host QQ116. These connections may be direct or indirect via one or more intermediate networks or devices. In other embodiments, the network nodes may be directly coupled to the hosts. The core network QQ106 includes one or more core network nodes (e.g., the core network node QQ108) that are comprised of hardware and software components. Features of these components may be substantially similar to those described with respect to the UEs, network nodes, and / or hosts, and therefore, those descriptions are generally applicable to the corresponding components of the core network node QQ108. Exemplary core network nodes include one or more of the following functions: a Mobile Switching Center (MSC), a Mobility Management Entity (MME), a Home Subscriber Server (HSS), an Access and Mobility Management Function (AMF), a Session Management Function (SMF), an Authentication Server Function (AUSF), a Subscription Identifier Unhiding Function (SIDF), a Unified Data Management (UDM), a Security Edge Protection Proxy (SEPP), a Network Exposure Function (NEF), and / or a User Plane Function (UPF).

[0095] Host QQ 116 may be under the ownership or control of a service provider other than the operator or provider of access network QQ 104 and / or telecommunications network QQ 102, and may be operated by or on behalf of the service provider. Host QQ 116 may host a variety of applications to provide one or more services. Examples of such applications include live and pre-recorded audio / video content, data collection services such as searching and compiling data on various ambient conditions detected by multiple UEs, analytics functions, social media, functions for controlling or otherwise interacting with remote devices, functions for an alarm and monitoring center, or any other such functions performed by a server.

[0096] Overall, the communication network QQ100 of Figure 12 enables connectivity between UEs, network nodes, and hosts. In that sense, the communication network can be configured to operate according to defined rules or procedures, including, but not limited to, Global System for Mobile Communications (GSM), Universal Mobile Telecommunications System (UMTS), Long Term Evolution (LTE), and / or other suitable 2G, 3G, 4G, 5G standards, or applicable future generation standards (e.g., 6G), Wireless Local Area Network (WLAN) standards, such as the Institute of Electrical and Electronics Engineers (IEEE) 802.11 standard (WiFi), and / or other suitable wireless communication standards, such as Worldwide Interoperability for Microwave Access (WiMax), Bluetooth, Z-Wave, Near Field Communication (NFC), ZigBee, LiFi, and / or any Low Power Wide Area Network (LPWAN) standards, such as LoRa and Sigfox.

[0097] In some examples, the telecommunications network QQ102 is a cellular network that implements 3GPP standardized features. Thus, the telecommunications network QQ102 may support network slicing to provide different logical networks to different devices connected to the telecommunications network QQ102. For example, the telecommunications network QQ102 may provide Ultra-Reliable Low-Latency Communication (URLLC) services to some UEs, while providing Enhanced Mobile Broadband (eMBB) services to other UEs, and / or Massive Machine Type Communication (mMTC) / Massive IoT services to additional UEs.

[0098] In an example, the hub QQ114 communicates with the access network QQ104 to facilitate indirect communication between one or more UEs (e.g., UEs QQ112c and / or QQ112d) and a network node (e.g., network node QQ110b). In some examples, the hub QQ114 may be a controller, a router, a content source and analysis, or any of the other communication devices described herein with respect to a UE. For example, the hub QQ114 may be a broadband router that enables access to the core network QQ106 for the UE. As another example, the hub QQ114 may be a controller that sends commands or instructions to one or more actuators in the UE. The instructions or instructions may be received from the UE, the network node QQ110, or by executable code, scripts, processes, or other instructions in the hub QQ114. As another example, the hub QQ114 may be a data collector that serves as a temporary store for UE data and may perform analysis or other processing of the data according to some embodiments. As another example, the hub QQ114 may be a content source. For example, for a UE that is a VR headset, display, loudspeaker, or other media distribution device, Hub QQ 114 may retrieve data related to VR assets, video, audio, or other media, or sensory information via a network node, which Hub QQ 114 then provides directly to the UE, either after performing local processing and / or adding additional local content. In yet another example, Hub QQ 114 acts as a proxy server or orchestrator for the UEs, especially when one or more of the UEs are low energy IoT devices.

[0099] Hub QQ114 may have a constant / persistent or intermittent connection to network node QQ110b. Hub QQ114 may also enable other communication schemes and / or schedules between hub QQ114 and UEs (e.g., UEs QQ112c and / or QQ112d) and between hub QQ114 and core network QQ106. In other examples, hub QQ114 is connected to core network QQ106 and / or one or more UEs via a wired connection. Furthermore, hub QQ114 may be configured to connect to an M2M service provider via access network QQ104 and / or to another UE via a direct connection. In some circumstances, a UE may establish a wireless connection with network node QQ110 while still being connected via hub QQ114 via a wired or wireless connection. According to some embodiments, hub QQ114 may be a dedicated hub, i.e., a hub whose main function is to route communications to / from network node QQ110b to UEs. According to other embodiments, hub QQ114 may be a non-dedicated hub, i.e., a device capable of operating to route communications between UEs and network node QQ110b, but further capable of operating as a communications origination and / or termination point for a particular data channel.

[0100] FIG. 13 illustrates a network node QQ300 according to some embodiments (e.g., the network node QQ300 may include a network node capable of implementing a computing device 900). As used herein, a network node refers to a configured, arranged, and / or operative device that can communicate directly or indirectly with UEs and / or other network nodes or devices in a communication network. Examples of network nodes include, but are not limited to, a server, an access point (AP) (e.g., a wireless access point), a base station (BS) (e.g., a radio base station, a Node B, an evolved Node B (eNB), and an NR Node B (gNB)).

[0101] Base stations may be classified based on the size of the coverage they provide (or, stated differently, their transmission power level) and may therefore be referred to as femto, pico, micro, or macro base stations depending on the size of the coverage provided. A base station may be a relay node or a relay donor node that controls a relay. A network node may also include one or more (or all) parts of a distributed radio base station, such as a centralized digital unit and / or a remote radio unit (RRU), sometimes referred to as a remote radio head (RRH). Such remote radio units may or may not be integrated with an antenna as an antenna-integrated radio. Parts of a distributed radio base station may also be referred to as nodes in a distributed antenna system (DAS).

[0102] Other examples of network nodes include a multi-transmission point (multi-TRP) 5G access node, a multi-standard radio (MSR) equipment such as an MSR BS, a network controller such as a radio network controller (RNC) or a base station controller (BSC), a base transceiver station (BTS), a transmission point, a transmitting node, a multi-cell / multicast coordination entity (MCE), an operation and maintenance (O&M) node, an operation support system (OSS), a self-organizing network (SON) node, a positioning node (e.g., an evolved serving mobile location center (E-SMLC)), a minimization of drive testing (MDT), and / or a cloud- or edge-implemented server.

Claims

1. 1. A method for managing transmission of power to at least one radio head from a local battery located proximate to the at least one radio head, the method being performed by a computing device (900) in a communication system, the method comprising: Making a decision on power transmission from the local battery to the at least one radio head in a future time window (1001), the decision being made by a machine learning model based on (i) determining a difference in output power data statistics including average power demand and peak power demand in a set of past time windows covering a defined period for the at least one radio head, and (ii) time and location dependent cost data of charging and / or discharging of the local battery and grid utilization in the future time window, outputting (1003) the decision on transmitting power from the local battery to the at least one wireless head in the future time window; A method comprising:

2. 2. The method of claim 1, wherein the output power data statistics include at least: (i) an average power demand of the at least one radio head in the set of past time windows; and (ii) a peak power demand of the at least one radio head in the set of past time windows.

3. 3. The method according to claim 1, wherein the time and location dependent cost data comprises a first cost for charging the local battery in the future time window and a second cost for using power from a power grid for an average power demand of the at least one wireless head in the future time window.

4. 4. The method of claim 1 , wherein the determination for the future time window includes one of: (i) transmitting power from the local battery to the at least one radio head during the future time window, where the future time window includes a period of peak power consumption in the at least one radio head; (ii) charging the local battery during the future time window; and (iii) not transmitting power from the local battery to the at least one radio head during the future time window.

5. 5. The method of claim 1 , wherein the determination is made by the machine learning model based on: (i) inputting the power data and the battery and cost data into the machine learning model; and (ii) determining a minimized total cost for transmitting power to the at least one wireless head for the future time window if the future time window is constrained by a number of constraints.

6. 6. The method of claim 5, wherein the plurality of constraints include (i) a first constraint set as an input power consistency such that the input power to the radio head is equal to a discharge level of the local battery to the radio head plus power from the grid, and (ii) a second constraint set as the discharge from the local battery being less than or equal to a charge level of the local battery at a start of the future time window plus a charge profile limit for the future time window.

7. 7. The method of claim 6, wherein the charge profile limit for the future time window includes further constraints including: (i) a third constraint setting the current charge level of the local battery to be the same as a previous charge level of the local battery plus a charge amount minus a discharge amount of the local battery in the current time window; and (ii) a fourth constraint setting the charge level of the local battery to be less than or equal to a charge capacity of the local battery in any one window in the set of time windows.

8. The method according to any one of claims 1 to 7, the machine learning model receives reward feedback for state and action pairs; the reward feedback being a value that minimizes the total cost of input power to the at least one radio head in a next window in the set of time windows; the state includes a current charge level of the local battery, a current input power of the at least one wireless head, and a current cost to charge the local battery; The method, wherein the action in the state and action pair includes the decision.

9. 9. The method according to claim 1, wherein the at least one radio head comprises a plurality of radio heads, further comprising: Dividing the defined period into a set of time windows (1101); making a decision for each radio head for each time window in the set of time windows (1103); A method comprising:

10. 10. The method of claim 1, wherein outputting the decision (1003) comprises outputting the decision for controlling the transmission of power to the at least one radio head based on the decision.

11. The method according to any one of claims 1 to 10, the power data is offline data; The method of claim 1, wherein the making of the determination (1001) and the outputting (1003) are performed during training of the machine learning model with the offline data.

12. The method according to any one of claims 1 to 11, The power data is online data, The machine learning model is installed in a communication system; The method, wherein the making of the decision (1001) and the outputting (1003) are performed by the installed machine learning model.

13. The method of any one of claims 1 to 12, wherein the computing device is located at one of a location proximate to the local battery and a cloud-based location.

14. A computing device (900) in a communication system for managing the transmission of power to at least one radio head from a local battery located proximate to the at least one radio head, the computing device comprising: At least one processor (903); and at least one memory (905) coupled to said at least one processor (903) and configured to store program code that is executed by said at least one processor to perform operations, said operations including: making a decision on transmitting power from the local battery to the at least one radio head in a future time window, the decision being made by a machine learning model based on (i) determining a difference in output power data statistics including average power demand and peak power demand in a set of past time windows covering a defined period for the at least one radio head, and (ii) time and location dependent cost data of charging and / or discharging of the local battery and grid utilization in the future time window; outputting the decision about transmitting power from the local battery to the at least one wireless head in the future time window; a computing device comprising:

15. A computing device (900) according to claim 14, wherein the at least one memory (905) is connected to the at least one processor (903) and stores program code executed by the at least one processor to perform the operations according to any one of claims 2 to 13.

16. A computing device (900), the computing device being in a communication system for managing transmission of power to at least one radio head from a local battery located proximate to the at least one radio head, the computing device being adapted to perform operations, the operations including: making a decision on transmitting power from the local battery to the at least one radio head in a future time window, the decision being made by a machine learning model based on (i) determining a difference in output power data statistics including average power demand and peak power demand in a set of past time windows covering a defined period for the at least one radio head, and (ii) time and location dependent cost data of charging and / or discharging of the local battery and grid utilization in the future time window; outputting the decision about transmitting power from the local battery to the at least one wireless head in the future time window; a computing device comprising:

17. A computing device according to claim 16, adapted to perform the operations according to any one of claims 2 to 13.

18. 1. A computer program comprising: program code executed by a processing circuit (903) of a computing device (900) in a communication system for managing transmission of power to at least one radio head from a local battery located proximate to the at least one radio head, whereby execution of the program code causes the computing device to perform operations, the operations comprising: making a decision on transmitting power from the local battery to the at least one radio head in a future time window, the decision being made by a machine learning model based on (i) determining a difference in output power data statistics including average power demand and peak power demand in a set of past time windows covering a defined period for the at least one radio head, and (ii) time and location dependent cost data of charging and / or discharging of the local battery and grid utilization in the future time window; outputting the decision about transmitting power from the local battery to the at least one wireless head in the future time window; A computer program comprising:

19. A computer program product according to claim 18, wherein execution of said program code causes said computing device (900) to perform the operations according to any one of claims 2 to 13.

20. 1. A computer program product comprising: a non-transitory storage medium comprising program code executed by a processing circuit (903) of a computing device (900) in a communication system, the program code comprising: managing transmission of power to at least one radio head from a local battery located proximate to the at least one radio head, whereby execution of the program code causes the computing device to perform operations, the operations including: making a decision on transmitting power from the local battery to the at least one radio head in a future time window, the decision being made by a machine learning model based on (i) determining a difference in output power data statistics including average power demand and peak power demand in a set of past time windows covering a defined period for the at least one radio head, and (ii) time and location dependent cost data of charging and / or discharging of the local battery and grid utilization in the future time window; outputting the decision about transmitting power from the local battery to the at least one wireless head in the future time window; 2. A computer program product comprising:

21. 21. A computer program product according to claim 20, wherein execution of said program code causes said computing device (900) to perform the operations according to any one of claims 2 to 13.

Citation Information

Patent Citations

  • Power demand adjustment device, power demand adjustment method and power demand adjustment program

    JP2016220384A

  • Power management device

    JP2017022918A

  • Method and apparatus for operating a smart system for optimizing power consumption - Patents.com

    JP2017516247A

  • Attribute estimation device, attribute estimation method, and attribute estimation program

    JP2018038206A

  • Power supply control system

    JP2018201267A