Development of an Energy Estimation Algorithm for LTE Mobile Access Networks

Transcription

1 Development of n Energy Estimtion Algorithm for LTE Mobile Access Networks 1 E. Obi, 2 S. Grb nd 2 S. M. Sni Deprtment of Electricl nd Computer Engineering, Ahmdu Bello University, Zri. Abstrct - This work proposes the development of n energy estimtion lgorithm for LTE mobile ccess networks. The LTE network environment nd the enodebs power consumption models were developed with view to implementing n energy estimtion lgorithm tht will estimte the energy consumption of the LTE ccess network. The energy estimtion lgorithm for the LTE enodebs ws developed nd implemented in MATLAB environment. The dily energy consumption of the LTE ccess network ws simulted nd nlysed in respect to 37 enodebs which ws used s cse study. The dily energy consumption of the LTE ccess network ws evluted while vrying the energy lod proportionlity constnt (q) which rnges from 0 1. The dily minimum nd mximum energy consumption of the LTE ccess network were found to be 87 kwh nd 1121 kwh for the energy lod proportionlity constnt of q = 0 nd q = 1 respectively. The result showed tht the energy consumption of the simulted LTE ccess network increses linerly s the vlues of the energy lod proportionlity constnt tends towrds 1. Key words - LTE Network, Lod Utiliztion Fctor, Bndwidth Efficiency, Dt rte nd Signl-to-interference-noise-rtio 1. INTRODUCTION The informtion nd communiction technology (ICT) systems consume up to 10% of the world s energy which is responsible for bout 2% of globl CO 2 emissions [1]. The telecommuniction network is one of the min energy consumer of the informtion nd communiction technology sector [1]. About 37% of the totl emissions from ICT devices nd systems re due to the telecommuniction infrstructure nd devices, where bout tenth of the estimte is due to cellulr mobile communiction networks. This ccounts for bout 0.2% of the globl CO 2 emissions nd 1% of the world energy consumption [2]. The mobile cellulr communictions sector lone consumes pproximtely 60 billion kwh per yer. Correspondingly, energy consumption s well s CO 2 footprint of mobile cellulr networks re incresing t n lrming rte due to the exponentil growth in mobile dt trffic [3]. This hs led to high network operting costs nd considerble contribution to the worsening globl wrming phenomenon respectively [4]. On the other hnd, cellulr mobile network trffic exhibits high-degree of temporl-sptil diversity, which mens tht trffic demnd vries both in time nd spce [5]. This vrition is directly relted to the rndom cll mking behviour nd mobility pttern of the mobile users [6]. However, under the current network opertion pproch, ll enodebs re kept powered on irrespective of trffic lod [7]. Thus, it is impertive to nlysis the energy consumption of this enodebs with respect to its utiliztion. The uthors in [8] developed n Alwys-On (AO) scheme to estimte the network energy consumption when ll bse sttions re lwys powered on for time period T. The lgorithm dpted the current network cpcity to the ctul trffic, while gurnteeing n dequte qulity of service to users. However, the scheme ws bsed on the UMTS stndrd of cellulr network. In view of this, there is lso need to develop robust energy estimtion lgorithm for the LTE cellulr ccess networks tht will incorporte the inherent temporl-time trffic diversity of cellulr ccess networks nd the lod-proportionl power consumption model of enodebs. 2. SYSTEM ARCHITECTURE AND MODELS 2.1 Architecture nd Power Consumption An rchitecture of n LTE, 4G enodebs hving three sectors nd four trnsceiver chins per sector is s shown in Figure 1 [9]. Figure 1: Architecture of Three Sector LTE 4G enodeb with Four Trnsmit Antenns A trnsceiver chin consists of rdio frequency (RF) module which is n equipment for generting trnsmit signls to the mobile sttions, power mplifier (PA) tht mplifies the trnsmit signls from the rdio frequency module to high power level suitble for trnsmission, n ntenn feeder, trnsmission ntenn for rditing the signls, switch/duplexer, bse-bnd module for both uplink nd downlink, power supply unit nd cooling system. A single power supply unit nd cooling system is normlly shred by ll the trnsceivers of n enodeb [9]. 2.2 Modeling the Power Consumption The enodeb power consumption model is used for evluting the energy consumption of n LTE cellulr network. The enodebs re modelled to consume power tht is prtly constnt nd prtly vries with the lod fctor t ny given instnt of time. The mthemticl representtion of the 313

2 instntneous power consumed by the jth enodeb is given s [10]: P j (t) = (1 q)ρ j (t)p j + qp j (1) Where: P j (t) is the opertionl power of the jth enodeb t time t ; P j is the mximum opertionl power of the jth enodeb; ρ j (t) is the lod utiliztion fctor of the jth enodeb t time t nd q [0,1] is clled the energy-lod proportionlity constnt of the enodebs which determines the level of dependency of the opertionl power of n enodeb on it lod utiliztion fctor. The opertionl power, P J of the jth enodeb is given s [11]: P J = P TX,j + b (2) Where: P TX,j is the trnsmit power of the jth enodeb; The prmeters nd b re termed s the power profile prmeters [11]. The lod utiliztion fctor t the jth enodeb, ρ j (t) t time t is given s [12]: ρ j (t) = N used.rb.j(t) N rb.j (3) Where: N used.rb.j (t) is the number of used physicl resource block t the jth enodeb t time t; N rb.j (t) is the number of vilble resource block t the jth enodeb. eqution (6) when considering dptive modultion nd coding [13]: 0 if γ i,j < γ min e i,j = { ξ log 2 (1 + γ i,j ) if γ min γ i,j < γ mx (6) e mx if γ i,j γ mx Where: 0 ξ 1 is the ttenution fctor ccounting the implementtion loss; γ min is the minimum signl-tointerference-noise-rtio; γ mx is the mximum signl-tointerference-noise-rtio; e mx is the mximum bndwidth efficiency nd γ i,j (t) is the instntneous received signl-tointerference-nd-noise rtio of the ith mobile sttion from the jth enodeb. The simulted trffic rrivl pttern of n enodeb follows poisson distribution model given s [14]: A(t) = p(t,μ) mx [p(t,μ)] p(t, μ) = μt t! e μ (8) Where: A(t) is normlized trffic t time t, p is the poisson distribution function, t is the specific time in dy nd μ is men vlue where pek number of trffic occurred. Figure 2 shows the pproximte trffic rrivl pttern of rel trffic rrivl pttern in cell with men vlue of 15 [14]. Thus, the pek trffic rte during dy occurs t 3:00pm (7) Also, the number of used physicl resource block t the jth enodeb t time t is given s [12]: N u N used.rb.j (t) = i=1 z i,j (t)w i,t (t) (4) Where: z i,j (t) is n ssignment indictor vrible which is equl to 1 when ith mobile sttion is served by jth enodeb t time t nd zero otherwise; w i,t (t) is the pproximte number of physicl resource block llocted by the jth enodeb to the ith mobile sttion t time t. The pproximte number of physicl resource block llocted by the jth enodeb to the ith mobile sttion t time t is given s [13]: w i,t (t) = R i,j(t) W RB e i,j (t) Where: W RB is the bndwidth per physicl resource block nd it is 180 khz; R i,j (t) is the bit rte requirement of the ith mobile sttion from the jth enodeb t time t; e i,j (t) is the verge bndwidth efficiency of the ith mobile sttion from the jth enodeb t time t. The verge bndwidth efficiency of the ith mobile sttion from the jth enodeb t time t is usully expressed using (5) Figure 2: Instntneous Trffic Normliztion Fctor The lod fctor of ech enodeb in the LTE ccess network is normlized using the instntneous trffic normliztion fctor A(t). This is to mke the trffic behviour t ech enodeb of the propose model exhibits the temporl-sptil diversity of typicl mobile cellulr network. The lod fctor of jth enodeb fter normliztion is given s: ρ j,new (t) = A(t)ρ j (t) (9) Where: ρ j,new (t) is the normlized instntneous lod fctor of the jth enodeb t time t; A(t) is the instntneous trffic normliztion fctor nd ρ j (t) is the clculted lod fctor of the jth enodeb t time t. Thus, the instntneous power consumed by the LTE ccess network of N enodebs t time t is given s: 314

3 N P N (t) = [(1 q)ρ j,new (t)p j=1 j + qp J ] (10) The totl energy consumed by the N enodebs over period of 24 hours cn be computed using eqution (11) nd is termed s the originl/bse-cse energy E N orig. E orig N = 24 t=0 P N (t) (11) 2.3 Modeling Mobile Sttion Distribution In the proposed model, ctive mobile sttions re selected rndomly from set of uniformly distributed mobile sttions. Ech mobile sttion is defined by its X nd Y coordintes. Mobility is simulted by rndomly selecting set (u x,i, u y,i ) of positions within prticulr cell. The distnce of mobile sttion i from n enodeb j is given s [15]: d i,j = ((x i x j ) 2 + (y i y j ) 2 ) (12) Mobile sttions re initilly generted uniformly cross the entire network. A set of ctive mobile sttions is selected from the group of N u uniformly distributed mobile sttions belonging to ech cell. Let D be the distribution fctor of the mobile sttions in ech cell, such tht the X xis is divided into 2(D + 1) sub-divisions nd the Y xis is divided into 4(D + 1) subdivisions. If ech point of intersection of the lines sub-dividing the X xis nd Y xis mrk the position of mobile sttions, the number of mobile sttions tht re locted within every cell for given vlue of D cn be expressed using: viii. Determine the totl number of resource blocks occupied by the entire ctive mobile sttion of n enodeb. ix. Compute the lod fctor of ech enodeb in the LTE ccess network. x. Compute the instntneous trffic normliztion fctor A(t). xi. Normlize the lod fctor of ech of the enodeb. xii. Compute the power consumed t ech enodeb t tht instnt nd store the result xiii. Increment timer. xiv. Repet (i) to (xiii) s long s the simultion time is greter thn timer redings xv. Output the totl energy consumed by the enodeb for the simultion time. 3. SIMULATION AND RESULT 3.1 Simultion Setup The performnce of the proposed energy estimtion lgorithm ws evluted by simultion. The simulted LTE ccess network consist of 37 mcro cells with distribution fctor of 3 nd 50 ctive mobile sttions per enodeb s cse study. The stndrd prmeters used for the simultion re shown in Tble 1 which re consisted with the simultion scenrio recommended by 3GPP [16] TABLE 1: STANDARD SIMULATION PARAMETERS Prmeter Trnsmit power of enodebs Vlue 46 dbm N u = 6D 2 + 8D + 3 (13) System bndwidth 20 MHz 2.4 The energy estimtion lgorithm This lgorithm comprises the step by step pproch required to estimte the energy consumed by enodebs in the LTE ccess network while considering the rndom nture of mobile sttions trffic clsses nd the dily trffic vrition of the enodebs. The following steps of instructions re executed logiclly in order to effectively estimte the energy consumed by the enodebs t ny time of the dy. However the proposed model considers hourly trffic vrition. The sequence of instructions re s follows: i. Initilize timer. ii. Generte the enodebs coordinte mtrices. iii. Generte the uniformly distributed mobile sttions coordinte mtrices. iv. Rndomly select ctive mobile sttions nd generte their coordinte mtrices from the set of uniformly distributed mobile sttions obtined in (iii). v. For ech of the ctive mobile sttion in (iv), rndomly select trffic type from the trffic ctegory mtrix (dt rte nd signl-to-noise-rtio). vi. Compute the verge bndwidth efficiency of ech ctive mobile sttion bsed on its dt rte nd signlto-noise-rtio. vii. Determine the mount of resource block occupied by ech ctive mobile sttion. Crrier frequency Bndwidth per Resource 2 GHz 180 khz Resource block per enodeb 100 The rdius of the mcro cell is chosen s 0.5 km. Adptive modultion nd coding (AMC) set prmeters re given s ξ = 0.75, γ min = 6.5 db, γ mx = 19 db nd e mx = 4.8 bps/hz following [13]. Five clsses of rel time constnt dt rte hving dt rtes equl to 64 kbps, 128 kbps, 256 kbps, 384 kbps nd 512 kbps re rndomly selected by mobile sttions. It is ssumed tht only one resource block cn be llocted to mobile sttion from ny clss. Thus, the required signl-to-noise-rtio of the five clsses, clculted using eqution (5) (6), re found equl to 4.1 db, 0.3 db, 4.3 db, 7.9 db, nd 11.1 db respectively. The enodebs power profile prmeters re: = nd b = for mcro cells. These prmeters provide the mximum operting power of the enodebs. The energy lod proportionlity constnt q rnges from 0 to 1. A snp shot of the simulted network is s shown in Figure

4 Figure 2: Simulted Network of 37 enodebs 3.2 Result nd Anlysis The instntneous power consumption of the 37 enodebs in the LTE ccess network ws simulted for 24 hours for the energy lod proportionlity constnts which rnges from q = 0 to q = 1 t the intervl of 0.1 using eqution (10). Figure 3 shows the plot of the results obtined. Figure 4: Hour Energy Consumption of the LTE Access Network Figure 4 demonstrtes the hourly energy consumption of the enodebs in dy while vrying the energy lod proportionlity constnt q from 0 to 1. The plot shows tht the energy consumption increses s the time of the dy increses. The energy consumption of the enodebs t prticulr time of the dy increses for higher vlue of energy lod proportionlity constnt. The dily energy consumption of the LTE ccess network while vrying the energy lod proportionlity constnt from 0 to 1 for 24 hours is tbulted s follows: Figure 3: Instntneous Power Consumption of the LTE Access Network Figure 3 demonstrtes the dependency of the instntneous power consumption of the LTE ccess network s function of the energy lod proportionlity constnt. For q 1, the instntneous power consumption by the LTE ccess network is highest nd constnt becuse the power consumption does not vry with the normlized instntneous trffic t the enodebs. However, s the vlue of q decreses the instntneous power consumption of the LTE ccess network decreses, but vries more with the instntneous normlised trffic t the enodebs. The instntneous power consumption of the LTE ccess network is minimum t q 0 nd vries completely with the normlized trffic t the enodebs. Also the mximum instntneous power consumption of the LTE ccess network ws found to be t 15 hours which correspond to the pek hour trffic for the vrious vlue of q. TABLE 2: VARIATION OF THE DAILY ENERGY CONSUMPTION WITH ENERGY LOAD PROPORTIONALITY CONSTANT q Dily Energy Consumption (kwh) The plot of the vrition of the dily energy consumption of the LTE ccess network with energy lod proportionlity constnt is given in Figure 5. The hourly energy consumption of the LTE ccess network ws simulted for 24 hours for energy lod proportionlity constnts which rnges from q = 0 to q = 1 t the intervl of 0.1. The simultion of the hourly energy consumption of the LTE ccess network ws done using eqution (11). Figure 4 shows the plot of the hourly energy consumption of the LTE ccess network enodebs. 316

5 REFERENCES Figure 5: Energy consumption with chnge in Energy Lod Proportionlity Constnt Figure 5 demonstrtes the vrition of the energy consumption of the network with energy lod proportionlity constnt. The dily mximum nd minimum energy consumption of the simulted 37 LTE network enodebs re 1121 kwh nd 87 kwh for n energy lod proportionlity constnt of q 1 nd q = 0 respectively. Since q 1 corresponds to constnt energy consumption of enodebs requiring constnt power consumption irrespective of trffic level, it resulted to the highest mount of energy consumption. Similrly q = 0 corresponds to the energy consumption of the enodebs tht completely vries with the trffic level of the enodebs leding to the lowest energy consumption in the network. 4. CONCLUSION The LTE network environment nd the enodebs power consumption model hve being developed. An energy estimtion lgorithm for the LTE enodebs hs being developed nd implemented in MATLAB version 2013b environment. The energy consumption in the enodebs hs been simulted nd nlysed. The dily minimum nd mximum energy consumption of the LTE network enodebs were found to be 87 kwh nd 1121 kwh for the energy lod proportionlity constnt of q = 0 nd q = 1 respectively. The result showed tht the energy consumption in the LTE enodebs increses linerly s the vlue of the energy lod proportionlity constnt increses. Also, it ws found tht the dependency of the instntneous power consumption of the enodebs increses s the vlues of energy lod proportionlity constnt (q) tends to zero. Future work will consider integrtion lod/trffic shring lgorithm into the energy estimtion lgorithm to form n energy sving lgorithm using sleep mode nd self-orgnizing network. [1] Yigitel, Mehmet Aykut, Incel, Ozlem Durmz, & Ersoy, Cem. (2014). Dynmic bse sttion plnning with power dpttion for green wireless cellulr networks. EURASIP Journl on Wireless Communictions nd Networking, 2014(1), [2] Son, Kyuho, Kim, Hongseok, Yi, Yung, & Kirshnmchri, Bhskr. (2013). Towrd Energy-Efficient Opertion of Bse Sttions in Cellulr Wireless Networks. In Jinsong Wu, Sundeep Rngn & Honggry Zhng (Eds.), Green Communictions. Theoriticl Fundmentls, Algorithms nd Appliction (pp ). [3] Wu, Junyong, Zhng, Ye, Zukermn, Moshe, & Yung, Edwrd. (2015). Energy-Efficient Bse Sttions Sleep Mode Techniques in Green Cellulr Networks: A Survey. IEEE Communictions(99), [4] Son, Kyuho, & Krishnmchri, Bhskr. (2012). Speedblnce: Speedscling-wre optiml lod blncing for green cellulr networks. INFOCOM, 2012 Proceedings IEEE, [5] Peng, Chunyi, Lee, Suk-Bok, Lu, Songwu, Luo, Hiyun, & Li, Hewu. (2011). Trffic-driven power sving in opertionl 3G cellulr networks. Proceedings of the 17th nnul interntionl conference on Mobile computing nd networking, [6] Pul, Utpl, Subrmnin, Annd Prbhu, Buddhikot, Milind M, & Ds, Smir R. (2011). Understnding trffic dynmics in cellulr dt networks. INFOCOM, 2011 Proceedings IEEE, [7] Wu, Yiqun, & Niu, Zhisheng. (2012). Energy efficient bse sttion deployment in green cellulr networks with trffic vritions. Communictions in Chin (ICCC), st IEEE Interntionl Conference on, [8] Chirviglio, Luc, Ciullo, Deli, Koutits, George, Meo, Michel, & Tssiuls, Lendros. (2012). Energy-efficient plnning nd mngement of cellulr networks. Wireless On-demnd Network Systems nd Services (WONS), th Annul Conference on, [9] Hn, Congzheng, Hrrold, Tim, Armour, Simon, Krikidis, Ionnis, Videv, Stefn, Grnt, Peter M, Hs, Hrld, Thompson, John S, Ku, Ivn, & Wng, Cheng-Xing. (2011). Green rdio: rdio techniques to enble energy-efficient wireless networks. Communictions Mgzine, IEEE, 49(6), [10] Correi, Luis M, Zeller, Dietrich, Blume, Oliver, Ferling, Dieter, Jding, Ylv, Gódor, István, Auer, Gunther, & Vn Der Perre, Liesbet. (2010). Chllenges nd enbling technologies for energy wre mobile rdio networks. Communictions Mgzine, IEEE, 48(11), [11] Shhb, Suhil Njm, Abdulkfi, Ayd Atiyh, & Zinun, Ayib Rosdi. (2015). Assessment of Are Energy Efficiency of LTE Mcro Bse Sttions in Different Environments. Journl of Telecommunictions nd Informtion Technology(1), 59. [12] Wng, Ho, Ding, Linghui, Wu, Ping, Pn, Zhiwen, Liu, Nn, & You, Xiohu. (2010). Dynmic lod blncing nd throughput optimiztion in 3gpp lte networks. Pper presented t the Proceedings of the 6th Interntionl Wireless Communictions nd Mobile Computing Conference. [13] Hossin, Md, Munsinghe, Kumudu S, & Jmlipour, Abbs. (2013). Distributed inter-bs coopertion ided energy efficient lod blncing for cellulr networks. Wireless Communictions, IEEE Trnsctions on, 12(11), [14] Tun, Khin Cho, & Kunvut, Kungorn. (2014). An Overview of Cell Zooming Algorithms nd Power Sving Cpbilities in Wireless Networks. KMUTNB: Interntionl Journl of Applied Science nd Technology, 7(3), [15] El-Beino, Wissm, El-Hjj, Ahmd M, & Dwy, Zher. (2012). A proctive pproch for LTE rdio network plnning with green considertions. Pper presented t the Telecommunictions (ICT), th Interntionl Conference on. [16] 3GPP, Technicl Specifiction Group Rdio Access Network; Evolved Universl Terrestril Rdio Access (E-UTRA); Rdio Frequency (RF) system scenrios, Technicl Report, 3GPP TR Ver Rel. 11, Sep