WO2013040739A1

WO2013040739A1 - System and method for communication in a cellular network

Info

Publication number: WO2013040739A1
Application number: PCT/CN2011/079807
Authority: WO
Inventors: Chunyi Peng
Original assignee: Peking University; LV, Songwu; Wang, Tao
Priority date: 2011-09-19
Filing date: 2011-09-19
Publication date: 2013-03-28
Also published as: CN103947237B; CN103947237A

Abstract

A grid-based location-dependent modeling scheme is provided to save power consumption in a cellular network. It exploits the traffic dynamics in a cellular network and completely powers off under-utilized BSs to save energy. In the novel scheme, the network is divided into one or more grids, each grid having one or more nodes that are equivalent to each other. The aggregate traffic for all nodes of a defined period in each gird is then estimated, and a set of active nodes of each grid for the defined period is determined to meet the estimated aggregate traffic, and the nodes in the grid that are not active are powered off.

Description

SYSTEM AND METHOD FOR COMMUNICATION IN A CELLULAR

NETWORK

TECHNICAL FIELD

The present invention relates to the field of communication network, and more particularly, to a system and a method for communication that is capable of saving power consumption in a cellular network.

BACKGROUND Cellular infrastructure is currently experiencing energy surge in the world. Recent reports show that energy consumption of mobile network would reach 124.4B KWh in 2011 , and the power bill is expected to double every 4~5 years until 2030. The accumulated energy bill within the time span of 5~6 years is equivalent to the total deployment cost of national 3G/4G networks, as shown by one of the largest 3G operators in the world.

SUMMARY OF THE INVENTION

The present invention proposes a system and a method for communication that is capable of saving power consumption in a cellular network. In an aspect of the invention, there is proposed a system for communication in a cellular network, comprising: a grid establishment unit configured to divide the network into one or more grids, each grid having one or more nodes arranged therein that are equivalent to each other, a grid traffic estimation unit configured to estimate aggregate traffic for all nodes of a defined period in each gird; and an active node set determination unit configured to determine a set of active nodes of each grid for the defined period based on the estimated aggregate traffic, and power off the nodes in the grid that are not active. In another aspect of the invention, there is proposed a method for communication in a cellular network, comprising: dividing the network into one or more grids, each grid having one or more nodes arranged therein that are equivalent to each other, estimating aggregate traffic for all nodes of a defined period in each gird; and determining a set of active nodes of each grid for the defined period based on the estimated aggregate traffic, and powering off the nodes in the grid that are not active.

In still another aspect of the invention, there is proposed a computer readable program for, then running on a computer, implementing the method.

BRIEF DESCRIPTION OF THE DRAWINGS

The above and other objects, features and advantages of the present invention will be clearer from the following detailed description about the non-limited embodiments of the present invention taken in conjunction with the accompanied drawings, in which:

Figure 1 is a diagram showing structure of a typical BS in a 3G UMTS network;

Figure 2 shows an illustrative BS location in four typical regions;

Figure 3 shows the spatial traffic diversity among neighboring BSs; Figure 4 shows the traffic load at four individual BSs of several days in different regions;

Figure 5 is a schematic block diagram to show the respective units operating in a system according to an embodiment of the present invention;

Figure 6 shows an example of a grid establishment result; Figure 7 illustrates how the BS selection algorithm works according to an embodiment of the present invention; and

Figure 8 is a flowchart to show the respective steps of a method according to an embodiment of the present invention. DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS

Hereunder, the present invention will be described in accordance with the drawings. In the following description, some embodiments are used for the purpose of description only, which shall not be understood as any limitation to the 5 present invention but the examples thereof. While it may blur the understanding of the present invention, the conventional structure or construction will be omitted.

To build a green cellular network, it may be more effective to improve the most critical node that is the dominant contributing factor to overall energy consumption. In the 3Gstandard context, the critical node may be the base station (BS). BSs i o consume about 80% of overall infrastructure energy, while the User Equipments (UEs) typically take around 1 % of overall infrastructure energy. BS is taken as a target for achieving the power consumption reduction effect in a 3G network in the description below. However, it is obvious that the invention is not limited to the BS and the 3G network. The invention is applicable to other forms of cellular networks

15 and critical nodes of such cellular networks.

Our analysis reveals that, traffic load of a 3G network exhibits wide-range fluctuations both in time and over space. However, energy consumption of current networks is not load adaptive. The used energy is unproportionally large under light traffic. The underlying cause is that each BS is not energy proportional, with 20 more than 50% spent on cooling, idle-mode signaling and processing, which are not related to the runtime traffic load.

[BS Power Consumption]

Figure 1 is a diagram showing structure of a typical BS in a 3G UMTS network. BS of a 3G UMTS network typically has a communication subsystem and a supporting

25 subsystem. The communication subsystem includes Remote Radio Unit (RRU), Base Band Unit (BBU), and Feeder. RRU is the radio specific hardware for each sector. Each BS may install several RRUs near antennas to provide different coverage and capacity. BBU, as the main unit, provides all other communication functions, including controlling, baseband processing, switching and lub interfaces

30 to Radio Network Controller (RNC). Each BS may have several BBUs. Feeder is the optical-fiber pair of cable that connects RRUs to BBUs. The supporting subsystem includes the cooling subsystem and other auxiliary devices. The cooling subsystem, including air conditioning and fans, maintains an appropriate operation temperature at the BS. The auxiliary devices include power supply and environment monitoring modules. From the energy efficiency perspective, the cooling subsystem and some transmission modules consume a significant portion of overall power at each BS, regardless of the traffic load intensity. Our measurement shows that it reaches 50% or more in an operational BS. This is a main factor that leads to energy inefficiency for the 3G infrastructure.

The total power consumption P at a BS may be given by

P~Ptx P misc,

Where the first part P_tx accounts for power used to provide network access to mobile clients. It includes power consumed by RRUs, BBUs, feeder and RNC transmission. The second part P_m/sc records the auxiliary power for cooling, power supply and monitoring. P_fx mainly changes with the carried load while P_misc typically remains constant given a fixed operation environment. A linear model can approximate P_fx. Two dominant components in P_tx are the power consumed by RRUs and BBUs. When the traffic load is heavy, RRU has to spend more power to support more active links. The power consumed by RRU, therefore, increases proportionally as the traffic load. On the other hand, BBU does baseband processing for all frequency carriers used by the BS. No matter how many links are active, its power consumption is mainly determined by the number of carriers unless it is in a sleep mode.

Cooling power consumption is a dominant factor in P_m/sc based on real measurement. It depends on the amount of the extracted heat and the desired operating temperature. It also varies with chillers that use a variety of compressors and drivers. Cooling may consume about 50% power at BSs. Cooling power mainly depends on the temperature.

In summary, our analysis shows that BS is not energy proportional to its traffic load, mainly due to the part P_{m sc}.

[Traffic Load: Diversity in Time and Space]

Each BS in a cellular network exhibits high traffic dynamic over time and across locations. Figure 2 shows an illustrative BS location in four typical regions, Region 1 is a large, populous city, Region 2 is a medium-size city, and Regions 3 and 4 are large cities in a large metropolitan area. The coverage area and the number of BSs in each region are given in Table 1 below, which are collected from an operational 3G network.

In hot spots of a city (e.g., subregion B), more BSs are provisioned, thus creating location-dependent diversity. The BS deployment density is quite diverse across different regions, as well as in the same region. As shown in Figure 2, a large number of BSs have multiple neighbors, especially in Regions 1 and 3. For example, for more than half of BSs in Region 1 , each has at least 10 neighbors within its 1 Km range. In contrast, Region 4 has the most sparse deployment; only 40% BSs have multiple neighbors. The dense BS deployment is partly due to the current practice that operators mostly ignore the traffic multiplexing effect. It is further observed that traffic load intensity be quite diverse even in each local neighborhood (i.e., traffic loads among the closely located BSs). Figure 3 shows the spatial traffic diversity among neighboring BSs. Each point represents, at any given time of the day, the traffic-volume ratio of the maximum-traffic BS and the minimum-traffic BS within 1 Km range of each BS in four regions. From Figure 3, it can be seen that max-to-min traffic ratio is larger than 5 in 50% cases, and larger than 10 in 30% cases. It can also be observed that such neighborhood-scale spatial traffic diversity be more evident during the peak time of a day.

Figure 4 shows the traffic load at four individual BSs of several days in different regions. Strong diurnal patterns can be observed on both daily and weekly basis, alternating between heavy-traffic and light-traffic durations. It also can be seen that the traffic patterns for weekend and for weekday are different, but those for weekend are similar, and those for weekday are similar.

We also observe that the traffic load be stable over the short term (e.g., the same time of consecutive days), while it may slowly evolve over a long term (e.g., 26% global increase in 2010). Although the traffic load fluctuates over time, the time of the day traffic load at each BS is quite stable over consecutive days. For example, BS 1 has a similar traffic load at 5pm on Days 1 and 2, Days 2 and 3, and so on.

[A Grid-based Location-dependent Modeling Scheme]

All the observations said above are taken into account, and a grid-based location-dependent modeling scheme and a novel system and method for implementing the scheme are proposed, to reduce power consumption in the network.

Figure 5 is a schematic block diagram to show the respective units operating in a system according to an embodiment of the present invention. Referring to Figure 5, the system 50 according to the embodiment comprises a grid establishment unit 510, a grid traffic estimation unit 520, and an active node set determination unit 530. The respective units are described in details as follows.

The grid establishment unit 510 divides the entire network into grids, so that BSs in each grid are equivalent. BSs are equivalent if they can replace each other when serving user equipments. Location information and transmission range of each BS may be used to decide whether BSs in spatial proximity are equivalent or not. Location coordinates can be obtained by GPS or other location systems when operators plan and deploy their infrastructure. Transmission range of BS may vary from 200m to 1 Km in cities and from 1 Km to 5Km in rural area. It may be different among BSs due to antenna configuration and replacement, transmit power and environment. Two BSs i and j are equivalent if ri + d(i, j) < Rj, rj + d(i, j) < Ri, where d(i, j) is the distance between BS i and BS j, ri and rj are the normal communication ranges of BSs, and Ri and Rj are the maximum possible communication ranges of BS i and BS j, respectively. Deploy density may vary, which is reflected by changing distance d(i, j).

Figure 6 shows an example of a grid establishment result, where the dotted circles in left part denote the coverage of the central BS, the circled numbers in right part denote respective BSs, r1 and r4 show the radii of BS 1 and BS 4 respectively, and R1 and R4 show the distances from BS1 to BS 4 and from BS1 to BS 6. In the figure, BS 1 is equivalent to BSs 2 and 3, but is not equivalent to BS 4. Accordingly, three grids are established, one including BSs 1 -3, the second including BSs 4 and 5, and the remaining one including BS 6 only.

A virtual grid is formed when all BSs in it are equivalent. Once a BS is not equivalent to every BS in the current grid, the grid establishment unit 510 forms a new grid. Different grid constructions may be formed when the grid establishment unit 510 starts with a different BS and towards to a different direction. In an embodiment, a simple heuristic "northwest rule" may be used to decide grid construction. It starts from the northwest corner in the BS deployment map (i.e., top-left corner in the network deployment), clusters all equivalent BSs from top to down and from left to right, and generates a new grid when a BS is found to not be equivalent to at least one BS in the current grid. The process repeats until the southeast corner is reached and all the BSs in the network are included in grids. In the illustrative example of Figure 6, three grids are thus formed following this rule. It is obvious that formation along other directions may generate different virtual grids, but would not much affect the goal of the embodiment. No matter what construction is formed, it does not change the inherent proximity. Close nodes (e.g., BS in the embodiment) belong to the same grid with high probability. For example, if we form the grid in a "northeast" rule (i.e., top-right first), three different grids may be formed: one including BSs 6 and 5, the second including BSs 4 and 3, and the third one including BSs 2 and 1. After the grid construction is established by the grid establishment unit 510, the grid traffic estimation unit 520 may estimate the aggregate traffic in each grid.

As described above and shown in Figure 4, the traffic load for BS is stable over a short term. A statistical scheme may be designed to estimate the aggregate traffic in a grid. Firstly, each day is divided into 24 hourly periods, the statistics of each hourly period is computed, and the aggregate traffic for the given hour is derived from the statistics. In an embodiment, the weekday may be differently treated from a weekend day, but all weekdays or weekend days are treated similarly. Holidays can be taken as weekend in another embodiment. Specifically, for /-th hour of k-th day that we stack together consecutive weeks, we compute the average S{i, k) and standard deviation D(i, k) as follows:

S(i, k) = (\ - a) - S(i, k - l) + a - S(i, k) ,

D{i, k) = (1 - ) - D(i, k - \) + fi- \ S{i, k) - S(i, k) I , where S(i, k) is the hourly sample value of the aggregate traffic in the grid for /^'-th hour during the /c-th day, and a , β are the smoothing parameters. Consequently, we estimate the hourly aggregate traffic as EV(i, k) = S(i, k) + γ^■ D(i, k) where γ is a design parameter that offers a tuning switch to balance between tight estimate and miss ratio. In our prototype, we choose a = 0.125, β = 0.25 and χ = 3. A skilled in the art can recognize that other parameters values can be applied based on the performance tradeoffs.

An alternative approach is to first obtain statistics of each individual BS and then sum up all in a grid as the grid statistics. It estimates each individual BS traffic load without extracting the multiplexing effect of traffic load among the BSs in a grid. The grid-based modeling may improve energy efficiency when traffic load is heavy.

In an embodiment, the short term for estimating the aggregate traffic may be half hour instead of one hour, or two hours. It is clear that the computation complexity and effect may be different if the period is set differently, but it does not change much the goal of the invention. The active node set determination unit 530 determines a set of active BSs for each grid based on the estimated aggregate traffic, and powers off under-utilized BSs. The power off of the under-utilized BSs shall not negatively affect the coverage and capacity requirements so that the network operates properly.

Given the 24-hour traffic statistics at a given grid, we may find the hour(s) with heaviest traffic. For this peak period, a set of active BSs in the grid is determined, denoted by S_max. Based on the fact that the residual energy at BS contributes to a large portion of energy that is irrelevant to radio transmission, the number of active BSs shall be reduced as much as possible so as to save energy. On the other hand, the aggregate capacity of the active BSs in the set has to be large enough to accommodate the aggregate traffic that has been estimated in the grid traffic estimation unit 520. In an embodiment, the BSs with larger capacity or higher energy efficiency are selected with higher priority. All the BSs in a grid are ranked in the decreasing order of their capacity values C(BSi), i.e. , C(BS₁)≥C{BS2), ... ,≥C(BS_n), when these BSs have homogeneous power models. The number m of active BSs of largest capacities are selected so that

∑™₌ (BS_k )≥EV_mm . The set of active BSs for the peak hour S_max is given by

S_ma = {BS_],..., BS . The remaining BSs that are not selected to be active are powered off so that the goal of saving power consumption in the network can be reached. This selection process ensures the minimum number of active BSs in the grid. Assume that all BSs in a grid use same power models, it can easily prove that the process is optimal to ensure minimum total energy in the grid. In an embodiment, BSs may have heterogeneous power models. In such a case, the high-energy-efficiency BSs are selected with higher priority if their capacity exceeds the traffic demand. The active node set determination unit 530 repeats the above process for each grid in the network, thus the set of active BSs for each grid during this heaviest traffic hour are obtained. It shall be noted that the heaviest traffic hour in different grid may be different.

It is further realized that frequent on/off switching of BS may be not desired for a couple of reasons, such as reduced lifetime of the cooling subsystem, negative effect on user equipments due to long ramp up time upon power on, and others. In order to minimize the number of on/off operations and reduce energy inefficiency, a continuous selection may be devised for the rest of the day. For the second heaviest traffic hour, active set is selected only from the superset S_max that is calculated for the heaviest traffic hour, rather than from all candidates in the gird. Similar selection policy is used to find the active set for all the other hours of a day. Figure 7 illustrates how the algorithm works, where S_max has 11 BSs and S_m/_n has 3 BSs. During the ramp-up transition from the idle hour with lightest traffic to the peak hour with heaviest traffic, we use an active set Sj at hour t, which is always a subset of S_max but a superset of the previous hour S_{i- 7}. A series of active sets S{f} is found that satisfies

S_mm = c S(i_t ) c S(/₂ )... <= S(t_p) = S_ma .where /, < t, < t₂ < ... < i_p denotes the hourly sequence from the idle hour to the peak hour. When migrating from hour t-1 to f, it only needs to power on those BSs not in S_f. _?, but keep all active BSs in S_f._/on. If S_f-ί is sufficient, it does not need to power on new BSs. Once a BS appears in S_t-i, it remains to power on at t and continues to appear in S_f. In the embodiment, BSs 4-10 will switch on sequentially based on the prediction of next hourly traffic from 6:00 am to 17:00, and switch off sequentially from 20:00 to 2:00 of the next day, as shown in Figure 7.This algorithm works well in case that the traffic for the hourly periods in the day increases or decreases orderly. In such cases, as shown in Figure 7, each BS is needed to be switched on or off at most once during each 24-hour duration. In case that there are two or more heaviest hours or two or more lightest traffic hours in a day, for example, two heaviest hours on 10:00 am and 3:00 pm, the selection algorithm will change to stick the same set of active BSs of a hour with neighboring hours as much as possible, to reduce the number of on/off switching of BSs. The active node set determination unit 530 may determine the sets of active BSs according to distribution of traffic of a whole day so that intersection of the sets for neighboring periods is as large as possible.

In an alternative, the sleeping BS is powered on ahead of the expected working time. It gives enough time for the cooling system to adjust the ambient temperature inside the sleeping BS. In another alternative, it always reserves a fraction (such as, 10%) of the capacity in a BS to be prepared for the worst-case scenario when determining the active BSs.

The proposed grid-based location-dependent modeling scheme is not limited to the system shown in Figure 5, and a system of different structure can also be used to implement the scheme. For example, the grid establishment unit 510 may be embodied as a table in RNC that is stored during the network deployment. The operator that deploys the network infrastructure knows the attributes of each BS, and may store the equivalence relationship for the deployed BSs in RNC in advance for later use. In another embodiment, the processor in the RNC may be designed to execute all the functions of the respective units of the system 50. Figure 8 is a flowchart to show the respective steps of a method according to an embodiment of the present invention. The method starts from step 810 where the network is divided into grids, so that BSs in each grid are equivalent to each other. After that, the method proceeds to step 820 where aggregate traffic of a period in each grid is estimated. Finally, in step 830, the set of active BSs for each grid is determined to meet the estimated aggregate traffic.

[Embodiment of 3G network Implementation]

The proposed grid-based location-dependent modeling scheme is standard compliant. Next, the implementation of the scheme in a 3Gnetwork is described in details.

In our scheme, after the grid construction is established and sets of active BSs are determined, the under-utilized BSs will be powered off during light-traffic period to save energy consumption. In such case, the active BSs need to extend their coverage to serve clients originally covered by the neighboring BSs that are powered off. "Cell Breathing" technique is well known in today's 3G network that can adjust cell boundaries. Cell breathing is traditionally used to adjust the cell size based on the number of client requests to achieve load balancing or capacity increase through micro-cell splitting. It can be used in the present scheme to the alternative purpose of power savings. Specifically, the effective service area may expand and contract according to the energy-saving requirement. By increasing the cell radius, an active BS can effectively extend the coverage area to neighboring BSs.

An alternative solution to cell breathing is to use dual BBU/RRU subsystems at a BS and switch between these two subsystems when adjusting the coverage area at peak or idle hours. For example, a transmission subsystem that works for a city area and another transmission subsystem that works for rural areas may be installed in a BS. Coverage provided by the BS can be adjusted by switching between these two subsystems. Another alternative is to use lower frequency bands at a given BS and extend its communication range.

Another practical issue that has to address for the embodiment of 3G network Implementation is how to effectively migrate existing user equipments from the about-to-power-off BSs to other active BSs. A network-controlled handoff (NCHO) mechanism in 3G standard may be used to with respect to this issue. For each active UE in an original BS (OBS), the following procedure is performed: (1 ) the OBS sends a handoff request to the neighboring active BS (ABS) via RNC; (2) the ABS acknowledges the handoff request and reserves resources for the migrating UE; (3) Upon receiving the handoff ACK from the ABS, the OBS sends the UE a handoff command; (4) the UE executes the handoff command via new association with the ABS. The OBS will defer its power-off if some UEs are still associated with it. In case of handoff failures, the OBS may repeat the above procedure with other active BSs until all UE handoffs succeed. Accordingly, the migration process in our power-saving mechanism can be readily made 3G standard compliant.

In the proposed scheme, BSs in the grid exchange traffic information to compute the aggregate traffic. A natural place to exchange such information is via the RNCs. Furthermore, in the handoff procedure said above, the OBS and ABS will exchange handoff request and UE information via RNC to complete the procedure. In an embodiment, BSs belonging to the same grid own the same RNC. In a case that BSs within a grid belong to different RNCs, information exchange between these RNCs is needed.

While our proposed scheme typically gives a reliable estimate on the aggregate traffic, rare case of traffic surge may occur. In order to prepare for such surges, each active BS is configured to monitor its traffic load. Whenever it sees sudden surge well beyond the traffic specified by the estimated traffic, it notifies its RNC of such case. The RNC may subsequently trigger emergency power on for the neighboring power-off BSs. The power-on number of BSs depends on the traffic surge volume the RNC is notified. In an embodiment where it is the RNC that estimates the aggregate traffic for grids and determines the active BSs, the transient traffic volume monitoring may be performed on the RNC instead of the active BSs.

Other arrangements of embodiments of the invention that are disclosed herein include software programs or firmware or hardware or configurations in other computing systems to perform the method embodiment steps and operations summarized above and disclosed in details below. More particularly, a computer program product is one embodiment that has a computer-readable medium including computer program logic encoded thereon that when performed in a computerized device provides associated operations implementing the grid-based location-dependent modeling scheme as explained herein. The computer program logic, when executed on at least one processor or other computing resource with a computing system, causes the processor or the computing resource to perform the operations (e.g., the methods) indicated herein as embodiments of the invention. Such arrangements of the invention are typically provided as software, code and/or other data structures arranged or encoded on a computer readable medium such as an optical medium (e.g., CD-ROM), floppy or hard disk or other a medium such as firmware or microcode in one or more ROM or RAM or PROM chips or as an Application Specific Integrated Circuit (ASIC) or as downloadable software images in one or more modules, shared libraries, or configurations in other computing systems, etc. The software or firmware or hardware or other configurations can be installed onto a computerized device to cause one or more processors in the computerized device to perform the techniques explained herein as embodiments of the invention. Software processes, firmware, hardware or configurations that operate in a collection of computerized devices, such as in a group of data communications devices or other entities can also provide the system of the invention. The system of the invention can be distributed between many software processes, firmware, hardware or configurations on several data communications devices, or all processes, firmware, hardware or configuration should run on a small set of dedicated computing systems or on one computing system alone.

It is to be understood that the embodiments of the invention can be embodied strictly as a software program, or as firmware, or as hardware and/or circuitry alone, or as configurations in other computing systems, or as any form of combination of them, such as within a RNC or other node in the network.

The foregoing description gives only the preferred embodiments of the present invention and is not intended to limit the present invention in any way. Thus, any modification, substitution, improvement or like made within the spirit and principle of the present invention should be encompassed by the scope of the present invention.

Claims

What is claimed is:

1 . A system for communication in a cellular network, comprising: a grid establishment unit configured to divide the network into one or more grids, each grid having one or more nodes arranged therein that are equivalent to each other, a grid traffic estimation unit configured to estimate aggregate traffic for all nodes of a defined period in each gird; and an active node set determination unit configured to determine a set of active nodes of each grid for the defined period based on the estimated aggregate traffic, and power off the nodes in the grid that are not active.

2. The system according to claim 1 , wherein the grid establishment unit is configured to decide that two nodes i and j are equivalent if ri + d(i, j) < Rj, rj + d(i, j) < Ri, where d(i, j)is the distance between Node i and Node j, ri and rj are the normal communication ranges of nodes, and Ri and Rj are the maximum possible communication ranges of Node i and Node j, respectively.

3. The system according to claim 2, wherein the grid establishment unit is configured to form grids from a certain node in the network and towards a predetermined direction, until all the nodes in the network are exhausted.

4. The system according to claim 3, wherein a northwest rule is used to form the grids, where the grid establishment unit is configured to form grids from the top-left node in the network from top to down and from left to right, or a northeast rule is used to form the grids, where the grid establishment unit is configured to form grids from the top-right node in the network from top to down and from right to left.

5. The system according to claim 1 , wherein the period is one of a half hour, one hour and two hours of a day.

6. The system according to claim 1 , wherein the grid traffic estimation unit is configured to estimate the aggregate traffic of the period of a day according to statistics of traffic of the same period of the previous days.

7. The system according to claim 6, wherein the aggregate traffic of the period for a weekday is estimated based on the statistics of weekdays, while the aggregate traffic of the period for a weekend is estimated based on the statistics of weekends.

8. The system according to claim 1 , wherein nodes in a grid are selected to be active in decreasing order of their energy efficiency and/or capacities.

9. The system according to claim 1 , wherein the active node set determination unit is configured to determine the set of active nodes starting from the period having heaviest aggregate traffic to the periods having less aggregate traffic, and the set of active nodes for a period other than the one having heaviest aggregate traffic is selected from the set of active nodes for the preceding period.

10. The system according to claim 1 , wherein the active node set determination unit is configured to determine the set of active nodes for each period of a day according to distribution of the aggregate traffic of a day so that intersection of the sets of active nodes for neighboring periods is as large as possible.

11 . The system according to claim 1 , wherein the node is a Base Station (BS), and the system is located within a Radio Network Controller (RNC). 2. A method for communication in a cellular network, comprising: dividing the network into one or more grids, each grid having one or more nodes arranged therein that are equivalent to each other, estimating aggregate traffic for all nodes of a defined period in each gird; and determining a set of active nodes of each grid for the defined period based on the estimated aggregate traffic, and powering off the nodes in the grid that are not active.

13. The method according to claim 12, wherein two nodes i and j are equivalent if ri + d(i, j) < Rj, rj + d(i, j) < Ri, where d(i, j)is the distance between Node i and Node j, ri and rj are the normal communication ranges of nodes, and Ri and Rj are the maximum possible communication ranges of Node i and Node j, respectively.

14. The method according to claim 13, wherein dividing the network into one or more grids comprises: forming grids from a certain node in the network and towards a predetermined direction, until all the nodes in the network are exhausted.

15. The method according to claim 14, wherein a northwest rule is used to form the grids, where forming grids comprises forming grids from the top-left node in the network from top to down and from left to right, or a northeast rule is used to form the grids, where forming grids comprises forming grids from the top-right node in the network from top to down and from right to left.

16. The method according to claim 12, wherein the period is one of a half hour, one hour and two hours of a day.

17. The method according to claim 12, wherein estimating the aggregate traffic of the period of a day comprises: estimating the aggregate traffic of the period of a day according to statistics of traffic of the same period of the previous days.

18. The method according to claim 17, wherein estimating the aggregate traffic of the period of a day comprises: estimating the aggregate traffic of the period for a weekday based on the statistics of weekdays, and estimating the aggregate traffic of the period for a weekend based on the statistics of weekends.

19. The method according to claim 12, wherein determining a set of active nodes based on the estimated aggregate traffic comprises: selecting nodes in a grid to be active in decreasing order of their energy efficiency and/or capacities, until the aggregate capacity of the selected nodes is no smaller than the estimated aggregate traffic of the grid.

20. The method according to claim 12, wherein determining a set of active nodes of each grid for the defined period comprises: determining the set of active nodes starting from the period having heaviest aggregate traffic to the periods having less aggregate traffic, and selecting the set of active nodes for a period other than the one having heaviest aggregate traffic from the set of active nodes for the preceding period.

21. The method according to claim 12, wherein determining a set of active nodes of each grid for the defined period comprises determining the set of active nodes for each period of a day according to distribution of the aggregate traffic of a day so that intersection of the sets of active nodes for neighboring periods is as large as possible.

22. The method according to claim 12, wherein the node is a Base Station (BS), and the system is located within a Radio Network Controller (RNC), and before powering off the BSs that are not active, the method further comprises: handing over users served by these BSs to the active BSs though a network-controlled handoff technique.

23. The method according to claim 22, wherein before powering off the BSs that are not active, the method further comprises: applying a cell breathing technique to have the active BSs adjust their coverage.

24. The method according to claim 22, wherein the BSs are provided with multiple transmission units of different communication coverage ranges.

25. Computer-readable program having a program code for performing, when running on a computer, the method of any of claims 12-24.

26. A storage medium having a computer-readable program stored thereon, the computer-readable program performs the method of any of claims 12-24 when running on a computer.