Research Article Energy-Saving Management Mechanism Based...

Research ArticleEnergy-Saving Management Mechanism Based onHybrid Energy Supplies for LTE Heterogeneous Networks

Peng Yu Lei Feng Wenjing Li and Xuesong Qiu

State Key Laboratory of Networking and Switching Technology Beijing University of Posts and TelecommunicationsBeijing 100876 China

Correspondence should be addressed to Peng Yu yupengbupteducn

Received 5 February 2016 Accepted 29 June 2016

Academic Editor Gabriel-Miro Muntean

Copyright copy 2016 Peng Yu et alThis is an open access article distributed under the Creative Commons Attribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Aiming at the lack of integrated energy-saving (ES) methods based on hybrid energy supplies in LTE heterogeneous networksa novel ES management mechanism considering hybrid energy supplies and self-organized network (SON) is proposed Themechanism firstly constructs ES optimization model with hybrid energy supplies And then a SON framework is proposed toresolve the model under practical networks According to the framework we divide the ES problem into four stages whichare traffic variation prediction regional Base Station (BS) mode determination BS-user association and power supply Andfour corresponding low-complexity algorithms are proposed to resolve them Simulations are taken on under LTE underlayheterogeneous networks Comparedwith other algorithms results show that ourmechanism can save 474 energy consumption ofthe network while keeping coverage interference and service quality above acceptable levels which takes on great green-economysignificance

1 Introduction

Recently energy consumption of Information and Com-munications Technology (ICT) Industry has an explosiveincrease particularly in wireless communication networkwhose energy consumption occupies a leading position [1]Since BSs in wireless cellular network consume roughly 60ndash80 energy [2] saving energy for BS becomes a researchhotspot With the development of cellular network technolo-gies LTEheterogeneous networks are commercially deployedstep by step and how to manage BS energy is getting moreand more attention as well Currently ES approaches for BSsmainly include five categories which are improvement onhardware components sleep mode technologies optimiza-tion in radio transmission process network planning anddeployment and adoption of renewable energy resources [3]As sleep mode techniques and renewable energy suppliesare two research focuses which do not require equipmentupgrade and benefit from low capital costs so this papermainly concentrates on these two issues

For sleep mode techniques several methods mainlyconsider energy and throughput tradeoffs [4] or maximizingES gains [5] which are just for homogeneous networks withsimilar BS deployments

Many BS sleep technologies have been proposed for LTEheterogeneous networks as well And these technologies canbe classified into two categories one is aiming at decreasingpower at one time point and the other is going to saveenergy during entire time domains For the first categoryBS sleep control is resolved through characterizing energy-delay tradeoff [6] dynamic sectorization [7] for cells orstochastic analysis of optimal Base Station energy [8] and soforth However these solutions ignore traffic fluctuations andcannot evaluate temporal ES gains For the second categorymethods in [9 10] analyze temporal and regional ES efficiencyof sleep nodes from the point of dynamic traffic reallocationandpractical cell topology adjustment respectivelyHoweverthey do not consider coverage and interference control fromuser perspectives Additionally methods in [11 12] mainlydiscuss BS sleep gains from the point of traffic prediction

Hindawi Publishing CorporationMobile Information SystemsVolume 2016 Article ID 3121538 13 pageshttpdxdoiorg10115520163121538

2 Mobile Information Systems

and BS power mode analysis But coverage compensationmethods and interference control methods for sleep BS aretheoretical

Beyond sleep mode techniques renewable energy sup-plies are another approach for saving energy of power gridIn [13] dynamic renewable energy allocation methods withminimal online power are investigated But BS sleep approachis ignored And tradeoff and deployment challenges forenergy harvesting BS are discussed in [14] but power controlis just taking from the streaming perspective and no detailedmethods are given as well Further BS sleep method withrenewable energy supply is analyzed in [15] however itmainly concentrates onmathematicalmodel and solution butdoes not refer to any management scheme for such greencontrols

In fact to avoid frequent adjustments to the networkES management (ESM) is defined in self-optimization casesunder SON specifications and it is considered as a popularmethod to switch off several BSs during low traffic period[16] Based on our previous work on BS sleep methodswith coverage and interference consideration [17] and hybridenergy supplies [18] we propose an ES management mecha-nism based on hybrid energy supplies for LTE heterogeneousnetwork This mechanism uses self-organized architectureand combines BS sleep technology and renewable energysupply together The contributions are as follows

Firstly by analyzing resource allocation Quality of Ser-vice (QoS) and energy consumption in LTE heterogeneousnetwork we build an ES optimizationmodel based on hybridenergy supplies in time domain The model fully considersconstraints of power interference coverage and resource andhas strong universality

Then to make the ES model be resolved under practicalnetwork we propose a self-organized framework whichdefines the management architecture and ES procedures Inaddition in order to reduce the computational complexitythe model is divided into four stages which are trafficvariation monitoring regional BS mode determination BS-user association and power supply

To address these four problems we design S-ARIMAbased traffic prediction algorithm low-complexity BS coop-eration algorithm based on geographical topology dis-tributed user allocation algorithm and sustainable powersupply algorithm respectively Finally the simulations areconducted in LTE underlay heterogeneous network Resultsshow that the mechanism can obtain more regional ESgains compared to other algorithms while maintaining usercoverage interference and QoS above acceptable levels

The rest of the paper is organized as follows Section 2presents ES optimization model in LTE heterogeneous net-work which develops resource allocation and BS energy con-sumptionmodel and three key points for resolving themodelare analyzed then Section 3 gives the self-organized frame-work for ESM which includes the management architec-ture the management entities and management proceduresBased on the framework Section 4 proposes correspondingSeasonal Auto-Regressive Integrated Moving Average (S-ARIMA) based traffic prediction algorithm BS cooperationalgorithms based on geographical topology distributed user

allocation algorithm and sustainable power supply algorithmwith complexity analysis Section 5 numerically evaluates themechanism and algorithm Conclusions and future work aregiven in the last section

2 ES Optimization Model

For ES optimization model this paper primary focuses onresource allocation method and BS energy consumptionmodel and constructs an optimizationmodel based on hybridenergy supplies Moreover key points to resolve the problemwill be discussed as well

21 Resource Allocation Method Assuming that there are119873 BSs in underlay LTE heterogeneous network including1198731macro BSs (eNodeB) and 119873

2micro BSs (microcell)

each eNodeB contains three sectors (except for those on theboundary) B

119872and B

119898denote the set of eNodeB and the set

of microcells and B = B119872cup B119898= 119861119895 The network has

119870 kinds of services Supposing that time domain is [0 119879]for 119861119895and user 119906

119894 M119895(119905) is the set of active users of 119861

119895at

time 119905 Assume that X(119905) = 119909119894119895(119905) is the BS association

matrix where 119909119894119895(119905) = 1 if user 119906

119894is served by 119861

119895 otherwise

119909119894119895(119905) = 0 Then Signal to Interference plus Noise Ratio

(SINR) experienced by user 119906119894from 119861

119895at time 119905 is given by

120574119894119895(119905) =

119901119894119895(119905) sdot 119892

119894119895(119905)

N0+I119894(119905) + 119868

119894119895(119905) 119909119894119895(119905) = 1

0 119909119894119895(119905) = 0

(1)

where 119901119894119895(119905) denotes the power for each resource block (RB)

and P(119905) = 119901119894119895(119905) is defined as its matrix 119892

119894119895(119905) denotes

the channel gain between 119861119895and user 119906

119894 N0denotes the

thermal noiseI119894(119905) and 119868

119894119895(119905) denote interference outside 119861

119895

and inside 119861119895for user 119906

119894 respectively

Assuming that AMC (Adaptive Modulation and Coding)is used then spectral efficiency 120593

119894119895(119905) (bpsHz) when user 119906

119894

is served by 119861119895is given by [7]

120593119894119895(119905) =

0 120574119894119895(119905) lt 120574min

120585 log2(1 + 120574

119894119895(119905)) 120574min le 120574119894119895 (119905) lt 120574max

120593max 120574119894119895(119905) ge 120574max

(2)

where 120574min is the minimum SINR and 120585 isin [0 1] is theattenuation factor 120574max and 120593max are the maximum of SINRand spectral efficiency respectively Further we canderive thenumber of RBs for user 119906

119894required from 119861

119895as

120573119894119895(119905) = lceil

119903119894119895(119905)

119882RB120593119894119895 (119905)rceil (3)

where 119903119894119895(119905) is the required service rate from serving119861

119895to user

119906119894and 119882RB is the bandwidth for each RB Then load factor

119871119895(119905) of BS can be defined as

119871119895(119905) =

1

120573119879

119895

|M119895(119905)|

sum

119894=1

120573119894119895(119905) (4)

Mobile Information Systems 3

where 120573119879119895represents the total number of RBs of BS 119895 Load

factor can be used as one important indicator to compute BSdynamic power

22 BS Energy Consumption Model For 119861119895 assume that its

transmit power is 119875119879119895(119905) at time 119905 and maximum operational

power is 119875119872119895 And ratio of BS static power to 119875119872

119895is denoted

as 120575119895 Then the power of BS 119861

119895at time 119905 can be expressed as

119875119895(119905) = (1 minus 120575

119895) 119871119895(119905) 119875119872

119895+ 120601119895(119905) 120575119895119875119872

119895

120601119895(119905) =

1 119871119895(119905) gt 0

120576

120575119895

119871119895(119905) = 0

(5)

where 120576 denotes energy proportion of sleep BSs to maintainbasic management functions relative to maximal operationalpowerWith (5) we can evaluate BS power in differentmodes

For hybrid energy supplies we assume that each outdooreNodeB has a panel powered by renewable energy such aswind energy and solar energy Still primary energy of panelof 119861119895is 1198640119895(KWH) and power generation rate is V

119895(119905) To

ensure system stability renewable energy is used accordingto the power level of eNodeB Therefore at time 119905 energygeneration rate 119890

119895(119905) and stored energy 119864

119895(119905) of each panel

can be expressed as

119890119895(119905) = 119891 (V

119895(119905) 119875119895(119905)) (6)

119864119895(119905) = 119864

0

119895+ int

119905

0

119890119895(119905) 119889119905 (7)

In (7) function 119891(sdot) is determined by the power supplypolicy Assuming that all derived energy of eNodeBs canbe stored into the battery then required power 119886

119895(119905) from

power grid and corresponding consumed energy 119860119895(119905) can

be written as

119886119895(119905) = ℎ (V

119895(119905) 119875119895(119905)) (8)

119860119895(119905) = int

119905

0

119886119895(119905) 119889119905 (9)

In (8) function ℎ(sdot) is still determined by the powersupply policy According to (6)ndash(9) we can obtain accuratevalue of energy consumption in power grid and renewableenergy system

After determining resource allocation method and BSenergy consumption model we need to select a reasonableoptimization model and solve it with proper algorithms asshown in the following sections

23 ES Optimization Model In this paper our target isto minimize energy consumption from power grid andminimize the stored energy in each battery while ensuringcoverage interference and QoS in whole region Thereforefor all users 119906

119894isin M119895(119905) and BSs 119861

119895isin B the optimization

model can be expressed as (10)

As shown in (10) the optimization objects are associationmatrix X(119905) and allocated power matrix P(119905) at time 119905 Thefirst is service quality constraint where 119875119861

119895119896(119905) is the service

blocking probability for service 119896 at BS 119895 and 119875119861119879

is thethreshold of service blocking probability for each serviceThe second constraint makes sure that one user can only beserved by no more than one BS simultaneously The thirdconstraint is used to guarantee that RB number of each BScan satisfy user demandThe next constraint is the restrictionfor transmit power with control factor 120572 The fifth constraintis related to the signal strength where 120590

119894119895(119905) = 119901

119894119895(119905) sdot 119892

119894119895(119905)

denotes the signal strength received by user 119906119894from119861

119895at time

119905Here Pr(119909)means accumulative probability for condition 119909It ensures that the accumulative probability of received signalstrength for active users (above threshold 120594min) is higher than119875120590The sixth constraint is involvedwith regional interference

That is the accumulative probability of SINR received byusers from serving BS (above threshold 120574min) is limited by thepredefined target 119875

120574 Consider

P minX(119905)P(119905)

sum

119895isinB119872119864119895(119879) + sum

119895isinB119872119860119895(119879)

st 119875119861119895119896(119905) le 119875

119861

119879 forall119895 119896 119905

119873

sum

119895=1

119909119894119895(119905) = 1 forall119894 119905

|M119895(119905)|

sum

119894=1

120573119894119895(119905) le 120573

119879 forall119895 119905

|M119895(119905)|

sum

119894=1

120573119894119895(119905) 119901119894119895(119905) le 120572119875

119879

119895(119905) forall119895 119905

Pr (120590119894119895(119905) ge 120594min119909119894119895 (119905)) ge 119875120590 forall119905


(10)

In this optimization model if we want to minimize119860119895(119879) and 119864

119895(119879) at the same time we should consider ES

schemes from two points one is maximizing power supplyof renewable energy with proper power scheduling andthe second is minimal BS operational power with BS sleepstrategyDue to the complexity of the optimizationmodel andpractical limits of cellular networks we will analyze the keypoints for resolving the model

24 Key Points for Resolving the Model Since the modelcontains a lot of constraints so much information shouldbe collected from the network to figure out whether theseconditions are satisfied Still resolutions with X(119905) and P(119905)require network control to change wireless parameter forBSs and reconnection for users all these actions requirethe help of network management functions Currently mostnetwork management work is done manually However ESaction always requires frequent adjustments which may putheavy burdens on network So traditional manual control


Distributed SON Distributed SON

Centralized SON

Energy supply Energy supply

Energy storage Energy storage

Energy controller Energy controller

eNodeB

OAM

SON agent

SON agent SON agent

eNodeB SON agent

X2

MicrocellMicrocell

Power lineManagement signal

middot middot middotmiddot middot middot

middot middot middot

Figure 1 Illustration of compensation under irregular scenario

may be high cost To make ES action be executed moreefficiently ESM defined by 3GPP in SON use cases [16] andself-organized BS cooperationmethod [19] will be adopted inthis paper as a suitablemanagement policy and compensationmethod As optimization objects are discrete matrix and con-tinuous matrix and the constraints are nonlinear from [7]we can derive that this problem is a nonconvex optimizationproblem and is hard to be resolved Tomake the optimizationmodel be executed under practical networks we should takethe following four points with low-complexity methods intoconsiderations

241 How to Determine When ES Actions Can Be ExecutedES actions can be carried out through BS sleep and corre-sponding BS parameter adjustments (such as power tilt andneighbor relationship) ESmodel considers ES problems dur-ing thewhole time domain butwe could not execute it at eachtime point so as to avoid frequent parameter adjustmentsTherefore execution frequency should be proper Still as ESactions is always triggered by the traffic variations accuratetraffic prediction method is profitable as well

242 How to Determine the BS Mode When ES Actions AreExecuted During ES period several BSswill be slept On onehand we want to sleep as more BSs as possible On the otherhand regional coverage and traffic should be accommodatedby the active BSs so geographic BS deployment and regionaltraffic load should be considered as well Thus we should findan efficient BS mode determination method

243 How to Keep Usersrsquo QoS during ES Period As BS sleepwill change the network topology and no doubt affect userQoS such as perceived signal strength interference level andservice blocking probability we should give a method to

adjustment parameters from user QoS perspective thus toguarantee regional user QoS and network performance aboveacceptable level As parameters are the same to each user themethod should leverage the parameter effect among BSs andusers

244 How to Maximize the Utilization of Renewable EnergySupplies As renewable energy supplies come from the solarenergy or wind energy which vary drastically along withenvironment the energy generation rate will change alongwith time as well However renewable energy takes on thebest green benefits so a method should be given to maximizethe utilization of renewable energy supplies Moreover themethod should guarantee that power supply for each eNodeBis stable and sustainable

Aiming at resolving above key points we propose a self-organized framework to address them which will be shownin Section 3

3 Self-Organized Framework for ESM

According to the ESM definition and the scenarios of LTEheterogeneous networks with hybrid supplies we give theself-organized framework for ESM in Figure 1

As shown in Figure 1 to make ESM more practical weshould consider the management architecture the manage-ment entities and management procedures in SON frame-work

31 Management Architecture There are three kinds ofmanagement architecture in SON which are centralizedSON distributed SON and hybrid SON Considering bothdistributed massive BSs and the OAM regional function weuse hybrid SON here as shown in Figure 1


Monitoring thetraffic load

Regional BS modedetermination

BS-userassociation

Power supplypolicy

Distributed SON ateNodeB and SON agent

microcell

Centralized SON at OAM


at microcell

Distributed SON ateNodeB

Figure 2 Procedures of ESM

Thehybridmanagement architecture includes centralizedSON and distributed SON Here we assume that distributedSON communicates with SON agents deployed on eacheNodeB Distributed SON is responsible for guaranteeingusersrsquo QoS under each eNodeB Still distributed SON com-municate with each other through X2 interface

Moreover centralized SON is deployed on OperationAdministration and Maintenance (OAM) system to manageregional information such as network topology and regionaltraffic load Centralized SON communicates with distributedSON at each eNodeB and SON agent at each microcellRegional control algorithms will be executed by centralizedSON as well

32 Management Entities As shown in Figure 1 to keep flu-ent management among different network elements (includ-ing battery eNodeB and microcell) we should set thefollowing management entities

(1) SON agent at each eNodeB which monitors thebasic parameters power key performance indicatorsand traffic data and pushes the ES actions fromdistributed SON to eNodeB

(2) SON agent at microcell which monitors the basicparameters power key performance indicators andtraffic data and pushes the ES actions from central-ized SON to microcell

(3) Centralized SON at OAM system which collectsregional information from the eNodeB andmicrocellsand gives regional ES action suggestions

(4) Distributed SON at each eNodeB which collects andstores the information of the entire eNodeB andenergy controller and gives local ES action suggestionfor users under current eNodeB

(5) Energy supply which controls the energy supply foreach eNodeB under hybrid energy sources

(6) Energy storage which stores the energy coming fromthe renewable energy sources and power lines

(7) Energy controller which determines the energy sup-ply method according to the energy storage eNodeBpower requirements and the control informationfrom the distributed SON

With above entities we can obtain effective energy supplyand energy saving with hybrid management architecture

33 Management Procedures As ESM is a looped controlfor the whole network next we will give the procedures forhow to implement it With the management architecture andmanagement entities the procedures are shown in Figure 2

As shown in Figure 2 our procedures have four stagesunder the management of different SONs

(1) For distributed SON at each eNodeB and SON agentat each microcell they monitor the traffic load ofeach eNodeB and each microcell and predicate thetraffic variations for the next period From [7] wecan assume that traffic per hour is unchanged whichmakes it possible to take the traffic prediction onlyfrom each hour

(2) According to traffic prediction results and regionalnetwork information the centralized SON deter-mines the BS mode and traffic accommodation poli-cies The determination should make sure that nocoverage hole exists in the network as well Here weset 119904119895(119905) as state of 119861

119895at time 119905 corresponding to the

mode (ie 119904119895(119905) = 0 as sleep mode and 119904

119895(119905) = 1 as

active mode)(3) After BS mode determination next distributed SON

at eNodeB and SON agent at microcell should findappropriate way to adjust parameters at each BS thusto make each user associate with proper BS aboveacceptable QoS levels


(4) Once the parameter adjustments are determined wecan obtain the power required for each BS Thendistributed SON at each eNodeB determines thepower supply policies to maximize the utilization forrenewable energy

After the energy supply policies are determined allthe parameter adjustments and power supply strategies willbe executed by each BS under the control of SON agentAnd then the whole network will return back to the trafficmonitoring stages

Above procedures just denote what to do for differentSON entities We should give the proper algorithms fordifferent stages as well To resolve the key points in the ESMprocedures we proposed S-ARIMA based traffic predictionalgorithms BS cooperation algorithm based on geographictopology distribution user allocation algorithm and sustain-able power supply algorithm to resolve them

4 Corresponding Practical Algorithm

41 S-ARIMA Based Traffic Prediction Algorithm To judgewhen ES actions can be triggered or rolled back we shouldknow how traffic will be changed along with time Heretraffic is taken as the load factor in (4) There are manytraffic prediction methods which have been used in BS sleepmethods such as Holt-Winters in [11] and online stochasticgame theoretic algorithm in [20] But they are not suitable fortraffic with small value and the accuracy can be improved Inthis paper according to the periodic features of traffic we useS-ARIMA traffic prediction algorithm to estimate the futuretraffic

The S-ARIMA model is given by

120601119909(119862)Φ

119883(119862119910) (1 minus 119862)

119889(1 minus 119862

119910)119863

(119905)

= 120579119902(119862)Θ

119876(119862119910) 120588 (119905)

(11)

where

(119905) =

119871 (119905) minus 120583 119889 = 119863 = 0

119871 (119905) others(12)

with

120601119909(119862) = 1 minus 120601

1119862 minus 120601

21198622minus sdot sdot sdot minus 120601

119909119862119909

Φ119883(119862119910) = 1 minus Φ

1119862119910minus Φ21198622119910minus sdot sdot sdot minus Φ

119883119862119883119910

120579119902(119862) = 1 minus 120579

1119862 minus 12057921198622minus sdot sdot sdot minus 120579

119902119862119902

Θ119876(119862119910) = 1 minus Θ

1119862119910minus Θ21198622119910minus sdot sdot sdot minus Θ

119876119862119876119910

(13)

In (11)sim(13) 120601119909(119862) and 120579

119902(119862) are the conventional autore-

gression operator and moving average operator respectivelyCorrespondingly Φ

119883(119862119910) and Θ

119876(119862119910) are seasonal autore-

gression operator and moving average operator 120588(119905) is thewhite noise with zero average and 120583 is a constant value 119862is the backward shift operator as 119862119871(119905) = 119871(119905 minus 1) 119871(119905) isthe load factor of BS and is taken as the time sequences

Moreover 119889 and 119863 are differential order and seasonaldifferential order respectively Then we call the model in (11)S-ARIMA(119909 119889 119902) times (119883119863119876)

119910model with season 119910

To obtain the proper S-ARIMAmodel for time sequenceswe should execute the following steps

Step 1 Compute the differencesnabla and seasonal differencesnabla119910

to obtain stationary series for the given time sequences

Step 2 Compute the Autocorrelation Function (ACF) andPartial Autocorrelation Function (PACF) for the stationarysequences and then match them to known values in S-ARIMA model If more than one combination of (119909 119889 119902) times(119883119863119876) is proper we then adopt the one with minimalAkaikersquos Information Criterion (AIC) as the tentative model

Step 3 Compute the initial estimation for model parametersin S-ARIMA(119909 119889 119902) times (119883119863119876)

119910withMaximum Likelihood

Estimation (MLE) or moment estimation

Step 4 After fitting check whether the residual sequencescan be considered as white noise with ACF and PACFcomputation If the checking is not passed improvementfor the parameters will be given and fitting and checkingprocedures will be executed until the checking is passed

As traffic variations in each BS take on obvious seasonablefeature S-ARIMA is an effective prediction algorithm forcellular traffic

In fact as time series prediction models require lotsof computations and iterations their computational com-plexities are determined by data volume the number ofparameters the estimation method and time cycle So itis hard to give an accurate mathematical expression fortime complexity However many tools such as RStudio haveintegrated S-ARIMA model into them and it is easy to usethis tool to predicate the time sequences

42 BS Cooperation Algorithm Based on Geographic Topol-ogy (BCAGT) For eNodeB since static power of each BSoccupiesmore than its 50 energy consumption as describedin [21] so the target of this part is to maximize number ofsleep BSs with global information at the centralized SON Inaddition three constraints should be taken into account

(i) After sleeping BSs and reallocating traffic load noactive node is overloaded

(ii) To reduce effect of frequent handovers caused by BSsleeping number of sleep times per BS during entiretime domain cannot exceed a threshold (eg 1 time)

(iii) To ensure satisfactory coverage each sleep BS has atleast one active neighbor BS

With above considerations BCAGTwhichmainly use thenetwork topology information can be obtained beforehand

For slept BSs one two or three neighbor BSs can cooper-ate to compensate coverage and capacity [22] as illustrated inFigure 3 Micro BS 119861

12is fully compensated by Macro BS 119861

11

which is called EP (Entire Pair) of 11986112 Additionally macro


Input B L(119905) 119904119895(119905) Output L(119905) 119904

119895(119905)

(1)B = BT = (2) whileB =

(3) 119861119895lowast lArr argmin

119861119895isinB119871119895(119905) | 119904

119895(119905) = 1 ampamp 119871

119895(119905) lt 1

(4) if 119861119895lowast is a micro - BS


119861119896isinH119895lowast (119905)119871119896(119905) | 119904

119896(119905) = 1 ampamp 119871

119896(119905) lt 1

(6) if 119861119896lowast exists ampamp 119871

119895lowast (119905) + 119871

119896lowast (119905) le 1

(7) 119904119895lowast (119905) = 0 119871

119896lowast (119905) lArr 119871

119895lowast (119905) + 119871

119896lowast (119905)

(8) end if(9) end if(10) if 119861

119895lowast is a Macro - BS

(11) TlArr OP119895lowast cup TP

119895lowast

(12) CPlowast lArr argmaxCPisinTprod119861119896isinCP(1 minus 119871119896(119905)) | forall119861119896 isin CP 119871119896(119905) lt 1 ampamp 119904119896(119905) = 1(13) if CPlowast exist ampamp for forall119861

119896isin CPlowast 119871

119896(119905) + 119908

119896119871119895lowast (119905) le 1

(14) 119871119896(119905) lArr 119871

119896(119905) + 119908

119896119871119895lowast (119905) 119904

119895lowast (119905) = 0

(15) end if(16) end if(17) BlArrB 119861

119895lowast

(18) end while

Algorithm 1 Description of BS cooperation algorithm

B1

B2

B3

B4

B5

B6

B7

B8

B9

B10

B11

B12


BS 1198619can be compensated by macro BS opposite pair (OP)

(1198618 11986110) and macro BS 119861

2can be compensated by macro BS

trigonal pair (TP) (1198611 1198614 and 119861

5) The definitions of OP and

TP can be seen in our previous work in [23 24]Based on definitions ofOP andTP the time domain [0 119879]

can be divided into four phases due to regional traffic states[17] In peak andmidnight phase the states of BSs remain thesame And in traffic decreasing phase this algorithm shouldbe executed at the beginning of each hour The process isshown as follows in Algorithm 1 This algorithm shows theprocess of sleep BS selection with load decline Similarlybased on symmetry of load variation in time domain thereverse process of BCAGT is used to recover sleep BSs duringtraffic increasing phase The four phases are determinedaccording to the fitting for historic traffic load Moreover thetraffic load used in this algorithm is the prediction traffic loadas well

Here L(119905) is the traffic prediction vector for regional BSsAs shown in Algorithm 1 firstly we find the active 119861

119895lowast with

the lowest load If 119861119895lowast is a micro BS select the active BS 119861

119896lowast

with the lowest load from its neighbor macro BS set H119895lowast(119905)

which can completely cover 119861119895lowast If 119861

119896lowast exists and is able to

absorb the load of 119861119895lowast then we can transfer the load to 119861

119896lowast

and sleep 119861119895lowast If 119861

119895lowast is a macro BS its OP set OP

119895lowast and TP

set TP119895lowast should be selected to form the set of compensation

elements denoted as T Then select compensation elementCPlowast which satisfies the conditions that forall119861

119896isin CPlowast is active

and not overloaded and the product of surplus load of all BSsin CPlowast is maximum If CPlowast exists and is able to absorb theload of 119861

119895lowast its load will be allocated by a ratio of119908

119896to BSs in

CPlowast and thenwe can sleep it According to [20]119908119896is defined

as

119908119896=

ℓ2

119896119895lowast

sum119894isinCPlowast ℓ

2

119894119895lowast

(14)

Here ℓ119894119895is the distance from BS 119894 to BS 119895

After selecting 119861119895lowast all BSs in this region should be

traversed until all BSs are analyzed We can easily findthat complexity of Algorithm 1 is 119874(|B

119898| sdot maxH

119895lowast(119905) +

|B119872| sdot max|OP

119895lowast | |TP

119895lowast |) Based on analysis from [17]

we know that max|OP119895lowast | |TP

119895lowast | le 20 Still neighbor

macro BS for each micro BS is known from the networktopology (often is no more than 3) so the complexity is lessthan 119874(3|B

119898| + 20|B

119872|) which means complexity is only

determined by regional BS numberSince this algorithm analyzes the compensatory method

only from view of BS load and state we need to considerregional and BS power constraint coverage constraint inter-ference constraint QoS constraint and so forth Aiming atsolving optimization problem from the perspective of usersthe paper designs distribution user allocation algorithm toachieve the optimal allocation for users next


Input B U(119905) X(119905) P(119905) Output X(119905) P(119905)(1)U(119905) = U(119905)(2) whileU(119905) = (3) for forall119894lowast isinU(119905) 119861

119895lowast lArr arg

119861119895isinBmax120590119894lowast119895(119905) | 119871

119895(119905) lt 1

(4) while 120590119894lowast119895lowast (119905) lt 120594 or 120590

119894lowast119895lowast (119905)(N

0+ sum119873

119896=1119896 =119895lowast 119875119879

119896(119905)119892119894lowast119896(119905)) lt 120574min

(5) 119901119894lowast119895lowast (119905) lArr 119901

119894lowast119895lowast (119905) + Δ119901 119909

119894lowast119895lowast (119905) = 1

(6) if exist119896 119875119861119895lowast119896(119905) gt 119875

119861

119879 break end if

(7) if sum|M119895lowast (119905)|119894=1

120573119894119895lowast (119905)119901119894119895lowast (119905) gt 120572119875

119879

119895lowast break end if

(8) if 119871119895lowast (119905) + 120573

119894lowast119895lowast (119905) gt 1 break end if

(9) end while(10) U(119905) lArrU(119905) 119894lowast

(11) end while

Algorithm 2 Description of distribution user allocation algorithm

Input 119875119895(119905) 119860

119895(119905) 119864119895(119905) Output 119860

119895(119905) 119864119895(119905)

(1) if 119864119895(119905) ge int

119905+1

119905119875119895(119905)119889119905

119864119895(119905) = 119864

119895(119905) minus int

119905+1

119905119875119895(119905)119889119905 + int

119905+1

119905V119895(119905)119889119905 and 119886

119895(119905) = 0

(2) else(3) 119864119895(119905) = 119864

119895(119905) + int

119905+1

119905V119895(119905)119889119905 and 119886

119895(119905) = 119875

119895(119905)

(4) end if

Algorithm 3 Description of sustainable power supply algorithm

43 Distribution User Allocation Algorithm (DUAA) AboveBS cooperation algorithm mainly concentrates on sleepnodes method and load reallocation Further user-BS asso-ciation needs to consider specific user allocation In this partthe regional power is minimized subject to the constraints in(10) The microscopic problem is a complex combinationaloptimization problem aswellThus this paper employs a low-complexity DUAA to solve it

We use U(119905) to designate the set of users at time 119905 whereU(119905) = cupM

119895(119905) For arbitrary user 119894lowast in U(119905) select the

corresponding BS 119895lowast with the strongest received signal Ifeither 120574

119894lowast119895lowast(119905) or 120590

119894lowast119895lowast(119905) does not meet the requirements it

can be considered that the serving BS of user 119894lowast is sleptAnd we can adjust power per RB 119901

119894lowast119895lowast(119905) of 119895lowast to satisfy

constraintsAccording to [7] in LTE system 119868

119894119895(119905) is generally set as

0 Based on RB conflict principleI119894(119905) can be written as

I119894(119905) =

119873

sum

119895=1119895 =119894

119871119894(119905) 119871119895(119905) 119875119879

119895(119905) 119892119894119895(119905) (15)

Then we have

120574119894119895(119905) =

120590119894119895(119905)

N0+ sum119873

119896=1119896 =119895119871119894(119905) 119871119896(119905) 119875119879

119896(119905) 119892119894119896(119905)

ge

120590119894119895(119905)

N0+ sum119873

119896=1119896 =119895119875119879

119896(119905) 119892119894119896(119905)

(16)

Obviously if the latter term in (16) is not less than 120574minit can be derived that 120574

119894119895(119905) ge 120574min Assuming that the

step to adjust power is Δ119901 this algorithm is described inAlgorithm 2 Since adjustable parameter is only power per RBallocated to each user which is irrelevant to other users andBS load DUAA is a distributed algorithmwithout centralizedcontrol

Given that the scope of 119901119894119895

is [119901min 119901max] and thecomplexity to compute 119875119861

119895lowast119896(119905) is Λ then the complexity of

DUAA is119874((119901maxminus119901min)Δ119901sdotΛsdot119870sdot|B|2 sdot |U(119905)| sdotmaxM119895(119905))

As 119870 is always a constant and the iteration upper limit isdefinite when range and step of 119901

119894119895are known computation

complexity is just 119874(Λ sdot |B|2 sdot |U(119905)|2) which is an acceptablequadratic polynomial

44 Sustainable Power Supply Algorithm With above threealgorithms we can obtain the BS modes the traffic realloca-tion methods and user-BS association strategies Howeverthey are all focusing on the power of BS requirement withoutconsidering hybrid energy supplies Here sustainable powersupply algorithm is proposed to maximize the green energyutilization For each eNodeB we will execute the algorithmas Algorithm 3 which determines function 119891(sdot) and ℎ(sdot) Tomake energy supply more stable the energy supply methodis consistent with the approach in [13 18] Still assume timeinterval during 119905 and 119905 + 1 is one hour here Algorithm 3 willbe executed at each time 119905 as well

That is only when the storage energy of renewable energyis higher than the eNodeB power required during the next


0 200 400 600 800 1000 1200 1400 1600 1800 20000

200

400

600

800

1000

1200

1400

1600

1800

2000

(m)

(m)

Figure 4 Illustration of simulation scenario

time interval will the renewable energy be used Otherwisethe energy will be stored for the next time intervals

In this algorithm as 119860119895(119905) and 119864

119895(119905) just need to be

computed at each time point with linear judgment for eacheNodeB so its complexity is only 119874(|B

119872|) which is linear

with eNodeB number

5 Simulation and Analysis

51 Simulation Scenario The simulation is performed in LTEunderlay heterogeneous network scenario as illustrated inFigure 4 This part of network covers a 2000m times 2000msquare area which includes 16 eNodeBs and 34 microcellsIn this figure blue asterisks denote the locations of eNodeBblue circles denote the locations of microcell and red bulletsdenote the users at a time point Still we assume that users areuniformly distributed in the network and we only consider512 kbps CRB services in the network The path loss employsCOST-231 HataModelThe BS carrier frequency penetrationloss antenna gain and thermal noise are 2GHz 10 dB 10 dBand minus174 dBmHz respectively

Moreover for resource allocation model the number ofRBs for eNodeB and microcell is 100 and 20 The attenuationfactor 120585 is 095 And 120574min and 120574max are minus13 dB and 20 dBrespectively 120593max is 48Mbps Bandwidth of each RB is180KHz

In BS energy consumption model and QoS evaluationmodel the maximal transmit power of eNodeB and microBS is 20W and 10W while the maximum operational poweris 500W and 15W respectively The ratio of static powerto maximum operational power of eNodeB and microcellis supposed to be 08 and 033 And 120576 and power amplifierefficiency are fixed as 005 and 02 for all BSs Primary energyof all eNodeB panels is set to be 0 Using S-ARIMA basedalgorithm in Section 4 for normalized traffic which comesfrom a city in China we predict traffic variations for Fridayas shown in Figure 5 We have found that S-ARIMA(1 1 1) times(0 1 1)

24is the most accurate model with highest correlation

coefficient 0996

002040608

1

1 9 17 25 33 41 49 57 65 73 81 89 97 105 113

Nor

mal

ized

traffi

c

Time (h)

Original trafficPredicated traffic

Figure 5 Traffic prediction for Friday with data from Monday toThursday

0005

01015

02025

03035

Serv

ice a

rriv

al ra

te (

s)

Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

001020304050607

Pow

er g

ener

atio

n ra

te (k

W)

Figure 6 Service arrival rate and power generation rate

Table 1 Simulation parameters

Parameter Value Unit119875119861

1198791

120572 09 mdash120594 minus105 dBm119875120590

97 119875120574

98 119901min 01 Watt119901max 1 WattΔ119901 005 Watt

According to the prediction results here we use a timeperiod of 29 hours predicated for Friday as the simulationtime Service arrival rate in the region and energy generationrate of solar panels are depicted in Figure 6 where theaverage service time is 5 minutes and the number of availableresource is the maximum resource number Here arrivalrate is consistent with the predicted results and the powergeneration rate is the same as [18] At the beginning ofeach hour user arrives at each BS with the same Poissonarrival process as shown in Figure 6 S-ARIMA algorithm isimplementedwith RStudio And the rest of the algorithms aresimulated under MATLAB The values of other parametersused in simulations are outlined in Table 1

According to the models and parameters above-mentioned simulation results are given as follows

52 Result Analysis The simulation is performed in LTEunder heterogeneous network and considers time-variant


Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

020406080

100120140160180200220240260280300

Accu

mul

ated

ener

gy (k

WH

)

Without ESES under power gridES under hybrid power supplies

Figure 7 Comparison of ES performance under different mecha-nisms

characters which is less studied yet Therefore this paperemphasizes the analysis of ES BSs numbers energy efficiencyand QoS coverage and interference parameters

It is true that executing ES algorithms and schemes alwaysputs additional computation and management burden of themanagement center and energy consumption may increaseas well However in our mechanism these algorithms andschemes are mainly executed in centralized SON at OAMsystem and distributed SON and SON agents on the BSsFor distributed SON and SON agents on the BS mainlyresponsible for ES action costs the energy costs have beentaken into consideration in (5) with ratio 120576 denoting energyproportion of sleep BSs to maintaining basic managementfunctions With these for active BSs with compensationactions we can assume that the control energy costs canbe accommodated by power increase For centralized SONlocated at OAM system the number of these nodes is fairlylower than number of BSs so their energy consumption ismuch lesser than BSs Besides as we adopt algorithms andschemes with low computation complexity their additionalenergy consumption is inappreciable compared to energy-saving gains for BSs Considering that these additionalenergies are minor and hard to be quantified we just ignorethem here

In the whole time domain themaximumnumber of sleepmacro BSs is 7 and sleep time intervals are 2sim9 and 24sim30 In addition all micro BSs can be slept under constraintsbetween 11 and 34 ones for different hours In time domain119879energy consumption of normal state is labeled as 119865(119879) andenergy consumption of using ES method is labeled as 1198651015840(119879)then ES gain in time domain 119866

119864(119879) can be expressed as

119866119864(119879) =

119865 (119879) minus 1198651015840(119879)

119865 (119879)times 100 (17)

Figure 7 shows the variation of regional accumulatedenergy consumption for three different methods which are

05

101520253035404550

OP

in [2

3]

TP in

[24]

Gre

enBS

N in

[5]

ES u

nder

pow

ergr

id

ES u

nder

hyb

ridpo

wer

supp

ly

ES-gain ()

Figure 8 ES gain comparison for different methods

method without ES mechanism method with ES underpower grid and method with ES under hybrid power sup-plies Here ES under power grid means only S-ARIMABCAGT andDUAA are adopted and ES under hybrid powersupplies mean that all the algorithms in this paper are usedCompared with conventional method energy consumptionof power grid can be saved more with renewable energyDuring time interval 10sim15 renewable energy system cansatisfy energy demands individually

As ESmethods in [23 24] just take ES actions once duringthe period there is no doubt that ES method proposed inthis paper will take on higher energy efficiency than themAs shown in Figure 8 compared with OP method in [23]TP method in [24] and classical GreenBSN in [5] (here wejust assume BS radius for eNodeB uses the value in [17]) wecan find that ES gains of our proposed ES mechanisms are3265 and 4740 respectively which are almost twice forOP (1732) and TP (1651) However GreenBSN takes onlittle higher ES efficiency (3386) than our ES under powergrid as it is a nearly optimal method But it is theoretical tosome extent as interference control is not preferred

Since ES mechanism has impact on system performancein the following we analyze coverage interference andQoS indicators respectively There is no doubt that ourmechanism is worse than methods in [23 24] as more BSsare slept So here we mainly explore the performance of ourmechanism after execution

To evaluate performance effect of our algorithm wechoose the time point with most sleep BSs (which is the 29thhour) and analyze the RSRP and SINR distributions for theactive eNodeB with highest traffic load at this time point Fig-ure 9 shows cumulative probability distribution of coverageindicator RSRP for the selected BS As DUAA just considerspower control for users under acceptable levels coverage andinterference effects for other active users should be evaluatedas well Here ES (users) means performance for user setwhose power has been adjusted through DUAA and ES(regional) means performance for all the active users in thisnetwork It can be seen that ESmechanism degrades coverage


minus120 minus110 minus100 minus90 minus80 minus70 minus60 minus50 minus40 minus30 minus200

01

02

03

04

05

06

07

08

09

1

RSRP (dBm)

Accu

mul

ativ

e pro

babi

lity

Without ESWith ES (regional)

With ES (users)

Figure 9 Cumulative probability distribution of RSRP

0

01

02

03

04

05

06

07

08

09

1

Accu

mul

ativ

e pro

babi

lity

minus20 minus15 minus10 minus5 0 5 10 15 20 25 30 35 40 45 50SINR (dB)


With ES (users)

Figure 10 Cumulative probability distribution of SINR

performance to some extent In the analysis we consider theeffect on active users as well as effect on overall coverageperformance of selected BS Because our mechanism mainlyemphasizes power control for active users under sleep BSsso RSRP cumulative probability distribution of active usersis generally better than all the users in the network Furthercumulative probabilities for active users and regional RSRP(more than minus105 dBm) are both 100 which proves thatcoverage performance conforms to constraints

Similarly from the perspective of interference cumu-lative probability distribution of interference indicator forselected BS is illustrated in Figure 10 We can see that ESmechanism can negatively affect regional interference as wellMoreover SINR cumulative probability distribution of activeusers also performs better than SINR distribution of overallcoverageMeanwhile cumulative probabilities of SINR (more

100 200 300 400 500 600 700 800 900 100025

30

35

40

45

50

55

60

65

70

75

Static power of BS (W)

ES effi

cien

cy (

)

ES under power gridES under hybrid power supplies

Figure 11 Regional ES gain with static power variation per BS

thanminus105 dBm) for active users under sleep BSs and for all theusers in the network are 100 and 981 respectively whichmeans interference meets constraints as well

As for QoS with computationmethod in [25] simulationresults indicate that maximum service blocking probability isless than the target 1 which indicates that it satisfies QoSconstraint

In order to verify scalability ES efficiency for BSs withdifferent static powers is further studied under simulationscenario As shown in Figure 11 on the premise that sleepnode method is determined ES efficiency decreases as BSstatic power increases which shows that BS static power isbottleneck of ES efficiency In other words reducing BS staticpower can enhance energy efficiency significantly WhenBS static power is lower than 500 Watt regional energyconsumption is less Thus it can be powered by renewableenergy At this point the ES mechanism mentioned in thispaper performs much better than conventional sleep nodemethods When BS static power is equal to 100 Watt bothmechanisms can achieve optimal energy gains which are7166 and 4688 respectively Conversely when BS staticpower is more than or equal to 500 Watt regional energyconsumption is more than available renewable energy whichmeans only power grid can be used Thus ES effects of twomethods tend to be the same and reach the peak efficiency3093 at 500 Watt It indicates that renewable energy hascertain limitations because of its low generation rate

Consequently the mechanism can reduce energy con-sumption of LTE heterogeneous network while maintainingsatisfactory coverage interference and QoS In addition itcan implement efficient ES for BSs with different powerthereby having strong adaptability

6 Conclusion

For LTE heterogeneous network this paper proposes anESM mechanism based on hybrid energy supplies With


simulations under irregular topology in LTE underlay het-erogeneous network this paper verifies that this mechanismcan save 474 energy while ensuring the acceptable regionalcoverage interference and QoS and has strong adaptabilityIn our further study we can take into account new charactersof LTELTE-A network Moreover new technologies suchas CoMP Relay and D2D can be used to achieve regionalcompensation thereby implementing ES reducing interfer-ence and enhancing resource utilization Additionally someinnovative indicators such as power per bit and power persquare can be set as optimization objectives to constructES models Still energy pool technologies which can sharethe renewable energy among different BS will be studiedWireless powering and energy-harvesting technologies for BSpower supply will be considered as well

Competing Interests

The authors declare that they have no competing interests

Acknowledgments

This research is supported by the National High Tech-nology Research and Development Program of China(2015AA01A705) and Natural Science Foundation of China(61271187)

References

[1] K Davaslioglu and E Ayanoglu ldquoQuantifying potential energyefficiency gain in green cellular wireless networksrdquo IEEE Com-munications Surveys amp Tutorials vol 16 no 4 pp 2065ndash20912014

[2] E Oh K Son and B Krishnamachari ldquoDynamic base stationswitching-onoff strategies for green cellular networksrdquo IEEETransactions on Wireless Communications vol 12 no 5 pp2126ndash2136 2013

[3] J Wu Y Zhang M Zukerman and E K-N Yung ldquoEnergy-efficient base-stations sleep-mode techniques in green cellularnetworks a surveyrdquo IEEE Communications Surveys and Tutori-als vol 17 no 2 pp 803ndash826 2015

[4] A Kumar and C Rosenberg ldquoEnergy and throughput trade-offs in cellular networks using base station switchingrdquo IEEETransactions on Mobile Computing vol 15 no 2 pp 364ndash3762016

[5] C Peng S-B Lee S Lu and H Luo ldquoGreenBSN enablingenergy-proportional cellular base station networksrdquo IEEETransactions onMobile Computing vol 13 no 11 pp 2537ndash25512014

[6] Z Niu X Guo S Zhou and P R Kumar ldquoCharacterizingenergy-delay tradeoff in hyper-cellular networks with basestation sleeping controlrdquo IEEE Journal on Selected Areas inCommunications vol 33 no 4 pp 641ndash650 2015

[7] M F Hossain K S Munasinghe and A Jamalipour ldquoEnergy-aware dynamic sectorization of base stations in multi-cellofdma networksrdquo IEEEWireless Communications Letters vol 2no 6 pp 587ndash590 2013

[8] J Peng PHong andKXue ldquoStochastic analysis of optimal basestation energy saving in cellular networks with sleep moderdquoIEEE Communications Letters vol 18 no 4 pp 612ndash615 2014

[9] N Deng M Zhao J Zhu and W Zhou ldquoTraffic-aware relaysleep control for joint macro-relay network energy efficiencyrdquoJournal of Communications and Networks vol 17 no 1 pp 47ndash57 2015

[10] L Suarez L Nuaymi and J-M Bonnin ldquoEnergy-efficient BSswitching-off and cell topology management for macrofemtoenvironmentsrdquo Computer Networks vol 78 pp 182ndash201 2015

[11] S Morosi P Piunti and E Del Re ldquoSleep mode managementin cellular networks a traffic based technique enabling energysavingrdquo Transactions on Emerging Telecommunications Tech-nologies vol 24 no 3 pp 331ndash341 2013

[12] D Paolo M Marco B Nicola and B Nicola ldquoA model toanalyze the energy savings of base station sleep mode in LTEHetNetsrdquo in Proceedings of the IEEE International Conference onand IEEE Cyber Physical and Social Computing and Internet ofThings Green Computing and Communications (GreenCom rsquo13)pp 1375ndash1380 Beijing China August 2013

[13] T Han and N Ansari ldquoOn optimizing green energy utilizationfor cellular networks with hybrid energy suppliesrdquo IEEE Trans-actions on Wireless Communications vol 12 no 8 pp 3872ndash3882 2013

[14] D Zordan M Miozzo P Dini and M Rossi ldquoWhen telecom-munications networks meet energy grids cellular networkswith energy harvesting and trading capabilitiesrdquo IEEE Commu-nications Magazine vol 53 no 6 pp 117ndash123 2015

[15] J Gong J S Thompson S Zhou and Z Niu ldquoBase stationsleeping and resource allocation in renewable energy poweredcellular networksrdquo IEEE Transactions on Communications vol62 no 11 pp 3801ndash3813 2014

[16] 3GPP ldquoEnergy Saving Management (ESM) concepts andrequirementsrdquo 3GPP TS 32551 Version 1130 2012

[17] P Yu L Feng Z Li W Li and X Qiu ldquoLow-complexity energyefficient base station cooperationmechanism in LTE networksrdquoKSII Transactions on Internet and Information Systems vol 9no 10 pp 3921ndash3944 2015

[18] P Yu J-P Cao S-X Zhang and W-J Li ldquoEnergy-savingmanagement mechanism based on hybrid energy supplies forwireless cellular networksrdquo Journal of Beijing University of Postsand Telecommunications vol 38 no 1 pp 46ndash50 2015

[19] M F Hossain K S Munasinghe and A Jamalipour ldquoOnthe energy efficiency of self-organizing LTE cellular accessnetworksrdquo in Proceedings of the IEEE Global CommunicationsConference (GLOBECOM rsquo12) pp 5314ndash5319 IEEE AnaheimCalif USA December 2012

[20] N Saxena B J R Sahu and Y S Han ldquoTraffic-aware energyoptimization in green LTE cellular systemsrdquo IEEE Communica-tions Letters vol 18 no 1 pp 38ndash41 2014

[21] M Deruyck E Tanghe W Joseph and L Martens ldquoModellingand optimization of power consumption in wireless accessnetworksrdquo Computer Communications vol 34 no 17 pp 2036ndash2046 2011

[22] M F Hossain K S Munasinghe and A Jamalipour ldquoOnthe eNB-based energy-saving cooperation techniques for LTEaccess networksrdquo Wireless Communications and Mobile Com-puting vol 15 no 3 pp 401ndash420 2015

[23] P Yu W-J Li and X-S Qiu ldquoA regional autonomic energy-saving management mechanism for cellular networksrdquo Journalof Electronicsamp Information Technology vol 34 no 11 pp 2707ndash2714 2012

[24] P Yu W Li and X Qiu ldquoSelf-organizing energy-savingmanagement mechanism based on pilot power adjustment in


cellular networksrdquo International Journal of Distributed SensorNetworks vol 2012 Article ID 721957 13 pages 2012

[25] L Chiaraviglio D Ciullo M Meo andM A Marsan ldquoEnergy-efficientmanagement ofUMTS access networksrdquo inProceedingsof the 21st International Teletraffic Congress (ITC 21 rsquo09) pp 1ndash8Paris France September 2009

Submit your manuscripts athttpwwwhindawicom

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Distributed Sensor Networks


Advances in

FuzzySystems

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014


ReconfigurableComputing

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014


Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

Artificial Intelligence

HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014

Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications


Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia


Biomedical Imaging


ArtificialNeural Systems

Advances in


RoboticsJournal of



Computational Intelligence and Neuroscience

Industrial EngineeringJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in



and BS power mode analysis But coverage compensationmethods and interference control methods for sleep BS aretheoretical

Beyond sleep mode techniques renewable energy sup-plies are another approach for saving energy of power gridIn [13] dynamic renewable energy allocation methods withminimal online power are investigated But BS sleep approachis ignored And tradeoff and deployment challenges forenergy harvesting BS are discussed in [14] but power controlis just taking from the streaming perspective and no detailedmethods are given as well Further BS sleep method withrenewable energy supply is analyzed in [15] however itmainly concentrates onmathematicalmodel and solution butdoes not refer to any management scheme for such greencontrols

In fact to avoid frequent adjustments to the networkES management (ESM) is defined in self-optimization casesunder SON specifications and it is considered as a popularmethod to switch off several BSs during low traffic period[16] Based on our previous work on BS sleep methodswith coverage and interference consideration [17] and hybridenergy supplies [18] we propose an ES management mecha-nism based on hybrid energy supplies for LTE heterogeneousnetwork This mechanism uses self-organized architectureand combines BS sleep technology and renewable energysupply together The contributions are as follows

Firstly by analyzing resource allocation Quality of Ser-vice (QoS) and energy consumption in LTE heterogeneousnetwork we build an ES optimizationmodel based on hybridenergy supplies in time domain The model fully considersconstraints of power interference coverage and resource andhas strong universality

Then to make the ES model be resolved under practicalnetwork we propose a self-organized framework whichdefines the management architecture and ES procedures Inaddition in order to reduce the computational complexitythe model is divided into four stages which are trafficvariation monitoring regional BS mode determination BS-user association and power supply

To address these four problems we design S-ARIMAbased traffic prediction algorithm low-complexity BS coop-eration algorithm based on geographical topology dis-tributed user allocation algorithm and sustainable powersupply algorithm respectively Finally the simulations areconducted in LTE underlay heterogeneous network Resultsshow that the mechanism can obtain more regional ESgains compared to other algorithms while maintaining usercoverage interference and QoS above acceptable levels

The rest of the paper is organized as follows Section 2presents ES optimization model in LTE heterogeneous net-work which develops resource allocation and BS energy con-sumptionmodel and three key points for resolving themodelare analyzed then Section 3 gives the self-organized frame-work for ESM which includes the management architec-ture the management entities and management proceduresBased on the framework Section 4 proposes correspondingSeasonal Auto-Regressive Integrated Moving Average (S-ARIMA) based traffic prediction algorithm BS cooperationalgorithms based on geographical topology distributed user

allocation algorithm and sustainable power supply algorithmwith complexity analysis Section 5 numerically evaluates themechanism and algorithm Conclusions and future work aregiven in the last section

2 ES Optimization Model

For ES optimization model this paper primary focuses onresource allocation method and BS energy consumptionmodel and constructs an optimizationmodel based on hybridenergy supplies Moreover key points to resolve the problemwill be discussed as well

21 Resource Allocation Method Assuming that there are119873 BSs in underlay LTE heterogeneous network including1198731macro BSs (eNodeB) and 119873

2micro BSs (microcell)

each eNodeB contains three sectors (except for those on theboundary) B

119872and B

119898denote the set of eNodeB and the set

of microcells and B = B119872cup B119898= 119861119895 The network has

119870 kinds of services Supposing that time domain is [0 119879]for 119861119895and user 119906

119894 M119895(119905) is the set of active users of 119861

119895at

time 119905 Assume that X(119905) = 119909119894119895(119905) is the BS association

matrix where 119909119894119895(119905) = 1 if user 119906

119894is served by 119861

119895 otherwise

119909119894119895(119905) = 0 Then Signal to Interference plus Noise Ratio

(SINR) experienced by user 119906119894from 119861

119895at time 119905 is given by

120574119894119895(119905) =

119901119894119895(119905) sdot 119892

119894119895(119905)

N0+I119894(119905) + 119868

119894119895(119905) 119909119894119895(119905) = 1

0 119909119894119895(119905) = 0

(1)

where 119901119894119895(119905) denotes the power for each resource block (RB)

and P(119905) = 119901119894119895(119905) is defined as its matrix 119892

119894119895(119905) denotes

the channel gain between 119861119895and user 119906

119894 N0denotes the

thermal noiseI119894(119905) and 119868

119894119895(119905) denote interference outside 119861

119895

and inside 119861119895for user 119906

119894 respectively

Assuming that AMC (Adaptive Modulation and Coding)is used then spectral efficiency 120593

119894119895(119905) (bpsHz) when user 119906

119894

is served by 119861119895is given by [7]

120593119894119895(119905) =

0 120574119894119895(119905) lt 120574min

120585 log2(1 + 120574

119894119895(119905)) 120574min le 120574119894119895 (119905) lt 120574max

120593max 120574119894119895(119905) ge 120574max

(2)

where 120574min is the minimum SINR and 120585 isin [0 1] is theattenuation factor 120574max and 120593max are the maximum of SINRand spectral efficiency respectively Further we canderive thenumber of RBs for user 119906

119894required from 119861

119895as

120573119894119895(119905) = lceil

119903119894119895(119905)

119882RB120593119894119895 (119905)rceil (3)

where 119903119894119895(119905) is the required service rate from serving119861

119895to user

119906119894and 119882RB is the bandwidth for each RB Then load factor

119871119895(119905) of BS can be defined as

119871119895(119905) =

1

120573119879

119895

|M119895(119905)|

sum

119894=1

120573119894119895(119905) (4)







119895is denoted



119875119895(119905) = (1 minus 120575

119895) 119871119895(119905) 119875119872

119895+ 120601119895(119905) 120575119895119875119872

119895

120601119895(119905) =

1 119871119895(119905) gt 0

120576

120575119895

119871119895(119905) = 0

(5)



119895(119905) To



119895(119905) of each panel

can be expressed as

119890119895(119905) = 119891 (V

119895(119905) 119875119895(119905)) (6)

119864119895(119905) = 119864

0

119895+ int

119905

0

119890119895(119905) 119889119905 (7)


119895(119905) from


be written as

119886119895(119905) = ℎ (V

119895(119905) 119875119895(119905)) (8)

119860119895(119905) = int

119905

0

119886119895(119905) 119889119905 (9)




119894isin M119895(119905) and BSs 119861




119895119896(119905) is the service



119894119895(119905) = 119901

119894119895(119905) sdot 119892

119894119895(119905)


119895at time



120574 Consider

P minX(119905)P(119905)

sum

119895isinB119872119864119895(119879) + sum

119895isinB119872119860119895(119879)

st 119875119861119895119896(119905) le 119875

119861

119879 forall119895 119896 119905

119873

sum

119895=1

119909119894119895(119905) = 1 forall119894 119905

|M119895(119905)|

sum

119894=1

120573119894119895(119905) le 120573

119879 forall119895 119905

|M119895(119905)|

sum

119894=1

120573119894119895(119905) 119901119894119895(119905) le 120572119875

119879

119895(119905) forall119895 119905



(10)







Centralized SON




eNodeB

OAM

SON agent

SON agent SON agent

eNodeB SON agent

X2

MicrocellMicrocell



















BS-userassociation

Power supplypolicy


microcell



at microcell




















119895(119905) = 1 as










120601119909(119862)Φ

119883(119862119910) (1 minus 119862)

119889(1 minus 119862

119910)119863

(119905)

= 120579119902(119862)Θ

119876(119862119910) 120588 (119905)

(11)

where

(119905) =

119871 (119905) minus 120583 119889 = 119863 = 0

119871 (119905) others(12)

with

120601119909(119862) = 1 minus 120601

1119862 minus 120601


119909119862119909

Φ119883(119862119910) = 1 minus Φ


119883119862119883119910

120579119902(119862) = 1 minus 120579


119902119862119902

Θ119876(119862119910) = 1 minus Θ


119876119862119876119910

(13)

In (11)sim(13) 120601119909(119862) and 120579



119883(119862119910) and Θ






















11



Input B L(119905) 119904119895(119905) Output L(119905) 119904

119895(119905)

(1)B = BT = (2) whileB =


119861119895isinB119871119895(119905) | 119904

119895(119905) = 1 ampamp 119871

119895(119905) lt 1



119861119896isinH119895lowast (119905)119871119896(119905) | 119904

119896(119905) = 1 ampamp 119871

119896(119905) lt 1


119895lowast (119905) + 119871

119896lowast (119905) le 1

(7) 119904119895lowast (119905) = 0 119871

119896lowast (119905) lArr 119871

119895lowast (119905) + 119871

119896lowast (119905)

(8) end if(9) end if(10) if 119861



119895lowast



119896(119905) + 119908

119896119871119895lowast (119905) le 1

(14) 119871119896(119905) lArr 119871

119896(119905) + 119908

119896119871119895lowast (119905) 119904

119895lowast (119905) = 0


119895lowast

(18) end while


B1

B2

B3

B4

B5

B6

B7

B8

B9

B10

B11

B12



(1198618 11986110) and macro BS 119861







119895lowast with


119896lowast





119896lowast



119895lowast and TP






119896to BSs in


as

119908119896=

ℓ2

119896119895lowast


2

119894119895lowast

(14)




119898| sdot maxH

119895lowast(119905) +


119895lowast | |TP





119898| + 20|B







119861119895isinBmax120590119894lowast119895(119905) | 119871

119895(119905) lt 1


119894lowast119895lowast (119905)(N

0+ sum119873

119896=1119896 =119895lowast 119875119879

119896(119905)119892119894lowast119896(119905)) lt 120574min


119894lowast119895lowast (119905) + Δ119901 119909

119894lowast119895lowast (119905) = 1


119861

119879 break end if

(7) if sum|M119895lowast (119905)|119894=1

120573119894119895lowast (119905)119901119894119895lowast (119905) gt 120572119875

119879


(8) if 119871119895lowast (119905) + 120573



(11) end while


Input 119875119895(119905) 119860

119895(119905) 119864119895(119905) Output 119860

119895(119905) 119864119895(119905)

(1) if 119864119895(119905) ge int

119905+1

119905119875119895(119905)119889119905

119864119895(119905) = 119864

119895(119905) minus int

119905+1

119905119875119895(119905)119889119905 + int

119905+1

119905V119895(119905)119889119905 and 119886

119895(119905) = 0

(2) else(3) 119864119895(119905) = 119864

119895(119905) + int

119905+1

119905V119895(119905)119889119905 and 119886

119895(119905) = 119875

119895(119905)

(4) end if






119894lowast119895lowast(119905) or 120590







I119894(119905) =

119873

sum

119895=1119895 =119894

119871119894(119905) 119871119895(119905) 119875119879

119895(119905) 119892119894119895(119905) (15)

Then we have

120574119894119895(119905) =

120590119894119895(119905)

N0+ sum119873

119896=1119896 =119895119871119894(119905) 119871119896(119905) 119875119879

119896(119905) 119892119894119896(119905)

ge

120590119894119895(119905)

N0+ sum119873

119896=1119896 =119895119875119879

119896(119905) 119892119894119896(119905)

(16)














0 200 400 600 800 1000 1200 1400 1600 1800 20000

200

400

600

800

1000

1200

1400

1600

1800

2000

(m)

(m)







with eNodeB number






coefficient 0996

002040608

1

1 9 17 25 33 41 49 57 65 73 81 89 97 105 113

Nor

mal

ized

traffi

c

Time (h)



0005

01015

02025

03035

Serv

ice a

rriv

al ra

te (

s)

Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

001020304050607

Pow

er g

ener

atio

n ra

te (k

W)




1198791

120572 09 mdash120594 minus105 dBm119875120590

97 119875120574






Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

020406080

100120140160180200220240260280300

Accu

mul

ated

ener

gy (k

WH

)







119866119864(119879) =

119865 (119879) minus 1198651015840(119879)

119865 (119879)times 100 (17)


05

101520253035404550

OP

in [2

3]

TP in

[24]

Gre

enBS

N in

[5]

ES u

nder

pow

ergr

id

ES u

nder

hyb

ridpo

wer

supp

ly

ES-gain ()








01

02

03

04

05

06

07

08

09

1

RSRP (dBm)

Accu

mul

ativ

e pro

babi

lity


With ES (users)


0

01

02

03

04

05

06

07

08

09

1

Accu

mul

ativ

e pro

babi

lity



With ES (users)




100 200 300 400 500 600 700 800 900 100025

30

35

40

45

50

55

60

65

70

75


ES effi

cien

cy (

)







6 Conclusion




Competing Interests


Acknowledgments


References



































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in









119895is denoted



119875119895(119905) = (1 minus 120575

119895) 119871119895(119905) 119875119872

119895+ 120601119895(119905) 120575119895119875119872

119895

120601119895(119905) =

1 119871119895(119905) gt 0

120576

120575119895

119871119895(119905) = 0

(5)



119895(119905) To



119895(119905) of each panel

can be expressed as

119890119895(119905) = 119891 (V

119895(119905) 119875119895(119905)) (6)

119864119895(119905) = 119864

0

119895+ int

119905

0

119890119895(119905) 119889119905 (7)


119895(119905) from


be written as

119886119895(119905) = ℎ (V

119895(119905) 119875119895(119905)) (8)

119860119895(119905) = int

119905

0

119886119895(119905) 119889119905 (9)




119894isin M119895(119905) and BSs 119861




119895119896(119905) is the service



119894119895(119905) = 119901

119894119895(119905) sdot 119892

119894119895(119905)


119895at time



120574 Consider

P minX(119905)P(119905)

sum

119895isinB119872119864119895(119879) + sum

119895isinB119872119860119895(119879)

st 119875119861119895119896(119905) le 119875

119861

119879 forall119895 119896 119905

119873

sum

119895=1

119909119894119895(119905) = 1 forall119894 119905

|M119895(119905)|

sum

119894=1

120573119894119895(119905) le 120573

119879 forall119895 119905

|M119895(119905)|

sum

119894=1

120573119894119895(119905) 119901119894119895(119905) le 120572119875

119879

119895(119905) forall119895 119905



(10)







Centralized SON




eNodeB

OAM

SON agent

SON agent SON agent

eNodeB SON agent

X2

MicrocellMicrocell



















BS-userassociation

Power supplypolicy


microcell



at microcell




















119895(119905) = 1 as










120601119909(119862)Φ

119883(119862119910) (1 minus 119862)

119889(1 minus 119862

119910)119863

(119905)

= 120579119902(119862)Θ

119876(119862119910) 120588 (119905)

(11)

where

(119905) =

119871 (119905) minus 120583 119889 = 119863 = 0

119871 (119905) others(12)

with

120601119909(119862) = 1 minus 120601

1119862 minus 120601


119909119862119909

Φ119883(119862119910) = 1 minus Φ


119883119862119883119910

120579119902(119862) = 1 minus 120579


119902119862119902

Θ119876(119862119910) = 1 minus Θ


119876119862119876119910

(13)

In (11)sim(13) 120601119909(119862) and 120579



119883(119862119910) and Θ






















11



Input B L(119905) 119904119895(119905) Output L(119905) 119904

119895(119905)

(1)B = BT = (2) whileB =


119861119895isinB119871119895(119905) | 119904

119895(119905) = 1 ampamp 119871

119895(119905) lt 1



119861119896isinH119895lowast (119905)119871119896(119905) | 119904

119896(119905) = 1 ampamp 119871

119896(119905) lt 1


119895lowast (119905) + 119871

119896lowast (119905) le 1

(7) 119904119895lowast (119905) = 0 119871

119896lowast (119905) lArr 119871

119895lowast (119905) + 119871

119896lowast (119905)

(8) end if(9) end if(10) if 119861



119895lowast



119896(119905) + 119908

119896119871119895lowast (119905) le 1

(14) 119871119896(119905) lArr 119871

119896(119905) + 119908

119896119871119895lowast (119905) 119904

119895lowast (119905) = 0


119895lowast

(18) end while


B1

B2

B3

B4

B5

B6

B7

B8

B9

B10

B11

B12



(1198618 11986110) and macro BS 119861







119895lowast with


119896lowast





119896lowast



119895lowast and TP






119896to BSs in


as

119908119896=

ℓ2

119896119895lowast


2

119894119895lowast

(14)




119898| sdot maxH

119895lowast(119905) +


119895lowast | |TP





119898| + 20|B







119861119895isinBmax120590119894lowast119895(119905) | 119871

119895(119905) lt 1


119894lowast119895lowast (119905)(N

0+ sum119873

119896=1119896 =119895lowast 119875119879

119896(119905)119892119894lowast119896(119905)) lt 120574min


119894lowast119895lowast (119905) + Δ119901 119909

119894lowast119895lowast (119905) = 1


119861

119879 break end if

(7) if sum|M119895lowast (119905)|119894=1

120573119894119895lowast (119905)119901119894119895lowast (119905) gt 120572119875

119879


(8) if 119871119895lowast (119905) + 120573



(11) end while


Input 119875119895(119905) 119860

119895(119905) 119864119895(119905) Output 119860

119895(119905) 119864119895(119905)

(1) if 119864119895(119905) ge int

119905+1

119905119875119895(119905)119889119905

119864119895(119905) = 119864

119895(119905) minus int

119905+1

119905119875119895(119905)119889119905 + int

119905+1

119905V119895(119905)119889119905 and 119886

119895(119905) = 0

(2) else(3) 119864119895(119905) = 119864

119895(119905) + int

119905+1

119905V119895(119905)119889119905 and 119886

119895(119905) = 119875

119895(119905)

(4) end if






119894lowast119895lowast(119905) or 120590







I119894(119905) =

119873

sum

119895=1119895 =119894

119871119894(119905) 119871119895(119905) 119875119879

119895(119905) 119892119894119895(119905) (15)

Then we have

120574119894119895(119905) =

120590119894119895(119905)

N0+ sum119873

119896=1119896 =119895119871119894(119905) 119871119896(119905) 119875119879

119896(119905) 119892119894119896(119905)

ge

120590119894119895(119905)

N0+ sum119873

119896=1119896 =119895119875119879

119896(119905) 119892119894119896(119905)

(16)














0 200 400 600 800 1000 1200 1400 1600 1800 20000

200

400

600

800

1000

1200

1400

1600

1800

2000

(m)

(m)







with eNodeB number






coefficient 0996

002040608

1

1 9 17 25 33 41 49 57 65 73 81 89 97 105 113

Nor

mal

ized

traffi

c

Time (h)



0005

01015

02025

03035

Serv

ice a

rriv

al ra

te (

s)

Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

001020304050607

Pow

er g

ener

atio

n ra

te (k

W)




1198791

120572 09 mdash120594 minus105 dBm119875120590

97 119875120574






Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

020406080

100120140160180200220240260280300

Accu

mul

ated

ener

gy (k

WH

)







119866119864(119879) =

119865 (119879) minus 1198651015840(119879)

119865 (119879)times 100 (17)


05

101520253035404550

OP

in [2

3]

TP in

[24]

Gre

enBS

N in

[5]

ES u

nder

pow

ergr

id

ES u

nder

hyb

ridpo

wer

supp

ly

ES-gain ()








01

02

03

04

05

06

07

08

09

1

RSRP (dBm)

Accu

mul

ativ

e pro

babi

lity


With ES (users)


0

01

02

03

04

05

06

07

08

09

1

Accu

mul

ativ

e pro

babi

lity



With ES (users)




100 200 300 400 500 600 700 800 900 100025

30

35

40

45

50

55

60

65

70

75


ES effi

cien

cy (

)







6 Conclusion




Competing Interests


Acknowledgments


References



































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





Centralized SON




eNodeB

OAM

SON agent

SON agent SON agent

eNodeB SON agent

X2

MicrocellMicrocell



















BS-userassociation

Power supplypolicy


microcell



at microcell




















119895(119905) = 1 as










120601119909(119862)Φ

119883(119862119910) (1 minus 119862)

119889(1 minus 119862

119910)119863

(119905)

= 120579119902(119862)Θ

119876(119862119910) 120588 (119905)

(11)

where

(119905) =

119871 (119905) minus 120583 119889 = 119863 = 0

119871 (119905) others(12)

with

120601119909(119862) = 1 minus 120601

1119862 minus 120601


119909119862119909

Φ119883(119862119910) = 1 minus Φ


119883119862119883119910

120579119902(119862) = 1 minus 120579


119902119862119902

Θ119876(119862119910) = 1 minus Θ


119876119862119876119910

(13)

In (11)sim(13) 120601119909(119862) and 120579



119883(119862119910) and Θ






















11



Input B L(119905) 119904119895(119905) Output L(119905) 119904

119895(119905)

(1)B = BT = (2) whileB =


119861119895isinB119871119895(119905) | 119904

119895(119905) = 1 ampamp 119871

119895(119905) lt 1



119861119896isinH119895lowast (119905)119871119896(119905) | 119904

119896(119905) = 1 ampamp 119871

119896(119905) lt 1


119895lowast (119905) + 119871

119896lowast (119905) le 1

(7) 119904119895lowast (119905) = 0 119871

119896lowast (119905) lArr 119871

119895lowast (119905) + 119871

119896lowast (119905)

(8) end if(9) end if(10) if 119861



119895lowast



119896(119905) + 119908

119896119871119895lowast (119905) le 1

(14) 119871119896(119905) lArr 119871

119896(119905) + 119908

119896119871119895lowast (119905) 119904

119895lowast (119905) = 0


119895lowast

(18) end while


B1

B2

B3

B4

B5

B6

B7

B8

B9

B10

B11

B12



(1198618 11986110) and macro BS 119861







119895lowast with


119896lowast





119896lowast



119895lowast and TP






119896to BSs in


as

119908119896=

ℓ2

119896119895lowast


2

119894119895lowast

(14)




119898| sdot maxH

119895lowast(119905) +


119895lowast | |TP





119898| + 20|B







119861119895isinBmax120590119894lowast119895(119905) | 119871

119895(119905) lt 1


119894lowast119895lowast (119905)(N

0+ sum119873

119896=1119896 =119895lowast 119875119879

119896(119905)119892119894lowast119896(119905)) lt 120574min


119894lowast119895lowast (119905) + Δ119901 119909

119894lowast119895lowast (119905) = 1


119861

119879 break end if

(7) if sum|M119895lowast (119905)|119894=1

120573119894119895lowast (119905)119901119894119895lowast (119905) gt 120572119875

119879


(8) if 119871119895lowast (119905) + 120573



(11) end while


Input 119875119895(119905) 119860

119895(119905) 119864119895(119905) Output 119860

119895(119905) 119864119895(119905)

(1) if 119864119895(119905) ge int

119905+1

119905119875119895(119905)119889119905

119864119895(119905) = 119864

119895(119905) minus int

119905+1

119905119875119895(119905)119889119905 + int

119905+1

119905V119895(119905)119889119905 and 119886

119895(119905) = 0

(2) else(3) 119864119895(119905) = 119864

119895(119905) + int

119905+1

119905V119895(119905)119889119905 and 119886

119895(119905) = 119875

119895(119905)

(4) end if






119894lowast119895lowast(119905) or 120590







I119894(119905) =

119873

sum

119895=1119895 =119894

119871119894(119905) 119871119895(119905) 119875119879

119895(119905) 119892119894119895(119905) (15)

Then we have

120574119894119895(119905) =

120590119894119895(119905)

N0+ sum119873

119896=1119896 =119895119871119894(119905) 119871119896(119905) 119875119879

119896(119905) 119892119894119896(119905)

ge

120590119894119895(119905)

N0+ sum119873

119896=1119896 =119895119875119879

119896(119905) 119892119894119896(119905)

(16)














0 200 400 600 800 1000 1200 1400 1600 1800 20000

200

400

600

800

1000

1200

1400

1600

1800

2000

(m)

(m)







with eNodeB number






coefficient 0996

002040608

1

1 9 17 25 33 41 49 57 65 73 81 89 97 105 113

Nor

mal

ized

traffi

c

Time (h)



0005

01015

02025

03035

Serv

ice a

rriv

al ra

te (

s)

Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

001020304050607

Pow

er g

ener

atio

n ra

te (k

W)




1198791

120572 09 mdash120594 minus105 dBm119875120590

97 119875120574






Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

020406080

100120140160180200220240260280300

Accu

mul

ated

ener

gy (k

WH

)







119866119864(119879) =

119865 (119879) minus 1198651015840(119879)

119865 (119879)times 100 (17)


05

101520253035404550

OP

in [2

3]

TP in

[24]

Gre

enBS

N in

[5]

ES u

nder

pow

ergr

id

ES u

nder

hyb

ridpo

wer

supp

ly

ES-gain ()








01

02

03

04

05

06

07

08

09

1

RSRP (dBm)

Accu

mul

ativ

e pro

babi

lity


With ES (users)


0

01

02

03

04

05

06

07

08

09

1

Accu

mul

ativ

e pro

babi

lity



With ES (users)




100 200 300 400 500 600 700 800 900 100025

30

35

40

45

50

55

60

65

70

75


ES effi

cien

cy (

)







6 Conclusion




Competing Interests


Acknowledgments


References



































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in






BS-userassociation

Power supplypolicy


microcell



at microcell




















119895(119905) = 1 as










120601119909(119862)Φ

119883(119862119910) (1 minus 119862)

119889(1 minus 119862

119910)119863

(119905)

= 120579119902(119862)Θ

119876(119862119910) 120588 (119905)

(11)

where

(119905) =

119871 (119905) minus 120583 119889 = 119863 = 0

119871 (119905) others(12)

with

120601119909(119862) = 1 minus 120601

1119862 minus 120601


119909119862119909

Φ119883(119862119910) = 1 minus Φ


119883119862119883119910

120579119902(119862) = 1 minus 120579


119902119862119902

Θ119876(119862119910) = 1 minus Θ


119876119862119876119910

(13)

In (11)sim(13) 120601119909(119862) and 120579



119883(119862119910) and Θ






















11



Input B L(119905) 119904119895(119905) Output L(119905) 119904

119895(119905)

(1)B = BT = (2) whileB =


119861119895isinB119871119895(119905) | 119904

119895(119905) = 1 ampamp 119871

119895(119905) lt 1



119861119896isinH119895lowast (119905)119871119896(119905) | 119904

119896(119905) = 1 ampamp 119871

119896(119905) lt 1


119895lowast (119905) + 119871

119896lowast (119905) le 1

(7) 119904119895lowast (119905) = 0 119871

119896lowast (119905) lArr 119871

119895lowast (119905) + 119871

119896lowast (119905)

(8) end if(9) end if(10) if 119861



119895lowast



119896(119905) + 119908

119896119871119895lowast (119905) le 1

(14) 119871119896(119905) lArr 119871

119896(119905) + 119908

119896119871119895lowast (119905) 119904

119895lowast (119905) = 0


119895lowast

(18) end while


B1

B2

B3

B4

B5

B6

B7

B8

B9

B10

B11

B12



(1198618 11986110) and macro BS 119861







119895lowast with


119896lowast





119896lowast



119895lowast and TP






119896to BSs in


as

119908119896=

ℓ2

119896119895lowast


2

119894119895lowast

(14)




119898| sdot maxH

119895lowast(119905) +


119895lowast | |TP





119898| + 20|B







119861119895isinBmax120590119894lowast119895(119905) | 119871

119895(119905) lt 1


119894lowast119895lowast (119905)(N

0+ sum119873

119896=1119896 =119895lowast 119875119879

119896(119905)119892119894lowast119896(119905)) lt 120574min


119894lowast119895lowast (119905) + Δ119901 119909

119894lowast119895lowast (119905) = 1


119861

119879 break end if

(7) if sum|M119895lowast (119905)|119894=1

120573119894119895lowast (119905)119901119894119895lowast (119905) gt 120572119875

119879


(8) if 119871119895lowast (119905) + 120573



(11) end while


Input 119875119895(119905) 119860

119895(119905) 119864119895(119905) Output 119860

119895(119905) 119864119895(119905)

(1) if 119864119895(119905) ge int

119905+1

119905119875119895(119905)119889119905

119864119895(119905) = 119864

119895(119905) minus int

119905+1

119905119875119895(119905)119889119905 + int

119905+1

119905V119895(119905)119889119905 and 119886

119895(119905) = 0

(2) else(3) 119864119895(119905) = 119864

119895(119905) + int

119905+1

119905V119895(119905)119889119905 and 119886

119895(119905) = 119875

119895(119905)

(4) end if






119894lowast119895lowast(119905) or 120590







I119894(119905) =

119873

sum

119895=1119895 =119894

119871119894(119905) 119871119895(119905) 119875119879

119895(119905) 119892119894119895(119905) (15)

Then we have

120574119894119895(119905) =

120590119894119895(119905)

N0+ sum119873

119896=1119896 =119895119871119894(119905) 119871119896(119905) 119875119879

119896(119905) 119892119894119896(119905)

ge

120590119894119895(119905)

N0+ sum119873

119896=1119896 =119895119875119879

119896(119905) 119892119894119896(119905)

(16)














0 200 400 600 800 1000 1200 1400 1600 1800 20000

200

400

600

800

1000

1200

1400

1600

1800

2000

(m)

(m)







with eNodeB number






coefficient 0996

002040608

1

1 9 17 25 33 41 49 57 65 73 81 89 97 105 113

Nor

mal

ized

traffi

c

Time (h)



0005

01015

02025

03035

Serv

ice a

rriv

al ra

te (

s)

Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

001020304050607

Pow

er g

ener

atio

n ra

te (k

W)




1198791

120572 09 mdash120594 minus105 dBm119875120590

97 119875120574






Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

020406080

100120140160180200220240260280300

Accu

mul

ated

ener

gy (k

WH

)







119866119864(119879) =

119865 (119879) minus 1198651015840(119879)

119865 (119879)times 100 (17)


05

101520253035404550

OP

in [2

3]

TP in

[24]

Gre

enBS

N in

[5]

ES u

nder

pow

ergr

id

ES u

nder

hyb

ridpo

wer

supp

ly

ES-gain ()








01

02

03

04

05

06

07

08

09

1

RSRP (dBm)

Accu

mul

ativ

e pro

babi

lity


With ES (users)


0

01

02

03

04

05

06

07

08

09

1

Accu

mul

ativ

e pro

babi

lity



With ES (users)




100 200 300 400 500 600 700 800 900 100025

30

35

40

45

50

55

60

65

70

75


ES effi

cien

cy (

)







6 Conclusion




Competing Interests


Acknowledgments


References



































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in










120601119909(119862)Φ

119883(119862119910) (1 minus 119862)

119889(1 minus 119862

119910)119863

(119905)

= 120579119902(119862)Θ

119876(119862119910) 120588 (119905)

(11)

where

(119905) =

119871 (119905) minus 120583 119889 = 119863 = 0

119871 (119905) others(12)

with

120601119909(119862) = 1 minus 120601

1119862 minus 120601


119909119862119909

Φ119883(119862119910) = 1 minus Φ


119883119862119883119910

120579119902(119862) = 1 minus 120579


119902119862119902

Θ119876(119862119910) = 1 minus Θ


119876119862119876119910

(13)

In (11)sim(13) 120601119909(119862) and 120579



119883(119862119910) and Θ






















11



Input B L(119905) 119904119895(119905) Output L(119905) 119904

119895(119905)

(1)B = BT = (2) whileB =


119861119895isinB119871119895(119905) | 119904

119895(119905) = 1 ampamp 119871

119895(119905) lt 1



119861119896isinH119895lowast (119905)119871119896(119905) | 119904

119896(119905) = 1 ampamp 119871

119896(119905) lt 1


119895lowast (119905) + 119871

119896lowast (119905) le 1

(7) 119904119895lowast (119905) = 0 119871

119896lowast (119905) lArr 119871

119895lowast (119905) + 119871

119896lowast (119905)

(8) end if(9) end if(10) if 119861



119895lowast



119896(119905) + 119908

119896119871119895lowast (119905) le 1

(14) 119871119896(119905) lArr 119871

119896(119905) + 119908

119896119871119895lowast (119905) 119904

119895lowast (119905) = 0


119895lowast

(18) end while


B1

B2

B3

B4

B5

B6

B7

B8

B9

B10

B11

B12



(1198618 11986110) and macro BS 119861







119895lowast with


119896lowast





119896lowast



119895lowast and TP






119896to BSs in


as

119908119896=

ℓ2

119896119895lowast


2

119894119895lowast

(14)




119898| sdot maxH

119895lowast(119905) +


119895lowast | |TP





119898| + 20|B







119861119895isinBmax120590119894lowast119895(119905) | 119871

119895(119905) lt 1


119894lowast119895lowast (119905)(N

0+ sum119873

119896=1119896 =119895lowast 119875119879

119896(119905)119892119894lowast119896(119905)) lt 120574min


119894lowast119895lowast (119905) + Δ119901 119909

119894lowast119895lowast (119905) = 1


119861

119879 break end if

(7) if sum|M119895lowast (119905)|119894=1

120573119894119895lowast (119905)119901119894119895lowast (119905) gt 120572119875

119879


(8) if 119871119895lowast (119905) + 120573



(11) end while


Input 119875119895(119905) 119860

119895(119905) 119864119895(119905) Output 119860

119895(119905) 119864119895(119905)

(1) if 119864119895(119905) ge int

119905+1

119905119875119895(119905)119889119905

119864119895(119905) = 119864

119895(119905) minus int

119905+1

119905119875119895(119905)119889119905 + int

119905+1

119905V119895(119905)119889119905 and 119886

119895(119905) = 0

(2) else(3) 119864119895(119905) = 119864

119895(119905) + int

119905+1

119905V119895(119905)119889119905 and 119886

119895(119905) = 119875

119895(119905)

(4) end if






119894lowast119895lowast(119905) or 120590







I119894(119905) =

119873

sum

119895=1119895 =119894

119871119894(119905) 119871119895(119905) 119875119879

119895(119905) 119892119894119895(119905) (15)

Then we have

120574119894119895(119905) =

120590119894119895(119905)

N0+ sum119873

119896=1119896 =119895119871119894(119905) 119871119896(119905) 119875119879

119896(119905) 119892119894119896(119905)

ge

120590119894119895(119905)

N0+ sum119873

119896=1119896 =119895119875119879

119896(119905) 119892119894119896(119905)

(16)














0 200 400 600 800 1000 1200 1400 1600 1800 20000

200

400

600

800

1000

1200

1400

1600

1800

2000

(m)

(m)







with eNodeB number






coefficient 0996

002040608

1

1 9 17 25 33 41 49 57 65 73 81 89 97 105 113

Nor

mal

ized

traffi

c

Time (h)



0005

01015

02025

03035

Serv

ice a

rriv

al ra

te (

s)

Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

001020304050607

Pow

er g

ener

atio

n ra

te (k

W)




1198791

120572 09 mdash120594 minus105 dBm119875120590

97 119875120574






Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

020406080

100120140160180200220240260280300

Accu

mul

ated

ener

gy (k

WH

)







119866119864(119879) =

119865 (119879) minus 1198651015840(119879)

119865 (119879)times 100 (17)


05

101520253035404550

OP

in [2

3]

TP in

[24]

Gre

enBS

N in

[5]

ES u

nder

pow

ergr

id

ES u

nder

hyb

ridpo

wer

supp

ly

ES-gain ()








01

02

03

04

05

06

07

08

09

1

RSRP (dBm)

Accu

mul

ativ

e pro

babi

lity


With ES (users)


0

01

02

03

04

05

06

07

08

09

1

Accu

mul

ativ

e pro

babi

lity



With ES (users)




100 200 300 400 500 600 700 800 900 100025

30

35

40

45

50

55

60

65

70

75


ES effi

cien

cy (

)







6 Conclusion




Competing Interests


Acknowledgments


References



































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




Input B L(119905) 119904119895(119905) Output L(119905) 119904

119895(119905)

(1)B = BT = (2) whileB =


119861119895isinB119871119895(119905) | 119904

119895(119905) = 1 ampamp 119871

119895(119905) lt 1



119861119896isinH119895lowast (119905)119871119896(119905) | 119904

119896(119905) = 1 ampamp 119871

119896(119905) lt 1


119895lowast (119905) + 119871

119896lowast (119905) le 1

(7) 119904119895lowast (119905) = 0 119871

119896lowast (119905) lArr 119871

119895lowast (119905) + 119871

119896lowast (119905)

(8) end if(9) end if(10) if 119861



119895lowast



119896(119905) + 119908

119896119871119895lowast (119905) le 1

(14) 119871119896(119905) lArr 119871

119896(119905) + 119908

119896119871119895lowast (119905) 119904

119895lowast (119905) = 0


119895lowast

(18) end while


B1

B2

B3

B4

B5

B6

B7

B8

B9

B10

B11

B12



(1198618 11986110) and macro BS 119861







119895lowast with


119896lowast





119896lowast



119895lowast and TP






119896to BSs in


as

119908119896=

ℓ2

119896119895lowast


2

119894119895lowast

(14)




119898| sdot maxH

119895lowast(119905) +


119895lowast | |TP





119898| + 20|B







119861119895isinBmax120590119894lowast119895(119905) | 119871

119895(119905) lt 1


119894lowast119895lowast (119905)(N

0+ sum119873

119896=1119896 =119895lowast 119875119879

119896(119905)119892119894lowast119896(119905)) lt 120574min


119894lowast119895lowast (119905) + Δ119901 119909

119894lowast119895lowast (119905) = 1


119861

119879 break end if

(7) if sum|M119895lowast (119905)|119894=1

120573119894119895lowast (119905)119901119894119895lowast (119905) gt 120572119875

119879


(8) if 119871119895lowast (119905) + 120573



(11) end while


Input 119875119895(119905) 119860

119895(119905) 119864119895(119905) Output 119860

119895(119905) 119864119895(119905)

(1) if 119864119895(119905) ge int

119905+1

119905119875119895(119905)119889119905

119864119895(119905) = 119864

119895(119905) minus int

119905+1

119905119875119895(119905)119889119905 + int

119905+1

119905V119895(119905)119889119905 and 119886

119895(119905) = 0

(2) else(3) 119864119895(119905) = 119864

119895(119905) + int

119905+1

119905V119895(119905)119889119905 and 119886

119895(119905) = 119875

119895(119905)

(4) end if






119894lowast119895lowast(119905) or 120590







I119894(119905) =

119873

sum

119895=1119895 =119894

119871119894(119905) 119871119895(119905) 119875119879

119895(119905) 119892119894119895(119905) (15)

Then we have

120574119894119895(119905) =

120590119894119895(119905)

N0+ sum119873

119896=1119896 =119895119871119894(119905) 119871119896(119905) 119875119879

119896(119905) 119892119894119896(119905)

ge

120590119894119895(119905)

N0+ sum119873

119896=1119896 =119895119875119879

119896(119905) 119892119894119896(119905)

(16)














0 200 400 600 800 1000 1200 1400 1600 1800 20000

200

400

600

800

1000

1200

1400

1600

1800

2000

(m)

(m)







with eNodeB number






coefficient 0996

002040608

1

1 9 17 25 33 41 49 57 65 73 81 89 97 105 113

Nor

mal

ized

traffi

c

Time (h)



0005

01015

02025

03035

Serv

ice a

rriv

al ra

te (

s)

Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

001020304050607

Pow

er g

ener

atio

n ra

te (k

W)




1198791

120572 09 mdash120594 minus105 dBm119875120590

97 119875120574






Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

020406080

100120140160180200220240260280300

Accu

mul

ated

ener

gy (k

WH

)







119866119864(119879) =

119865 (119879) minus 1198651015840(119879)

119865 (119879)times 100 (17)


05

101520253035404550

OP

in [2

3]

TP in

[24]

Gre

enBS

N in

[5]

ES u

nder

pow

ergr

id

ES u

nder

hyb

ridpo

wer

supp

ly

ES-gain ()








01

02

03

04

05

06

07

08

09

1

RSRP (dBm)

Accu

mul

ativ

e pro

babi

lity


With ES (users)


0

01

02

03

04

05

06

07

08

09

1

Accu

mul

ativ

e pro

babi

lity



With ES (users)




100 200 300 400 500 600 700 800 900 100025

30

35

40

45

50

55

60

65

70

75


ES effi

cien

cy (

)







6 Conclusion




Competing Interests


Acknowledgments


References



































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in






119861119895isinBmax120590119894lowast119895(119905) | 119871

119895(119905) lt 1


119894lowast119895lowast (119905)(N

0+ sum119873

119896=1119896 =119895lowast 119875119879

119896(119905)119892119894lowast119896(119905)) lt 120574min


119894lowast119895lowast (119905) + Δ119901 119909

119894lowast119895lowast (119905) = 1


119861

119879 break end if

(7) if sum|M119895lowast (119905)|119894=1

120573119894119895lowast (119905)119901119894119895lowast (119905) gt 120572119875

119879


(8) if 119871119895lowast (119905) + 120573



(11) end while


Input 119875119895(119905) 119860

119895(119905) 119864119895(119905) Output 119860

119895(119905) 119864119895(119905)

(1) if 119864119895(119905) ge int

119905+1

119905119875119895(119905)119889119905

119864119895(119905) = 119864

119895(119905) minus int

119905+1

119905119875119895(119905)119889119905 + int

119905+1

119905V119895(119905)119889119905 and 119886

119895(119905) = 0

(2) else(3) 119864119895(119905) = 119864

119895(119905) + int

119905+1

119905V119895(119905)119889119905 and 119886

119895(119905) = 119875

119895(119905)

(4) end if






119894lowast119895lowast(119905) or 120590







I119894(119905) =

119873

sum

119895=1119895 =119894

119871119894(119905) 119871119895(119905) 119875119879

119895(119905) 119892119894119895(119905) (15)

Then we have

120574119894119895(119905) =

120590119894119895(119905)

N0+ sum119873

119896=1119896 =119895119871119894(119905) 119871119896(119905) 119875119879

119896(119905) 119892119894119896(119905)

ge

120590119894119895(119905)

N0+ sum119873

119896=1119896 =119895119875119879

119896(119905) 119892119894119896(119905)

(16)














0 200 400 600 800 1000 1200 1400 1600 1800 20000

200

400

600

800

1000

1200

1400

1600

1800

2000

(m)

(m)







with eNodeB number






coefficient 0996

002040608

1

1 9 17 25 33 41 49 57 65 73 81 89 97 105 113

Nor

mal

ized

traffi

c

Time (h)



0005

01015

02025

03035

Serv

ice a

rriv

al ra

te (

s)

Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

001020304050607

Pow

er g

ener

atio

n ra

te (k

W)




1198791

120572 09 mdash120594 minus105 dBm119875120590

97 119875120574






Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

020406080

100120140160180200220240260280300

Accu

mul

ated

ener

gy (k

WH

)







119866119864(119879) =

119865 (119879) minus 1198651015840(119879)

119865 (119879)times 100 (17)


05

101520253035404550

OP

in [2

3]

TP in

[24]

Gre

enBS

N in

[5]

ES u

nder

pow

ergr

id

ES u

nder

hyb

ridpo

wer

supp

ly

ES-gain ()








01

02

03

04

05

06

07

08

09

1

RSRP (dBm)

Accu

mul

ativ

e pro

babi

lity


With ES (users)


0

01

02

03

04

05

06

07

08

09

1

Accu

mul

ativ

e pro

babi

lity



With ES (users)




100 200 300 400 500 600 700 800 900 100025

30

35

40

45

50

55

60

65

70

75


ES effi

cien

cy (

)







6 Conclusion




Competing Interests


Acknowledgments


References



































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




0 200 400 600 800 1000 1200 1400 1600 1800 20000

200

400

600

800

1000

1200

1400

1600

1800

2000

(m)

(m)







with eNodeB number






coefficient 0996

002040608

1

1 9 17 25 33 41 49 57 65 73 81 89 97 105 113

Nor

mal

ized

traffi

c

Time (h)



0005

01015

02025

03035

Serv

ice a

rriv

al ra

te (

s)

Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

001020304050607

Pow

er g

ener

atio

n ra

te (k

W)




1198791

120572 09 mdash120594 minus105 dBm119875120590

97 119875120574






Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

020406080

100120140160180200220240260280300

Accu

mul

ated

ener

gy (k

WH

)







119866119864(119879) =

119865 (119879) minus 1198651015840(119879)

119865 (119879)times 100 (17)


05

101520253035404550

OP

in [2

3]

TP in

[24]

Gre

enBS

N in

[5]

ES u

nder

pow

ergr

id

ES u

nder

hyb

ridpo

wer

supp

ly

ES-gain ()








01

02

03

04

05

06

07

08

09

1

RSRP (dBm)

Accu

mul

ativ

e pro

babi

lity


With ES (users)


0

01

02

03

04

05

06

07

08

09

1

Accu

mul

ativ

e pro

babi

lity



With ES (users)




100 200 300 400 500 600 700 800 900 100025

30

35

40

45

50

55

60

65

70

75


ES effi

cien

cy (

)







6 Conclusion




Competing Interests


Acknowledgments


References



































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




Time (h)1 3 5 7 9 11 13 15 17 19 21 23 25 27 29

020406080

100120140160180200220240260280300

Accu

mul

ated

ener

gy (k

WH

)







119866119864(119879) =

119865 (119879) minus 1198651015840(119879)

119865 (119879)times 100 (17)


05

101520253035404550

OP

in [2

3]

TP in

[24]

Gre

enBS

N in

[5]

ES u

nder

pow

ergr

id

ES u

nder

hyb

ridpo

wer

supp

ly

ES-gain ()








01

02

03

04

05

06

07

08

09

1

RSRP (dBm)

Accu

mul

ativ

e pro

babi

lity


With ES (users)


0

01

02

03

04

05

06

07

08

09

1

Accu

mul

ativ

e pro

babi

lity



With ES (users)




100 200 300 400 500 600 700 800 900 100025

30

35

40

45

50

55

60

65

70

75


ES effi

cien

cy (

)







6 Conclusion




Competing Interests


Acknowledgments


References



































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





01

02

03

04

05

06

07

08

09

1

RSRP (dBm)

Accu

mul

ativ

e pro

babi

lity


With ES (users)


0

01

02

03

04

05

06

07

08

09

1

Accu

mul

ativ

e pro

babi

lity



With ES (users)




100 200 300 400 500 600 700 800 900 100025

30

35

40

45

50

55

60

65

70

75


ES effi

cien

cy (

)







6 Conclusion




Competing Interests


Acknowledgments


References



































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





Competing Interests


Acknowledgments


References



































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in













Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in



Research Article Energy-Saving Management Mechanism Based...

Documents

Transcript of Research Article Energy-Saving Management Mechanism Based...