research article joint radio resource allocation and base

Research ArticleJoint Radio Resource Allocation and Base StationLocation Selection in OFDMA Based Private Wireless AccessNetworks for Smart Grid

Peng Du1 and Yuan Zhang2

1College of Automation Nanjing University of Posts and Telecommunications Nanjing 210023 China2National Mobile Communications Research Laboratory Southeast University Nanjing China

Correspondence should be addressed to Yuan Zhang yzhangseueducn

Received 3 February 2016 Revised 11 July 2016 Accepted 1 August 2016

Academic Editor George Tsoulos

Copyright copy 2016 P Du and Y Zhang This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

This paper studies the base stations deployment problem in orthogonal frequency-divisionmultiple access (OFDMA) based privatewireless access networks for smart grid (SG) Firstly we analyze the differences between private wireless access networks for SG andpublic cellular access networks Then we propose scheduling and power control based algorithms for the radio resource allocationsubproblem andK-means simulated annealing (SA) and particle swarm optimization (PSO) based algorithms for the base station(BS) location selection subproblem and iterate over these two sets of algorithms to solve the target problem Simulation results showthat the proposed method can effectively solve the target problem Specifically the combination of power control based resourceallocation algorithm and PSO based location selection algorithm is recommended

1 Introduction

It is critical that the underlying communication technol-ogy shall support efficient data exchange between variousdomains comprising smart grid (SG) [1] This paper studiesthe orthogonal frequency-divisionmultiple access (OFDMA)[2] based private wireless access networks for SG Actuallywireless technologies can be used in the grid for monitoringmetering and data gathering [3ndash17] Specifically SG devicessuch as switching station distribution circuit and distributedenergy sites will produce various information data and sendthem periodically to the base station (BS) to realize theautomation of power distribution and electricity informationacquisition In order to achieve this requirement many tech-nical problems need to be solved Among them this paperaddresses the problem of how to optimize the deployment ofBSs

This problemhas been extensively studied in the literaturefor the public cellular access networks However the scenarioof private wireless access networks for SG is quite differentfrom that of the public cellular access networks

Property 1 (the locations of devices can be considered asfixed) Due to the peculiarity of SG the locations of mostpower devices can be considered as fixed or quasi-fixed

Property 2 (the uplink transmission is dominant [18 19])The smart grid network is an uplink dominated network asthe main data flow is from power devices to control centerThe most often requirement of communications in SG isdevices periodically reporting status monitoring data to thecontrol center via BSs Therefore the direction of most datatransmissions will be uplink

Property 3 (the transmission rate requirement of devices canbe considered as fixed [19]) The communications happen-ing in the SG belong to the type of machine-to-machine(M2M) communications and the uplink data transmissionrate requirement of each device can be considered as fixedor quasi-fixed

Property 4 (the frequency separation requirement [19])Wireless networks are more vulnerable than their wiredcounterparts due to the potential for direct access to the

Hindawi Publishing CorporationJournal of Electrical and Computer EngineeringVolume 2016 Article ID 7948018 13 pageshttpdxdoiorg10115520167948018

2 Journal of Electrical and Computer Engineering

transport medium Hence security must be considered atevery layer of the protocol stack in the private wirelessaccess networks for SG In addition to the authenticationauthorization and encryption considered at the applicationlayer the frequency separation mechanism at the physicallayer shall also be considered which is explained as followsActually data produced by different types of devices in smartgrid shall be transmitted to different destination systems Forexample as shown in Figure 1 the data produced by dataterminal unit (DTU) shall be transmitted to the productionservice system while the data produced by video terminalunit (VTU) shall be transmitted to the management servicesystem Due to the security purpose the transmission pathsused by different types of data shall be separated as muchas possible The separation can be achieved physically orlogically For example four different approaches to constructthe access network are illustrated in Figure 1 where thedata paths in Figure 1(a) are the most separated that is theseparation is achieved physically while the data paths inFigure 1(d) are the least separated that is the separation canbe achieved logically Further in addition to the separationfor the wireline segment data transmission over the wirelesssegment of data path shall also be separated for differenttypes of devices This requires that different types of devicesshall use different frequency channels to transmit their datathat is different types of devices sharing the same frequencychannel will not be allowed This is the frequency separationrequirement considered in this work

As a first step towards addressing the above issuesthis paper investigates the problem of how to deploy BSsand allocate wireless resources so that the uplink trans-mission requirements are efficiently met For this problemwe propose to decompose it into the resource allocationsubproblem and the location selection subproblem and solvethese two subproblems in an iterative fashionThe remainderof the paper is organized as follows Section 2 formulatesthe joint resource allocation and location selection problemSection 3 presents the overall framework to address thisproblem Sections 4 and 5 propose resource allocation andlocation selection algorithms respectively Simulation resultsare reported in Section 6 Finally we conclude in Section 7

2 Problem Formulation

Consider a set of SG devices scattered in an area Ψ Let Hdenote the set of devices For each device 119895 isin H let119862119895 denotethe minimum uplink data rate requirement and 119875119895 the uplinktransmission power The value of 119875119895 shall satisfy 0 le 119875119895 le119875max where119875max is the upper bound For convenience let119862 =119862119895 and 119875 = 119875119895 respectively Assume that all devices areclassified into119870 different types LetH119894 denote the set of type-119894 SG devices 1 le 119894 le 119870

Assume that the private wireless access network consistsof 119861 BSs which are located in the area Ψ Let z119887 = (1199111119887 1199112119887)denote the deployment location of the 119887th BS 1 le 119887 le 119861where 1199111119887 and 1199112119887 are the horizontal and vertical ordinate of thedeployment location respectively For convenience let z =z119887 Not every location inΨ can be the candidate location for

BS Assume thatΘ denotes the candidate BS location set inΨand the deployment location of BS can only be selected fromthe elements of Θ That is we restrict z119887 isin Θ In additionlet Ω = 1198781 1198782 119878119861 denote the relationship between SGdevices and BSs where 119878119887 is the set of devices served bythe 119887th BS For simplicity and without loss of generality weassume that the value of 119878119887 is determined by the distancebased rule That is a device will be served by the BS whichis the closest to it

Consider a OFDMA based private wireless access net-work The radio resource is defined as follows In thefrequency domain assume that the total bandwidth is dividedinto 119873 channels Let 119882 and 1198820 denote total and channelbandwidth in Hertz respectively In the time domain assumethat the time axes are organized into consecutive slots and1198710 consecutive slots constitute a frame The basic resourceunit for data transmission is a resource block (RB) which isdefined as one channel in the frequency domain and one slotin the time domain respectively In each frame assume that119871 slots can be used for uplink communications Therefore foreach channel there are 119871 RBs which are allocatable Finallydefine binary variable 119884119899119897119895 to denote the results of radioresource allocation which is valued 1 if the 119897th RB of the 119899thchannel is allocated to device 119895 and 0 otherwise Each deviceshall be allocated a number of RBs to meet its minimum datarate requirement For convenience let 119884 = 119884119899119897119895

Given that the RB (119899 119897) has been allocated to device 119895 thereceived signal-interference-noise-ratio (SINR) experiencedby device 119895 on this RB at BS 119887 can be written as

120574119899119897119895 = 119875119895119866119895119887119875N + 119875I (1)

where 119866119887119895 is the path loss from device 119895 to BS 119887 119875N isthe power of background noise 119875I = sum119896 =119895119896isinD119899119897 119875119896119866119896119887 isthe power of interference and D119899119897 is the set of deviceswhich share the same RB with device 119895 For simplicity andwithout loss of generality we assume that the path lossmainlydepends on the distance and can be calculated accordingto the formula PL(119909) for a distance separation of 119909 metersand we assume that there is no interference between distantdevices Let 119862119899119897119895 denote the uplink data rate achieved bydevice 119895 on RB (119899 119897) which is calculated by the Shannonformula as

119899119897119895 = 1198820log (1 + 120574119899119897119895)1198710 (2)

Then the total data rate achieved by device 119895 denoted as 119895can be calculated as

119895 = sum(119899119897)119884119899119897119895=1

119899119897119895 (3)

For convenience let 119862 = 119895Finally the problem addressed in this paper can be

formulated as given the parameters Ψ H 119862 119870 H119894119875max 119861 Θ 119882 1198820 119873 1198710 and 119871 how to determine thevalues of deployment location z transmission power 119875 and

Journal of Electrical and Computer Engineering 3

(eg video surveillance)Management servicedistribution automation)Production service (eg

Wireline segmentWirelesssegment

Evolved packet core (EPC)

Baseband unit (BBU)

Remote radio unit (RRU)

Video terminal unit (VTU)

Data terminal unit (DTU)

Time frequency wireless resource

(a)




Baseband unit (BBU)





(b)




Baseband unit (BBU)





(c)




Baseband unit (BBU)





(d)

Figure 1 The separation requirement


radio resource allocation 119884 so that the achieved data rate approaches 119862 as much as possible The symbols used in thispaper are summarized in List of Symbols

3 The Framework

It is difficult to solve z 119875 and 119884 simultaneously Thereforewe decompose the problem into two subproblems The firstis the location selection subproblem which determines zthe second is the resource allocation subproblem whichdetermines 119875 and 119884 Specifically the resource allocationsubproblem determines 119875 and 119884 based on z produced by thelocation selection subproblemThen the payoff of the currentz is calculated Let 119881 denote the payoff of a given z

The general expression of the payoff function can bewritten as

119881 = sum119895

(119880119895 (119895) minus 119868119895 (119875119895)) (4)

where 119880119895(sdot) is an increasing function representing the utilityof device 119895 and 119868119895(sdot) is also an increasing function repre-senting the cost of device 119895 In this paper we firstly let119880119895(119895) = 119862119895119862119895 where 119862119895 is the minimum uplink datarate requirement of device 119895 Secondly since the locations ofdevices in smart grid are fixed and the power can be suppliedby alternating current adapter we just let 119868119895(119875119895) = 0 Thisis a difference between wireless communications for smartgrid and for land mobile users Therefore we define thesatisfaction ratio 119888119895 of device 119895 as the ratio between achieveddata rate and required data rate that is

119888119895 = 119862119895119862119895 (5)

and we then define 119881 as the sum of satisfaction ratio over alldevices that is

119881 = sum119895

119888119895 (6)

which is used to measure how good the given z isThe problem can be solved by solving these two subprob-

lems in an iterative fashion The value of 119881 for the currentz will be fed back to the location selection subproblem forguided search of the better z The next two sections will solvethese two subproblems in sequence

4 Resource Allocation Methods

The task of resource allocation is to determine 119875 and 119884 givenz Two different methods based on different principles arepresented The first is scheduling based for which uplinkswhich are far away from each other are scheduled to sharethe same RB The second is power control based for whichthe transmission power of each uplink is controlled so thatuplinks which are not far away from each other can also sharethe same RB

41 Scheduling Based Resource Allocation This method con-sists of four steps which are described in sequence as follows

411 Uplink Transmission Power Setting This subsectiondetermines the transmission power 119875119895 for each device 119895 Asstated before for this method uplinks which are far awayfrom each other (ie do not interfere with each other) will bescheduled to share the same RBTherefore for the schedulingbased method it can be expected that the interference power119875I in (1) is negligible That is we assume that there is nointerference between distant devices Thus given the RBallocated to device 119895 the received SINR experienced by device119895 on this RB at BS 119887 can be approximately written as

120574119895 asymp 119875119895119866119895119887119875N ge Γ (7)

where device 119895 is served by BS 119887 (ie 119895 isin 119878119887) and Γ isthe minimum SINR requirement Γ is a system parameterand common to all devices and RBs Therefore the uplinktransmission power 119875119895 can be set to

119875119895 = min(119875N sdot Γ119866119895119887 119875max) (8)

That is since in this method distant devices between whichthere is no interference are scheduled simultaneously there isno power control and power is strictly a function of the targetminimum SINR requirement

412 Interference Graph Construction The interferencegraph is used to indicate whether any two devices canreuse the same RB due to the interference between themAs indicated by Property 4 different types of devices shalltransmit data over different channels So we need to constructinterference graph for each type respectively Let G119894(119881119894 119864119894)denote the interference graph for the 119894th type 1 le 119894 le 119870where 119881119894 is the vertex set in which each vertex represents adevice of the 119894th type and 119864119894 is the edge set in which eachedge 119890119895119896 represents devices 119895 and 119896 which cannot reuse thesame RB There are two rules to decide if edge 119890119895119896 existsAssume that devices 119895 and 119896 are served by BS 119887119895 and 119887119896respectively The first rule is if 119887119895 = 119887119896 then edge 119890119895119896 existsThe second rule is if 119887119895 = 119887119896 but the interference causedto each other is too large then edge 119890119895119896 exists Specificallyif the distance between device 119895 and BS 119887119896 is less than theinterference radius 119877119895 of device 119895 or if the distance betweendevice 119896 and BS 119887119895 is less than the interference radius 119877119896 ofdevice 119896 then edge 119890119895119896 exists

The calculation of interference radius is as follows Fordevice 119895 the interference radius 119877119895 is defined as the distanceat which the received SINR is 120578 where 120578 is the SINR require-ment to ensure that the device does not cause nonnegligibleinterference to other uplinks that are out of the range ofinterference radius According to (7) we have the equationfor 119877119895 as

119875119895 sdot PL (119877119895)119875N = 120578 (9)

fromwhich the value of119877119895 can be solved After calculating theinterference radius for each device the interference graphG119894


Require zEnsure G119894 1 le 119894 le 119870(1) for 119894 = 1 to119870 do(2) for any two devices 119895 and 119896 inH119894 do(3) if 119887119895 = 119887119896 then(4) connect vertexes 119895 and 119896 inG119894(5) end if(6) if dis(119895 119887119896) lt 119877119895 or dis(119896 119887119895) lt 119877119896 then(7) connect vertexes 119895 and 119896 inG119894(8) end if(9) end for(10) end forAlgorithm 1 Interference graph construction

can be constructedThe procedure is outlined in Algorithm 1where dis(119895 119887) in line (6) represents the distance betweendevice 119895 and BS 119887413 Utility Function Calculation For each 119894 1 le 119894 le 119870 theutility function 119865119894119899 is defined as the sum of satisfaction ratioover all devices of the 119894th type given that a total of 119899 channelshave been allocated to them To calculate 119865119894119899 define Δ119865119894119899 asthe sum of satisfaction ratio over all devices of the 119894th typegiven that the 119899th channel has been allocated to them Thenthe value of 119865119894119899 can be obtained according to

119865119894119899 = 119865119894119899minus1 + Δ119865119894119899 (10)

where 1198651198940 = 0 Further to calculate Δ119865119894119899 define Δ119865119894119899119897 as thesum of satisfaction ratio over all devices of the 119894th type giventhat the 119897th RB of the 119899th channel has been allocated to themThen the value of Δ119865119894119899 can be obtained according to

Δ119865119894119899 =119871sum119897=1

Δ119865119894119899119897 (11)

Let H119894119899119897 denote the set of devices of the 119894th type that sharethe 119897th RB of the 119899th channel Then the value of Δ119865119894119899119897 can becalculated as

Δ119865119894119899119897 = sum119895isinH119894119899119897

119899119897119895119862119895 (12)

where 119862119899119897119895 can be obtained by (2) Finally we say a deviceis feasible in slot 119897 if the total power allocated to this devicein this slot does not exceed 119875max The procedure to calculateutility function is outlined inAlgorithm 2where the setH119894119899119897is determined in a heuristic manner in lines (4)ndash(10)414 RB Allocation This subsection presents the RB allo-cation algorithm As indicated by Property 4 due to thesecurity consideration different types of devices shall usedifferent frequency channels and different types sharing thesame frequency channel are not allowedThis is the constraintwhich channel allocation shall satisfy The procedure of the

scheduling based RB allocation is outlined in Algorithm 3where 119899119894 denotes the number of channels which have beenallocated to the 119894th type Specifically after the type which isallocated to the 119899th channel has been selected in line (3) theRBs of the 119899th channel shall be allocated according to H119894119899119897which has been obtained in Algorithm 2 as shown in line (6)42 Power Control Based Resource Allocation This methodconsists of four steps which are described in sequence asfollows

421 Grouping Let 119878119894119887 denote the set of type-119894 devices whichare served by BS 119887 1 le 119894 le 119870 The value of 119878119894119887 can be derivedfrom the value of 119878119887 which can be derived from the value ofz LetH119894 = H1198941 H119894119892 H119894119866119894 denote the grouping forthe 119894th type where H119894119892 is the set of devices of the 119894th typewhich can share the same RB and 119866119894 is the number of groupsThe procedure of grouping is outlined in Algorithm 4

422 Uplink Transmission Power Control Since all devicesin H119894119892 share the same RB the received SINR in (1) can berewritten as

120574119899119897119895 = 119875119895119866119895119887119875N + sum119896 =119895119896isinH119894119892 119875119896119866119896119887 ge Γ (13)

where 119895 isin H119894119892 and Γ is the minimum SINR requirementSimilarly Γ is a system parameter and common to all devicesand RBs

We propose an iterative update algorithm for finding theminimum transmission power satisfying the above equationSpecifically for the 119905th iteration the optimal power 119875[119905]119895 tobe used by device 119895 can be obtained by solving the followingequation

119875[119905]119895 119866119895119887119875N + sum119896 =119895119896isinH119894119892 119875[119905minus1]119896 119866119896119887 = Γ (14)

where 119875[119905minus1]119896

is the power settings obtained at iteration 119905 minus 1According to (14) the value of 119875[119905]119895 can be easily obtainedusing the bisection method [20] Additionally if the value of119875[119905]119895 is greater than 119875max it will be set as 119875max The update ofthe values of transmission power proceeds in iterations untilthe power convergence

423 Utility Function Definitions Theutility function 119865119894119899119892 isdefined as the sum of satisfaction ratio over all devices inH119894119892given that a total of 119899 channels have been allocated to them Tocalculate 119865119894119899119892 define Δ119865119894119899119892119897 as the sum of satisfaction ratioover all devices in H119894119892 given that the first 119897 RBs of the 119899thchannel have been allocated to them Then the value of 119865119894119899119892can be obtained according to

119865119894119899119892 = 119865119894119899minus1119892 + Δ119865119894119899119892119871 (15)


RequireG119894 1 le 119894 le 119870Ensure 119865119894119899 and H119894119899119897

(1) for 119894 = 1 to119870 do(2) for 119899 = 1 to119873 do(3) for 119897 = 1 to 119871 do(4) initializeH119894119899119897 = 0(5) delete fromG119894 devices which are not feasible in slot 119897 anymore(6) while G119894 = 0 do(7) determine device 119895lowast with the lowest satisfaction ratio inG119894(8) put 119895lowast intoH119894119899119897(9) delete 119895lowast and devices connected to it fromG119894(10) end while(11) calculate Δ119865119894119899119897(12) recoverG119894(13) update satisfaction ratio of all devices inG119894(14) end for(15) calculate Δ119865119894119899(16) calculate 119865119894119899(17) end for(18) end for

Algorithm 2 Utility function calculation

Require 119865119894119899 and H119894119899119897Ensure 119884119899119897119895

(1) initialize 119899119894 = 0 1 le 119894 le 119870(2) for 119899 = 1 to119873 do(3) allocate the 119899th channel to the 119894lowastth type with the minimum 119865119894119899119894 and break the tie arbitrarily(4) let 119899119894lowast = 119899119894lowast + 1(5) for 119897 = 1 to 119871 do(6) allocate the 119897th RB to devices inH119894lowast 119899119894lowast 119897

(7) let 119884119899119897119895 = 1 for each 119895 isin H119894lowast 119899119894lowast 119897

(8) end for(9) end for

Algorithm 3 Scheduling based RB allocation

Require zEnsure H119894119892

(1) for 119894 = 1 to119870 do(2) let 119866119894 = max |1198781198941| |1198781198942| |119878119894119861|(3) for 119892 = 1 to 119866119894 do(4) for 119887 = 1 to 119861 do(5) if 119878119894119887 = 0 then(6) select any device 119895 from 119878119894119887 and put intoH119894119892(7) delete device 119895 from 119878119894119887(8) end if(9) end for(10) end for(11) end for

Algorithm 4 Grouping


Require H119894119892Ensure 119884119899119897119895

(1) initialize 119899119894 = 0 1 le 119894 le 119870(2) for 119899 = 1 to119873 do(3) allocate the 119899th channel to the 119894lowastth type with the minimum 119865119894119899119894 and break the tie arbitrarily(4) let 119899119894lowast = 119899119894lowast + 1(5) for 119897 = 1 to 119871 do(6) selectH119894lowast119892lowast which is feasible in slot 119897 and has the minimum Δ119865119894lowast 119899119894lowast 119892119897minus1(7) let 119884119899119897119895 = 1 for each 119895 isin H119894lowast119892lowast (8) calculateΔ119865119894lowast 119899119894lowast 119892119897 for each 119892(9) end for(10) calculate 119865119894119899119894 for each 119894(11) end for

Algorithm 5 Power control based RB allocation

where 1198651198940119892 = 0 and the value of Δ119865119894119899119892119871 can be obtainedaccording to

Δ119865119894119899119892119897 =

Δ119865119894119899119892119897minus1 sum119895isinH119894119892

119884119899119897119895 = 0

Δ119865119894119899119892119897minus1 + sum119895isinH119894119892

119899119897119895119862119895 otherwise (16)

where 1198651198941198991198920 = 0 and 1 le 119897 le 119871424 RB Allocation This subsection presents the RB alloca-tion algorithm Similarly different types of devices are notallowed to share the same frequency channel which is theconstraint which channel allocation shall satisfy

For convenience we define function 119865119894119899 as119865119894119899 =

119866119894sum119892=1

119865119894119899119892 (17)

In addition we say a groupH119894119892 is feasible in slot 119897 if the totalpower allocated to each device 119895 isin H119894119892 in this slot does notexceed 119875max The procedure of the power control based RBallocation is outlined in Algorithm 5 where 119899119894 also denotesthe number of channels which have been allocated to the119894th type Specifically after the type which is allocated to the119899th channel has been selected in line (3) the RBs of the 119899thchannel shall be allocated according to H119894119892 which has beenobtained in Algorithm 4 as shown in line (6)5 Location Selection Methods

The task of location selection is to search for the location zThree different location selection methods are presentedThefirst is K-means based [21] This method is raw and is usedas the benchmark in this work The next two are simulatedannealing (SA) based [22] and particle swarm optimization(PSO) based [23] respectively

51 119870-Means Based Location Selection Initially z119887 = (1199111119887 1199112119887)is randomly selected from the candidate location set Θ as

the deployment locations of BSs where 1199111119887 and 1199112119887 are thehorizontal and vertical ordinate of the deployment locationrespectively Then we can obtain the corresponding Ω =1198781 1198782 119878119861 which describes the relationship between SGdevices andBSsNext the BS locations are updated as followsAssume that the locations of device 119895 are x119895 = (1199091119895 1199092119895)where 1199091119895 and 1199092119895 are the horizontal and vertical ordinate ofthe location of device 119895 respectively The new BS locationscan be calculated as

119911ℎ119887 = 110038161003816100381610038161198781198871003816100381610038161003816 sum119895isin119878119887

119909ℎ119895 (18)

where 1 le 119887 le 119861 ℎ isin 1 2 and |119878119887| is the number of devicesserved by the 119887th BS For each 119887 if the calculated z119887 does notbelong to Θ it shall be set as the element in Θ which is theclosest to the calculated value

52 SA Based Location Selection The location selection is toiterate over all candidate locations to find the best locationthat maximizes the satisfaction ratio Since the enumerationis practically impossible an algorithm with controllablecomplexity which can output a solution within the giventime limit is desirable We consider a stochastic local searchalgorithm which progressively traverses from one locationto its neighbor in a probabilistic manner for finding theglobal optimal solution Specifically an algorithm based onsimulated annealing is proposed as outlined in Algorithm 6

Beginning with an initial location the variable zbestrecords the location with the highest payoff obtained so faras the algorithm proceeds In lines (4) and (9) the resourceallocation methods in Section 4 are used to determine thevalues of 119875 and 119884 At each iteration a new location znextamong the neighborhood of current location z is chosen inline (8)The new location znext is determined as follows Firstfor the current z we can obtain Ω = 1198781 1198782 119878119861 and thencalculate the satisfactory ratio of each 119878119887 1 le 119887 le 119861 Foreach iteration only one BS location is changed We choose BS119887lowast with the lowest satisfactory ratio to change the locationSpecifically we select a candidate BS location from Θ whichis no more than 119889meters away from the original BS location


(1) initialize 119888 = 0(2) initialize 119905 = 119905init(3) initialize z(4) determine the values of 119875 and 119884 given z(5) determine the value of 119881 given z 119875 and 119884(6) initialize zbest = z and 119881best = 119881(7) while 119888 lt 119888max do(8) update znext(9) update 119875next and 119884next given znext(10) update119881next given znext 119875next and 119884next(11) if 119881next gt 119881 then(12) update z = znext and 119881 = 119881next(13) if 119881next gt 119881best then(14) update zbest = znext and 119881best = 119881next(15) end if(16) else(17) update z = znext and 119881 = 119881next with probability 119890(119881nextminus119881)119905(18) end if(19) let 119888 = 119888 + 1(20) let 119905 = 120572119905(21) end while(22) return zbest

Algorithm 6 SA based iterative procedure

as the new BS location where 119889 is a parameter If znext yieldsa better payoff than z the search proceeds with znext for thenext iteration Otherwise znext is still chosen with probability119890(119881nextminus119881)119905 based on the concept of simulated annealing inline (17) In line (20) the temperature 119905 decreases after eachiteration according to an annealing schedule 119905 = 120572119905 where0 lt 120572 lt 1 is also a parameter Different values of 119888max 120572 and119889 can be set to control the speed of cooling

53 PSO Based Location Selection In this subsection aparticle swarm optimization based algorithm is presentedto search for the location Assume that the swarm consistsof 119872 particles and the search space is 119861 dimensional LetZ119898 = (z1198981 z119898119887 z119898119861) represent the position ofthe 119898th particle where z119898119887 is a two-dimensional vectorrepresenting the deployment location of the 119887th BS Letk119898 = (k1198981 k119898119887 k119898119861) represent the velocity of the119898thparticle where k119898119887 = (V1119898119887 V2119898119887) is a two-dimensional vectorfor which V1119898119887 and V2119898119887 represent the horizontal and verticalvelocity respectively Let P119898 = (p1198981 p119898119887 p119898119861)represent the position of the best solution found by the119898th particle and let Plowast = (plowast1 plowast119887 plowast119861) represent theposition of the best solution found by all particles duringthe search The position of each particle is updated by usingZ[119905+1]119898 = Z[119905]119898 + k[119905+1]119898 where Z[119905]119898 is the position of the 119898thparticle at iteration 119905 and k[119905+1]119898 is the new velocity of the119898th particle at iteration 119905 + 1 The velocities of the particlesare updated according to k[119905+1]119898 = 119908k[119905]119898 + 1198881120585(P[119905]119898 minus Z[119905]119898 ) +1198882120578(Plowast[119905] minusZ[119905]119898 ) where P[119905]119898 is the position of the best solutionfound by the119898th particle at iteration 119905 Plowast[119905] is the position ofthe best solution found by all particles during the search so

far and 120585 and 120578 are random values generated by the uniformdistribution in the interval [0 1]

Additionally for the PSO based algorithm there are twotypes of collisions For the first type the particles could beattracted to regions outside the feasible search space Θ forthe second type the velocity of particles could be too largeThe anticollision mechanisms for preserving the feasibility ofsolution are as follows For the first type of collision if z119898119887 notinΘ occurs we set z119898119887 randomly selected location inΘ For thesecond type of collision if it occurs we set

Vℎ119898119887 = Vmax if Vℎ119898119887 gt Vmax

minusVmax if Vℎ119898119887 lt minusVmax (19)

where ℎ isin 1 2 and Vmax is the velocity limitThe procedure for PSO based algorithm is outlined in

Algorithm 7 where 119888max is the iteration limit

6 Performance Evaluation

61 Parameter Setting Assume there are a total of 119870 = 3types of SG devices In the case of no particular descriptionthe required uplink data rate of each type is 1198621 = 100 kbps1198622 = 400 kbps and 1198623 = 800 kbps respectively and thenumber of devices of each type is 50 50 and 50 respectivelyWe randomly distribute these devices in a circle regionΨ witha radius of 1200 meters Further we assume that Θ contains atotal of 350 candidate BS locations which are also randomlygenerated in Ψ Based on the simulation settings in [24 25]wireless communication related parameters are set as followsThe maximum transmission power 119875max is 20 dBm The pathloss formula is PL(119909) = 6 + 4268 log(119909) dB for a distance


(1) initialize 119888 = 1(2) for particle119898 = 1 to119872 do(3) initialize the velocity k119898 in [minusVmax Vmax](4) initialize the position Z119898 in Θ(5) determine the value of 119881119898 given Z119898(6) initialize P119898 = Z119898 and 119881best119898 = 119881119898(7) end for(8) calculate119881best = max 119881best1 119881best2 119881best119872 and determine Plowast(9) while 119888 lt 119888max do(10) for particle119898 = 1 to119872 do(11) update the velocity k119898(12) update the position Z119898(13) determine the value of 119881119898 given Z119898(14) if 119881119898 gt 119881best119898 then(15) update 119881best119898 = 119881119898(16) update P119898(17) end if(18) end for(19) calculate1198811015840best = max 119881best1 119881best2 119881best119872(20) if 1198811015840best gt 119881best then(21) update119881best = 1198811015840best(22) update Plowast(23) end if(24) let 119888 = 119888 + 1(25) end while(26) return Plowast

Algorithm 7 PSO based iterative procedure

separation of 119909meters The total bandwidth 119882 is 5MHz andthe bandwidth of each channel 1198820 is 180 kHz Assume thatthe power of background noise 119875N = 1198730119882 where the noisepower spectrum density 1198730 = minus174 dBmHz The minimumSINR requirement Γ is 3 dB which is used in (7) and (13) todetermine transmit powerThe SINR requirement 120578 is minus2 dBwhich is used in (9) to determine interference radius Finallythe number of slots in each frame 1198710 is 20 In the case ofno particular description assume that the number of usableslots 119871 is also 20 For SA there are three parameters 119905init120572 and 119889 For 119905init and 119889 based on the recommendations in[20 26 27] we set 119905init = 1000 and 119889 = 30 For 120572 we have runmany simulation experiments to find an appropriate value ofit Simulation results show that the larger the value of 120572 isthe better the supporting ratio is Therefore since the valueof 120572 shall be between 0 and 1 we set 120572 = 099 For PSOthere are five parameters 119872 119881max 119908 1198881 and 1198882 For 119872 1199081198881 and 1198882 based on the recommendations in [27 28] we set119872 = 10 119908 = 07 1198881 = 2 and 1198882 = 2 For 119881max we have runmany simulation experiments to find an appropriate value ofit Simulation results show that the value of 119881max shall notbe too small or too large Specifically if the value of 119881maxis too small the convergence rate of PSO will be very slowif the value of 119881max is too large PSO will oscillate and notconverge Therefore after many simulation experiments wehave selected 119881max = 150 to achieve acceptable convergencerate Finally for both algorithms the iteration limit 119888max is setto be 1000

Combining different resource allocation and locationselection algorithms we have a total of six different schemes

We evaluate the performance of above schemes for differentparameter configurations For each parameter configurationwe run simulation experiments for 1000 times and averagethe results

62 Simulation Results This subsection presents the perfor-mance evaluation results of the proposed schemes under dif-ferent scenarios and the effects of various system parametersare evaluated and compared

621 Convergence We show in Figure 2 a typical trace ofthe progression of benefits for guided stochastic searchin all schemes where ldquoPCrdquo and ldquoSchedrdquo represent powercontrol and scheduling based resource allocation algorithmrespectively We can find that the payoff of the best locationselection is increased gradually and will be converged to aconstant value finally Therefore the curves in Figure 2 showthat the proposed schemes are converged to a steady stateAdditionally we can observe that the solution quality and therequired number of iterations to converge are significantlydifferent from each other Firstly the final values of payofffor different schemes are different Specifically the ldquoPC +PSOrdquo scheme can achieve the highest payoff (ie 1285030)among all schemes Recall that the payoff is defined as thesumof satisfaction ratio over all devices where the satisfactionratio of a device is defined as the ratio between achieved datarate and the required data rate For this set of simulationexperiments since there are totally 150 devices (as stated inthe beginning of Section 61) the value of payoff will not behigher than 150 Therefore a payoff of 1285030 means that


PC + K-meansPC + SAPC + PSO

Sched + K-meansSched + SASched + PSO

80

85

90

95

100

105

110

115

120

125

130

Payo

ff

10 20 30 40 50 60 70 801Number of iterations

Figure 2 Convergence of the proposed schemes

Supp

ortin

g ra

tio



0303504

04505

05506

06507

07508

08509

0951

4 5 6 7 8 93Channel bandwidth (MHz)

Figure 3 Impact of the number of channels

most data rate requirements have been satisfied Secondlyfor K-means related schemes (ie the ldquoPC + K-meansrdquo andldquoSched + K-meansrdquo schemes) although their payoff is nothigh (ie 1048572 and 1001876) the required numbers ofiterations to converge (ie 2 and 2) are much smaller thanother schemes that is they converge much faster than otherschemes Therefore we can conclude that different schemescan achieve different tradeoffs between solution quality andconvergence rate

For any device 119895 if its uplink data rate requirement is met(ie 119862119895 ge 119862119895) we say this device is satisfied Further wedefine the supporting ratio as the ratio between the number

Supp

ortin

g ra

tio

0404505

05506

06507

07508

08509

0951

9 10 11 12 13 14 158Number of BSs



Figure 4 Impact of the number of BSs

70 80 90 10060Number of devices of the second type

Supp

ortin

g ra

tio



0303504

04505

05506

06507

07508

08509

0951

Figure 5 Impact of the number of devices

of devices which have been satisfied and the total numberof devices In the following simulation experiments we willevaluate the impact of the number of channels (ie the totalbandwidth) the number of BSs and the number of deviceson the performance (ie the supporting ratio) of all thesesix schemes Additionally we would like to claim that all thevalues plotted in Figures 3 4 and 5 are obtained after thealgorithms have converged to a steady state

622 Impact of the Number of Channels The number ofchannels is equal to lfloor1198821198820rfloor where119882 is the total bandwidthFigure 3 shows the supporting ratio of all proposed schemes


when the total bandwidth 119882 or equivalently the number ofchannels is varied For this set of simulation experimentsthere are totally 150 devices for which the sum of data raterequirements is 50times1198621+50times1198622 +50times1198623 = 65MbpsWe setthe number of BSs 119861 to be 10 It can be observed that whenthe total bandwidth (ie the number of channels) increasesthe supporting ratio increases Specifically when the totalbandwidth is 9MHz (ie the number of channels is 50) thesupporting ratio of the ldquoPC+ PSOrdquo and ldquoPC+ SArdquo schemes isas high as 1 (ie the data rate requirements of all 150 deviceshave been satisfied) but the supporting ratio of the ldquoPC + K-meansrdquo and ldquoSched + K-meansrdquo schemes is just 08713 and08605 respectively (ie there are still 150 times (1 minus 08713) =20 and 150 times (1 minus 08605) = 21 devices whose data raterequirements are not satisfied resp)

Finally it can be observed that the ldquoPC + PSOrdquo schemeis the best among all other schemes which will be val-idated again by the following simulation results This isdue to two aspects of reasons For the first reason SAand PSO are metaheuristics which efficiently explore thesearch space to find near-optimal solutions By searchingover a large set of feasible solutions they can find goodsolutions with less computational effort compared to simpleheuristics (eg the K-means method) Therefore SA andPSO are superior to K-means in finding good solutions Forthe second reason if two devices are close to each otherthey could interfere with each other if they use the sameRB For the PC method the transmission power of eachdevice is controlled so that devices which are close to eachother can also share the same RB for the Sched methodonly devices which are far away from each other can sharethe same RB Since the PC method allows devices whichare close to each other to transmit data simultaneously itcan admit more devices than the Sched method On theother hand for the PC method since there exists inter-ference among neighbor devices each device will have toincrease its transmission power to combat such interferencetomeet the minimum SINR requirement This makes devicesusing the PC method consume more power resource thanthe Sched method Therefore the PC method can admitmore devices than the Sched method via consuming morepower

623 Impact of the Number of BSs Figure 4 shows thesupporting ratio of all proposed schemeswhen the number ofBSs 119861 is varied We can observe that when the number of BSsincreases the supporting ratio increases since the averagedistance between devices and access points is shortenedSpecifically when the number of BSs is 15 the supportingratio of the ldquoPC + PSOrdquo and ldquoPC + SArdquo schemes is 1 and09767 respectively (ie there are zero and 150 times (1 minus09767) = 4 devices whose data rate requirements are notsatisfied resp) but the supporting ratio of the ldquoPC + K-meansrdquo and ldquoSched + K-meansrdquo schemes is just 08447 and07743 respectively (ie there are still 150 times (1 minus 08447) =24 and 150 times (1 minus 07743) = 34 devices whose datarate requirements are not satisfied resp) Therefore we canconclude that the ldquoPC + PSOrdquo scheme is the best one and forthe simulated scenario at least 15 BSs shall be deployed so that

the supporting ratio of one can be achieved For the followingsimulations we will set the value of 119861 to be 15

624 Impact of the Number of Devices Figure 5 shows thesupporting ratio of all proposed schemes when the number ofdevices is varied For convenience let 119873119894 denote the numberof devices of the 119894th type Let119873119894init denote the initial value of119873119894 As stated in the beginning of Section 61 we set 1198731init =50 1198732init = 50 and 1198733init = 50 We will collect theperformance metrics (ie the supporting ratio) which is afunction of (119873111987321198733) However it is hard to visualize high-dimensional data when the dimension is greater than twoTherefore we run the simulation for three times For the 119895th(119895 = 1 2 3) run we change the values of 119873119895 = 119873119895init + 119899while keeping the values of other 119873119894 (119894 = 119895) fixed to be119873119894initwhere 119899 = 10 20 30 40 50 Due to the limited space weonly plot the simulation results of the second run in Figure 5where the horizontal axis represents the number of devicesof the second type We can observe that when the numberof devices increases the supporting ratio decreases since theradio resource consumed by each type of devices increasesSpecifically when the number of devices of the second typeis increased to be 100 there are totally 50 + 100 + 50 =200 devices for which the sum of data rate requirementsis 50 times 1198621 + 100 times 1198622 + 50 times 1198623 = 85Mbps For thisscenario the supporting ratio of the ldquoPC + PSOrdquo and ldquoPC+ SArdquo schemes is still 08626 and 08420 respectively (iethere are 150 times 08626 = 129 and 150 times 08420 = 126 deviceswhose data rate requirements can be satisfied resp) but thesupporting ratio of the ldquoPC + K-meansrdquo and ldquoSched + K-meansrdquo schemes is only 06600 and 06015 respectively (iethere are only 150 times 06600 = 99 and 150 times 06015 = 90devices whose data rate requirements have been satisfiedresp) Comparing these curves we can also conclude that theldquoPC + PSOrdquo scheme is more preferable than other schemes

7 Conclusions

In this paper we study the joint BS location selectiontransmission power control and wireless channel allocationproblem in OFDMA based private wireless access networksfor smart grid We transform the joint problem into channelallocation and site selection subproblems and solve these twosubproblems iteratively According to the simulation resultsthe combination of power control based resource allocationalgorithm and PSO based location selection algorithm isrecommended to solve the joint problem

List of Symbols

Ψ The area in which a set of SG devices isscattered

H The set of devices119862119895 The minimum uplink data raterequirement of device 119895119875119895 The uplink transmission power of device 119895119875max The maximum uplink transmission power119870 The number of types of devices

H119894 The set of SG devices of the 119894th type


119861 The number of BSsz119887 The deployment location of the 119887th BSΘ The set of candidate BS locations119878119887 The set of devices served by the 119887th BS119882 The total bandwidth in Hertz119873 The number of channels into which the

total bandwidth is divided1198820 The channel bandwidth in Hertz1198710 The number of slots in a frame119871 The number of slots which can be used foruplink communications in each frame119884119899119897119895 The binary variable indicating whether the119897th RB of the 119899th channel is allocated todevice 119895120574119899119897119895 The received SINR experienced by device119895 on the RB (119899 119897) at BS 119887119866119887119895 The path loss from device 119895 to BS 119887119875N The power of background noise119875I The power of interference

D119899119897 The set of devices which share the sameRB with device 119895

PL(119909) The path loss for a distance separation of 119909meters119862119899119897119895 The uplink data rate achieved by device 119895on RB (119899 119897)119862119895 The total data rate achieved by device 119895119862 The set of all 119862119895119875 The set of all 119875119895

z The set of all z119887Ω The set of all 119878119887119884 The set of all 119884119899119897119895119862 The set of all 119895119888119895 The satisfaction ratio of device 119895119881 The sum of satisfaction ratio over alldevicesΓ The minimum SINR requirement

G119894 The interference graph for the 119894th type119881119894 The vertex set inG119894119864119894 The edge set in G119894119890119895119896 The edge which represents devices 119895 and 119896cannot reuse the same RB119877119895 The interference radius of device 119895120578 The SINR requirement to calculate 119877119895

dis(119895 119887) The distance between device 119895 and BS 119887119865119894119899 The sum of satisfaction ratio over alldevices of the 119894th type given that a total of119899 channels have been allocated to themΔ119865119894119899 The sum of satisfaction ratio over alldevices of the 119894th type given that the 119899thchannel has been allocated to themΔ119865119894119899119897 The sum of satisfaction ratio over alldevices of the 119894th type given that the 119897thRB of the 119899th channel has been allocatedto them

H119894119899119897 The set of devices of the 119894th type that sharethe 119897th RB of the 119899th channel119899119894 The number of channels which have beenallocated to the 119894th type

119878119894119887 The set of type-119894 devices which are servedby BS 119887

H119894119892 The set of devices of the 119894th type which canshare the same RB119866119894 The number of groups

H119894 The set of allH119894119892119875[119905]119895 The power setting obtained at iteration 119905119865119894119899119892 The sum of satisfaction ratio over all

devices inH119894119892 given that a total of 119899channels have been allocated to themΔ119865119894119899119892119897 The sum of satisfaction ratio over alldevices inH119894119892 given that the first 119897 RBs ofthe 119899th channel have been allocated tothem119865119894119899 The sum of 119865119894119899119892 over all groups

x119895 The locations of device 119895Competing Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This work was supported by the National Natural ScienceFoundation of China (no 61571111)

References

[1] R Ma H-H Chen Y-R Huang and W Meng ldquoSmartgrid communication its challenges and opportunitiesrdquo IEEETransactions on Smart Grid vol 4 no 1 pp 36ndash46 2013

[2] E Dahlman S Parkvall and J Skold 4G LTELTE-Advancedfor Mobile Broadband Academic Press New York NY USA2013

[3] X S Shen ldquoEmpowering the smart grid with wireless technolo-giesrdquo IEEE Network vol 26 no 3 pp 2ndash3 2012

[4] H Gharavi and B Hu ldquoMultigate communication network forsmart gridrdquoProceedings of the IEEE vol 99 no 6 pp 1028ndash10452011

[5] C Gentile D Griffith and M Souryal ldquoWireless networkdeployment in the smart grid design and evaluation issuesrdquoIEEE Network vol 26 no 6 pp 48ndash53 2012

[6] Q-D Ho Y Gao and T Le-Ngoc ldquoChallenges and researchopportunities in wireless communication networks for smartgridrdquo IEEE Wireless Communications vol 20 no 3 pp 89ndash952013

[7] Z Zhu S Lambotharan W H Chin and Z Fan ldquoOverviewof demand management in smart grid and enabling wirelesscommunication technologiesrdquo IEEE Wireless Communicationsvol 19 no 3 pp 48ndash56 2012

[8] A Abdrabou and A M Gaouda ldquoUninterrupted wirelessdata transfer for smart grids in the presence of high powertransientsrdquo IEEE Systems Journal vol 9 no 2 pp 567ndash577 2015

[9] P-Y Kong ldquoWireless neighborhood area networks with QoSsupport for demand response in smart gridrdquo IEEE Transactionson Smart Grid vol 7 no 4 pp 1913ndash1923 2015

[10] W-Z Song D De S Tan S K Das and L Tong ldquoA wirelesssmart grid testbed in labrdquo IEEE Wireless Communications vol19 no 3 pp 58ndash64 2012


[11] B Fateh M Govindarasu and V Ajjarapu ldquoWireless networkdesign for transmission line monitoring in smart gridrdquo IEEETransactions on Smart Grid vol 4 no 2 pp 1076ndash1086 2013

[12] H Gharavi and B Hu ldquoScalable synchrophasors commu-nication network design and implementation for real-timedistributed generation gridrdquo IEEE Transactions on Smart Gridvol 6 no 5 pp 2539ndash2550 2015

[13] M M Aly and M A El-Sayed ldquoEnhanced fault locationalgorithm for smart grid containing wind farm using wirelesscommunication facilitiesrdquo IET Generation Transmission ampDistribution vol 10 no 9 pp 2231ndash2239 2016

[14] X Wang and P Yi ldquoSecurity framework for wireless communi-cations in smart distribution gridrdquo IEEE Transactions on SmartGrid vol 2 no 4 pp 809ndash818 2011

[15] T Liu Y Liu Y Mao et al ldquoA dynamic secret-based encryptionscheme for smart grid wireless communicationrdquo IEEE Transac-tions on Smart Grid vol 5 no 3 pp 1175ndash1182 2014

[16] B Hu andH Gharavi ldquoSmart gridmesh network security usingdynamic key distribution withmerkle tree 4-way handshakingrdquoIEEETransactions on Smart Grid vol 5 no 2 pp 550ndash558 2014

[17] F Salvadori C S Gehrke A C de Oliveira M de Campos andP S Sausen ldquoSmart grid infrastructure using a hybrid networkarchitecturerdquo IEEE Transactions on Smart Grid vol 4 no 3 pp1630ndash1639 2013

[18] S Chen ldquoA novel TD-LTE frame structure for heavy uplinktraffic in smart gridrdquo in Proceedings of the 2014 IEEE InnovativeSmart Grid Technologies-Asia (ISGTAsia rsquo14) pp 158ndash163 KualaLumpur Malaysia May 2014

[19] Jiangsu Electric Power Company of China ldquoElectric powerbroadbandwirelessmulti-service bearer networkrdquoWhite Paper2015

[20] H-Y Hsieh S-E Wei and C-P Chien ldquoOptimizing small celldeployment in arbitrary wireless networks with minimum ser-vice rate constraintsrdquo IEEE Transactions on Mobile Computingvol 13 no 8 pp 1801ndash1815 2014

[21] J Macqueen ldquoOn convergence of K-means and partitions withminimum average variancerdquo Annals of Mathematical Statisticsvol 36 no 3 pp 1084ndash1090 1965

[22] H Keinanen ldquoSimulated annealing for multi-agent coalitionformationrdquo in Agent and Multi-Agent Systems Technologies andApplications Lecture Notes in Computer Science pp 30ndash39Springer Berlin Germany 2009

[23] J Kennedy and R Eberhart ldquoParticle swarm optimizationrdquoin Proceedings of the IEEE International Conference on NeuralNetworks pp 1942ndash1948 Perth Australia November 1995

[24] 3GPP ldquoFurther advancements for E-UTRA physical layeraspectsrdquo 3GPP TR 36814 v900 2010

[25] 3GPP ldquoLTE coverage enhancementsrdquo 3GPP TR 36824 v11002012

[26] Y Lu Y Lin Q Peng and Y Wang ldquoA review of improvementand research on parameters of simulated annealing algorithmrdquoCollege Mathematics vol 31 no 6 pp 96ndash103 2015

[27] D Wang Intelligent Optimization Methods Higher EducationPress Beijing China 2007

[28] A I S Nascimento and C J A Bastos-Filho ldquoA particleswarmoptimization based approach for themaximumcoverageproblem in cellular base stations positioningrdquo in Proceedings ofthe 10th International Conference on Hybrid Intelligent Systems(HIS rsquo10) pp 91ndash96 IEEE Atlanta Ga USA August 2010

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



transport medium Hence security must be considered atevery layer of the protocol stack in the private wirelessaccess networks for SG In addition to the authenticationauthorization and encryption considered at the applicationlayer the frequency separation mechanism at the physicallayer shall also be considered which is explained as followsActually data produced by different types of devices in smartgrid shall be transmitted to different destination systems Forexample as shown in Figure 1 the data produced by dataterminal unit (DTU) shall be transmitted to the productionservice system while the data produced by video terminalunit (VTU) shall be transmitted to the management servicesystem Due to the security purpose the transmission pathsused by different types of data shall be separated as muchas possible The separation can be achieved physically orlogically For example four different approaches to constructthe access network are illustrated in Figure 1 where thedata paths in Figure 1(a) are the most separated that is theseparation is achieved physically while the data paths inFigure 1(d) are the least separated that is the separation canbe achieved logically Further in addition to the separationfor the wireline segment data transmission over the wirelesssegment of data path shall also be separated for differenttypes of devices This requires that different types of devicesshall use different frequency channels to transmit their datathat is different types of devices sharing the same frequencychannel will not be allowed This is the frequency separationrequirement considered in this work

As a first step towards addressing the above issuesthis paper investigates the problem of how to deploy BSsand allocate wireless resources so that the uplink trans-mission requirements are efficiently met For this problemwe propose to decompose it into the resource allocationsubproblem and the location selection subproblem and solvethese two subproblems in an iterative fashionThe remainderof the paper is organized as follows Section 2 formulatesthe joint resource allocation and location selection problemSection 3 presents the overall framework to address thisproblem Sections 4 and 5 propose resource allocation andlocation selection algorithms respectively Simulation resultsare reported in Section 6 Finally we conclude in Section 7

2 Problem Formulation

Consider a set of SG devices scattered in an area Ψ Let Hdenote the set of devices For each device 119895 isin H let119862119895 denotethe minimum uplink data rate requirement and 119875119895 the uplinktransmission power The value of 119875119895 shall satisfy 0 le 119875119895 le119875max where119875max is the upper bound For convenience let119862 =119862119895 and 119875 = 119875119895 respectively Assume that all devices areclassified into119870 different types LetH119894 denote the set of type-119894 SG devices 1 le 119894 le 119870

Assume that the private wireless access network consistsof 119861 BSs which are located in the area Ψ Let z119887 = (1199111119887 1199112119887)denote the deployment location of the 119887th BS 1 le 119887 le 119861where 1199111119887 and 1199112119887 are the horizontal and vertical ordinate of thedeployment location respectively For convenience let z =z119887 Not every location inΨ can be the candidate location for

BS Assume thatΘ denotes the candidate BS location set inΨand the deployment location of BS can only be selected fromthe elements of Θ That is we restrict z119887 isin Θ In additionlet Ω = 1198781 1198782 119878119861 denote the relationship between SGdevices and BSs where 119878119887 is the set of devices served bythe 119887th BS For simplicity and without loss of generality weassume that the value of 119878119887 is determined by the distancebased rule That is a device will be served by the BS whichis the closest to it

Consider a OFDMA based private wireless access net-work The radio resource is defined as follows In thefrequency domain assume that the total bandwidth is dividedinto 119873 channels Let 119882 and 1198820 denote total and channelbandwidth in Hertz respectively In the time domain assumethat the time axes are organized into consecutive slots and1198710 consecutive slots constitute a frame The basic resourceunit for data transmission is a resource block (RB) which isdefined as one channel in the frequency domain and one slotin the time domain respectively In each frame assume that119871 slots can be used for uplink communications Therefore foreach channel there are 119871 RBs which are allocatable Finallydefine binary variable 119884119899119897119895 to denote the results of radioresource allocation which is valued 1 if the 119897th RB of the 119899thchannel is allocated to device 119895 and 0 otherwise Each deviceshall be allocated a number of RBs to meet its minimum datarate requirement For convenience let 119884 = 119884119899119897119895

Given that the RB (119899 119897) has been allocated to device 119895 thereceived signal-interference-noise-ratio (SINR) experiencedby device 119895 on this RB at BS 119887 can be written as

120574119899119897119895 = 119875119895119866119895119887119875N + 119875I (1)

where 119866119887119895 is the path loss from device 119895 to BS 119887 119875N isthe power of background noise 119875I = sum119896 =119895119896isinD119899119897 119875119896119866119896119887 isthe power of interference and D119899119897 is the set of deviceswhich share the same RB with device 119895 For simplicity andwithout loss of generality we assume that the path lossmainlydepends on the distance and can be calculated accordingto the formula PL(119909) for a distance separation of 119909 metersand we assume that there is no interference between distantdevices Let 119862119899119897119895 denote the uplink data rate achieved bydevice 119895 on RB (119899 119897) which is calculated by the Shannonformula as

119899119897119895 = 1198820log (1 + 120574119899119897119895)1198710 (2)

Then the total data rate achieved by device 119895 denoted as 119895can be calculated as

119895 = sum(119899119897)119884119899119897119895=1

119899119897119895 (3)

For convenience let 119862 = 119895Finally the problem addressed in this paper can be

formulated as given the parameters Ψ H 119862 119870 H119894119875max 119861 Θ 119882 1198820 119873 1198710 and 119871 how to determine thevalues of deployment location z transmission power 119875 and





Baseband unit (BBU)





(a)




Baseband unit (BBU)





(b)




Baseband unit (BBU)





(c)




Baseband unit (BBU)





(d)




3 The Framework



119881 = sum119895

(119880119895 (119895) minus 119868119895 (119875119895)) (4)


119888119895 = 119862119895119862119895 (5)


119881 = sum119895

119888119895 (6)







120574119895 asymp 119875119895119866119895119887119875N ge Γ (7)


119875119895 = min(119875N sdot Γ119866119895119887 119875max) (8)




119875119895 sdot PL (119877119895)119875N = 120578 (9)





119865119894119899 = 119865119894119899minus1 + Δ119865119894119899 (10)


Δ119865119894119899 =119871sum119897=1

Δ119865119894119899119897 (11)


Δ119865119894119899119897 = sum119895isinH119894119899119897

119899119897119895119862119895 (12)





120574119899119897119895 = 119875119895119866119895119887119875N + sum119896 =119895119896isinH119894119892 119875119896119866119896119887 ge Γ (13)




where 119875[119905minus1]119896



119865119894119899119892 = 119865119894119899minus1119892 + Δ119865119894119899119892119871 (15)














Require H119894119892Ensure 119884119899119897119895




Δ119865119894119899119892119897 =

Δ119865119894119899119892119897minus1 sum119895isinH119894119892

119884119899119897119895 = 0

Δ119865119894119899119892119897minus1 + sum119895isinH119894119892

119899119897119895119862119895 otherwise (16)



119866119894sum119892=1

119865119894119899119892 (17)





119911ℎ119887 = 110038161003816100381610038161198781198871003816100381610038161003816 sum119895isin119878119887

119909ℎ119895 (18)




























80

85

90

95

100

105

110

115

120

125

130

Payo

ff



Supp

ortin

g ra

tio



0303504

04505

05506

06507

07508

08509

0951





Supp

ortin

g ra

tio

0404505

05506

06507

07508

08509

0951

9 10 11 12 13 14 158Number of BSs





Supp

ortin

g ra

tio



0303504

04505

05506

06507

07508

08509

0951










7 Conclusions


List of Symbols



















Acknowledgments


References
































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation













Baseband unit (BBU)





(a)




Baseband unit (BBU)





(b)




Baseband unit (BBU)





(c)




Baseband unit (BBU)





(d)




3 The Framework



119881 = sum119895

(119880119895 (119895) minus 119868119895 (119875119895)) (4)


119888119895 = 119862119895119862119895 (5)


119881 = sum119895

119888119895 (6)







120574119895 asymp 119875119895119866119895119887119875N ge Γ (7)


119875119895 = min(119875N sdot Γ119866119895119887 119875max) (8)




119875119895 sdot PL (119877119895)119875N = 120578 (9)





119865119894119899 = 119865119894119899minus1 + Δ119865119894119899 (10)


Δ119865119894119899 =119871sum119897=1

Δ119865119894119899119897 (11)


Δ119865119894119899119897 = sum119895isinH119894119899119897

119899119897119895119862119895 (12)





120574119899119897119895 = 119875119895119866119895119887119875N + sum119896 =119895119896isinH119894119892 119875119896119866119896119887 ge Γ (13)




where 119875[119905minus1]119896



119865119894119899119892 = 119865119894119899minus1119892 + Δ119865119894119899119892119871 (15)














Require H119894119892Ensure 119884119899119897119895




Δ119865119894119899119892119897 =

Δ119865119894119899119892119897minus1 sum119895isinH119894119892

119884119899119897119895 = 0

Δ119865119894119899119892119897minus1 + sum119895isinH119894119892

119899119897119895119862119895 otherwise (16)



119866119894sum119892=1

119865119894119899119892 (17)





119911ℎ119887 = 110038161003816100381610038161198781198871003816100381610038161003816 sum119895isin119878119887

119909ℎ119895 (18)




























80

85

90

95

100

105

110

115

120

125

130

Payo

ff



Supp

ortin

g ra

tio



0303504

04505

05506

06507

07508

08509

0951





Supp

ortin

g ra

tio

0404505

05506

06507

07508

08509

0951

9 10 11 12 13 14 158Number of BSs





Supp

ortin

g ra

tio



0303504

04505

05506

06507

07508

08509

0951










7 Conclusions


List of Symbols



















Acknowledgments


References
































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











3 The Framework



119881 = sum119895

(119880119895 (119895) minus 119868119895 (119875119895)) (4)


119888119895 = 119862119895119862119895 (5)


119881 = sum119895

119888119895 (6)







120574119895 asymp 119875119895119866119895119887119875N ge Γ (7)


119875119895 = min(119875N sdot Γ119866119895119887 119875max) (8)




119875119895 sdot PL (119877119895)119875N = 120578 (9)





119865119894119899 = 119865119894119899minus1 + Δ119865119894119899 (10)


Δ119865119894119899 =119871sum119897=1

Δ119865119894119899119897 (11)


Δ119865119894119899119897 = sum119895isinH119894119899119897

119899119897119895119862119895 (12)





120574119899119897119895 = 119875119895119866119895119887119875N + sum119896 =119895119896isinH119894119892 119875119896119866119896119887 ge Γ (13)




where 119875[119905minus1]119896



119865119894119899119892 = 119865119894119899minus1119892 + Δ119865119894119899119892119871 (15)














Require H119894119892Ensure 119884119899119897119895




Δ119865119894119899119892119897 =

Δ119865119894119899119892119897minus1 sum119895isinH119894119892

119884119899119897119895 = 0

Δ119865119894119899119892119897minus1 + sum119895isinH119894119892

119899119897119895119862119895 otherwise (16)



119866119894sum119892=1

119865119894119899119892 (17)





119911ℎ119887 = 110038161003816100381610038161198781198871003816100381610038161003816 sum119895isin119878119887

119909ℎ119895 (18)




























80

85

90

95

100

105

110

115

120

125

130

Payo

ff



Supp

ortin

g ra

tio



0303504

04505

05506

06507

07508

08509

0951





Supp

ortin

g ra

tio

0404505

05506

06507

07508

08509

0951

9 10 11 12 13 14 158Number of BSs





Supp

ortin

g ra

tio



0303504

04505

05506

06507

07508

08509

0951










7 Conclusions


List of Symbols



















Acknowledgments


References
































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












119865119894119899 = 119865119894119899minus1 + Δ119865119894119899 (10)


Δ119865119894119899 =119871sum119897=1

Δ119865119894119899119897 (11)


Δ119865119894119899119897 = sum119895isinH119894119899119897

119899119897119895119862119895 (12)





120574119899119897119895 = 119875119895119866119895119887119875N + sum119896 =119895119896isinH119894119892 119875119896119866119896119887 ge Γ (13)




where 119875[119905minus1]119896



119865119894119899119892 = 119865119894119899minus1119892 + Δ119865119894119899119892119871 (15)














Require H119894119892Ensure 119884119899119897119895




Δ119865119894119899119892119897 =

Δ119865119894119899119892119897minus1 sum119895isinH119894119892

119884119899119897119895 = 0

Δ119865119894119899119892119897minus1 + sum119895isinH119894119892

119899119897119895119862119895 otherwise (16)



119866119894sum119892=1

119865119894119899119892 (17)





119911ℎ119887 = 110038161003816100381610038161198781198871003816100381610038161003816 sum119895isin119878119887

119909ℎ119895 (18)




























80

85

90

95

100

105

110

115

120

125

130

Payo

ff



Supp

ortin

g ra

tio



0303504

04505

05506

06507

07508

08509

0951





Supp

ortin

g ra

tio

0404505

05506

06507

07508

08509

0951

9 10 11 12 13 14 158Number of BSs





Supp

ortin

g ra

tio



0303504

04505

05506

06507

07508

08509

0951










7 Conclusions


List of Symbols



















Acknowledgments


References
































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation






















Require H119894119892Ensure 119884119899119897119895




Δ119865119894119899119892119897 =

Δ119865119894119899119892119897minus1 sum119895isinH119894119892

119884119899119897119895 = 0

Δ119865119894119899119892119897minus1 + sum119895isinH119894119892

119899119897119895119862119895 otherwise (16)



119866119894sum119892=1

119865119894119899119892 (17)





119911ℎ119887 = 110038161003816100381610038161198781198871003816100381610038161003816 sum119895isin119878119887

119909ℎ119895 (18)




























80

85

90

95

100

105

110

115

120

125

130

Payo

ff



Supp

ortin

g ra

tio



0303504

04505

05506

06507

07508

08509

0951





Supp

ortin

g ra

tio

0404505

05506

06507

07508

08509

0951

9 10 11 12 13 14 158Number of BSs





Supp

ortin

g ra

tio



0303504

04505

05506

06507

07508

08509

0951










7 Conclusions


List of Symbols



















Acknowledgments


References
































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation

































80

85

90

95

100

105

110

115

120

125

130

Payo

ff



Supp

ortin

g ra

tio



0303504

04505

05506

06507

07508

08509

0951





Supp

ortin

g ra

tio

0404505

05506

06507

07508

08509

0951

9 10 11 12 13 14 158Number of BSs





Supp

ortin

g ra

tio



0303504

04505

05506

06507

07508

08509

0951










7 Conclusions


List of Symbols



















Acknowledgments


References
































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation















7 Conclusions


List of Symbols



















Acknowledgments


References
































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation
























Acknowledgments


References
































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation






























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









research article joint radio resource allocation and base

Documents