Cellular networks are one of the most important technological innovations witnessed in recent times, improving the quality of human life and contributing to economic growth. Recently, cellular networks have developed tremendously to offer video streaming data, which explains the rapid growth in mobile subscribers [1
]. Mobile subscriptions grew approximately 3% between 2017 and 2018 to reach 7.9 billion in Q3 2018, while mobile data traffic grew by approximately 79% in the same period to reach approximately 20.7 ExaBytes (EB) per month in Q3 2018 [2
]. Moreover, the number of mobile subscribers will increase exponentially due to continuous development of cellular networks, the advent of the fifth generation (5G) of cellular networks, and the Internet of Things (IoT), which is considered the backbone of emerging applications, moving the telecommunications forward to contribute to quality of life and grow the world’s economy [3
]. By 2024, mobile subscriptions are expected to reach 8.9 billion; meanwhile, cellular IoT connections will reach 4.1 billion, with mobile data traffic using approximately up to 136 EB per month, wherein 74% will be utilized for mobile video traffic, and 5G networks will carry 25% of global mobile data traffic [2
]. Figure 1
presents the growth of global mobile subscribers and mobile data traffic based on cellular generations.
The 5G technology is expected to address some of the critical requirements of network issues because this technology will not only provide personal mobile service but also massive machine communications [4
]. However, 5G is not intended to replace the legacy of cellular networks but rather to enhance and support it to provide superior performance of cellular networks (CNs) [5
]. Thus, the number of base stations, considered the main source of cellular networks energy utilization, will increase considerably [6
]. Judging from the cellular network operators (CNOs) context, energy efficiency (EE) is of paramount interest to CNOs based on its significant economic and ecological influence in the coming years. Therefore, a novel research discipline termed “green communication” has emerged [7
]. What actually constitutes “green communication” is the first question that needs attention. Secondly, is it measurable, and if so, how do we define the degree of “greenness” in CNs? It is plausible to use carbon footprint or CO2
emissions as a measure of “greenness,” but CNs’ carbon emissions are negligible. Apart from the carbon emissions issue, there are other drivers of “green” wireless technology that include financial gains (lower energy costs). With the adoption of economic gains in the “green” concept, energy savings or energy efficiency appears to be a more appropriate way to measure “greenness.” Hence, the concept of “green communication” technology in wireless systems can be made meaningful with a comprehensive evaluation of energy savings and performance in a practical system [8
]. Thus, the goal of green communication is to enhance the EE of BSs, decrease OPEX, and eradicate greenhouse gases. For further information on green wireless communications network, see [9
Environmental and energy concerns are core drivers of green technology.
According to [13
], the world’s yearly electricity usage for the telecommunications sector is increasing and expected to reach 51% of global electricity in 2030 unless the electricity efficiency of wireless access networks and access/data center networks is sufficiently improved. CNs are considered the cardinal contributor to the appreciable rise in energy utilization in the telecommunication sector [14
]. Figure 2
shows the expected global electricity usage of cellular networks based on different cellular generations to 2030. BSs are the major consumer of energy in CNs and account for 57% of the total energy consumed [15
]. The number of BSs worldwide is increasing, which led OPEX to rise conspicuously because a greater percentage of the OPEX entails electricity bills [16
CNOs contribute significantly to greenhouse gas (GHG) emissions. According to [17
], the carbon dioxide (CO2
) quantity emitted by the mobile sector is envisaged to increase to 179 MtCO2
by 2020, which translates to 51% of the carbon footprint of the information and communication technology (ICT) sector.
Many interesting studies [9
] have been conducted to find a “greener” cellular network and reduce OPEX. These studies can be categorized into two main approaches: first, enhance the energy efficiency of hardware [18
]; and second, intelligent management of core network elements driven by dynamic traffic load fluctuations, such as switching BSs and cell zooming [22
]. However, BS switching on/off is agreed to be a justifiable approach to improve the energy efficiency for the following reasons. (i) The approach of the improving energy efficiency of the hardware entails high implementation costs. Thus, network owners must meticulously consider the operational and economic dimensions of this approach before initiating hardware replacement [9
]. (ii) BS switching off/on schemes can be easily executed using the current network topology without hardware replacement. (iii) Out of the total energy consumption of CNs, BSs are responsible for 57% [9
]. (iv) The infrastructures of CNs are fashioned to support daytime traffic. Diurnal traffic loads differ from nocturnal traffic loads. As a result, energy wastage arising because of inefficient use of resources cannot be overemphasized, especially for low traffic loads [23
]. (v) The BS switch-off approach is a system-level approach that operates in an area covered by multiple cells, where those cells may use different radio access technologies that offer various services.
In the philosophy of the BS switching on/off technique, the BSs monitor and exploit the traffic load of the cellular network to create a trade-off between network performance and EE. This can be implemented by a dynamic BSs switching on/off strategy motivated by the traffic demand conditions. In areas where there is low traffic, it is advisable that the BSs be switched off. However, with a minimum number of BSs switched on for service provisioning, users’ coverage and the basic operations of the network can be supported.
Operations of BSs switching on/off should adopt certain operational guidelines, which should be considered for execution and evaluation along CN. The typical questions that arise are as follows: (i) Which cells should be switched off and which cells should remain active? (ii) How can we manage the cooperation between BSs via a dynamic BSs switching off/on technique? (iii) What policies do we adopt for the BSs switching implementation? (iv) Parameter-wise, which are the key ones to be considered when executing the switch-off session?
The key contributions of this study are summarized as follows:
To propose a mechanism of cooperation between the LTE and next-generation wireless networks, such as 5G. Thereby, we create an equilibrium between network performance and EE via a 5G BSs switching off/on strategy driven by the network instantaneous traffic load demand while guaranteeing service coverage for mobile subscribers by the remaining active LTE BSs. The proposed BSs switching on/off decision-making algorithm is presented in Section 3.1
To determine the optimum criteria of the active LTE BSs (transmission power, the total antenna gain, bandwidth/spectrum, and SINR) that achieves the maximum coverage for the entire area during the 5G BS switch-off session.
1.3. Paper Organization
describes the work conducted. In Section 3
, the detailed system model and problem formulation are presented. Optimization programming and simulation setup are presented in Section 4
. Results and a comparison with previous works are presented in Section 5
to evaluate the performance of the proposed mechanism. Section 6
gives our conclusions.
2. Related Work
A plethora of works have studied the switch-off approach. Various strategies for switching off BSs based on Universal Mobile Telecommunications System (UMTS) CNs during off-peak traffic conditions have been implemented [25
]. Chiaraviglio et al. (2008) explored the likelihood of switching off some cells and BSs in the UMTS network during off-peak traffic durations, at the same time guaranteeing quality of service (QoS) constraints from the perspective of blocking probability and electromagnetic exposure limits. The researchers analyzed three kinds of scenarios: non-commercial, commercial, and hierarchical. It was shown that there was a 50% power saving in all the three scenarios utilizing BS switching. The previous work was extended by proposing a responsive network planning framework for BSs switching off/on involving uniform and hierarchical scenarios [26
]. The results were extended in [27
], and a realistic regular cell architecture was proposed wherein each of these configurations has a specific energy-saving ratio by turning three out of four or eight out of nine BSs off, thereby achieving an energy savings on the order of 25% to 30%. Furthermore, two energy-saving mechanisms have been proposed in [28
]: (i) A greedy centralized algorithm, which is based on the concept that each BS determines which BS is fit to be switched off based on the traffic, and (ii) a decentralized algorithm involving an individual BS locally estimating its traffic load and making independent decisions as to whether it will be switched off. Gong et al. (2010) adopted a dynamic switch on/off algorithm motivated by blocking probabilities. The BSs switched off are moderated based on the traffic variation with regards to a blocking probability constraint [29
]. Xiang et al. (2011) analyzed the minimal number of active BSs that will be utilized based on the balance between fixed and dynamic powers [30
]. Lorincz et al. (2012) proposed a new UMTS cellular access network energy-saving optimization model [31
]. Using the average distance among BSs and UEs as a decision-making metric, Bousia et al. (2012) proposed an algorithm in which the BSs with maximum average distance will be switched off [32
This study proposes an approach to cooperation between the LTE and 5G technologies, which aims to strike an equilibrium between network performance and energy savings via switching off/on the 5G BSs influenced by instantaneous load traffic, ensuring the service coverage of the remaining mobile subscribers by the remaining active LTE BSs.
5. Results and Discussion
This section first discusses the performance evaluation of the LTE cell coverage area optimization issue during the period when 5G BSs are switched off based on the PSO under constraints, the transmission power of the BS, the total antenna gains, the flexible BW, and the propagation environment (SINR, path loss, and fading) according to the optimization problem formula given in Equation (11) and the setup parameters listed in Table 1
. The second part of the discussion focuses on the data rate that will be delivered to the UEs during 5G BSs’ switching off and on, because the EE is a function of both the data rate and the total BS power consumption. The final part of this section evaluates the EE based on the proposed BSs switching-off/on model.
and Figure 8
describe the behavior of the LTE BSs coverage when the 5G BSs are switched off at low traffic load and the behavior of the constraints that affect the coverage, respectively.
shows that at the start, with random particle positions and zero velocities (at iterations equal to zero), the LTE cell coverage area was 99.05%, with Ptx
= 43.08 dBm
= 7.97 dB, at SINR
= 1.49 dB and σ
= 5.50 dB, as shown in Figure 6
. From Figure 6
, when the SINR
decreased, both Ptx
increased to maintain the maximum coverage area. The simulation results demonstrated that the optimum transmitting power Ptx
and the antenna gain G
that maintain maximum coverage at the edge (where the SINR
was the lowest at −5.1 dB and where the shadowing is at 4.8 dB) are 43.1 dBm
and 6.3 dB, respectively. In addition, the optimal BW
is 10 MHz during the period that the 5G BSs are switched off; the optimal BW
is proportional to Pmin
, as given in Equation (11). However, for a high data rate, a BW greater than 10 MHz can be used. In this case, the full coverage is not secure because the Pmin
required is high. It is plausible that big bandwidth has the possibility of improving the EE when compared to a small bandwidth within the same size coverage area based on the premise that big bandwidth can support more resource blocks, eventually leading to a higher data rate.
For the downlink (DL), i.e., from the BS to the UE, the BS typically selects the MCS based on the channel quality indicator (CQI) feedback characteristics of the UE receiver, i.e., the SINR via an adaptive link strategy. Based on assumptions made in [38
], Figure 9
shows the relationship between the radius of the cell, Pmin
, and the MCS. The received power level is obtained from the sensitivity calculation in Equation (9), and the values of SINR
, etc., are retrieved from Table 1
reduces, the MCS reduces simply because the demodulation error rate rises as a result of the rise in both the noise and interference. This analysis is of a cell radius of 500 m (edge of the LTE cell), denoting a cell experiencing a low-traffic case; the lowest modulation rate (QPSK) supports a cell radius of 500 m. For LTE with a 10 MHz BW, this case involves 50 resource blocks (RBs) with each RB including 12 subcarriers, each subcarrier having seven symbols for normal CP, and the time of the slot set to 0.5 ms. Hence, the total number of symbols per RB is 12 × 7 × 2 = 168 symbols per ms. Therefore, 8400 symbols per ms are identified in this case. When 1/8 QPSK is used (2 bits per symbol), the data rate will be 2.1 Mbps for a single chain, and with 2 × 2 MIMO (2T
), the data rate will be twice that of a single chain, i.e., 4.2 Mbps, for the worst case of SINR. Figure 10
shows the data rate versus cell radii, with Ptx
= 43.1 dBm
= 10 MHz.
For a 5G mm-wave small cell with a 500 MHz BW, 694 RBs are available. LTE technology comprises a RB having 12 subcarriers with 60 kHz subcarrier space. Then, each of the subcarrier has seven symbols for normal CP and six symbols for extended CP. Both the normal and extended CP have a slot time of 0.1 ms [47
]. Mathematically, it is possible to estimate the number of symbols per RB as 12 × 7 × 10 = 840 symbols per ms, corresponding to 582,960 symbols per ms. On the other hand, if 1/8 QPSK is deployed for edge users (2 bits per symbol with a coding rate of 1/8), the data rate is 0.146 Gbps for a single chain. Therefore, with eight antenna sectors, the data rate is 8 times that of a single chain, for a total of 1.167 Gbps. In addition, with a high-order modulation 64QAM, the total data rate with 500 MHz BW and eight antennas is up to 22.4 Gbps. Figure 11
summarizes the data rate versus MCS and BW for eight antennas. Moreover, Figure 12
summarizes the data rate that can be achieved with various numbers of antennas for different BWs at the edge of 5G small cells.
Hence, in a high-traffic case, both the LTE and 5G cells are active and serve the users in the network. However, the priority goes to the 5G small cells that can provide the minimum data rate at the edge of the 5G cell of 1.167 Gbps at 1/8 QPSK with BW
= 500 MHz and eight antennas used, as shown in Figure 9
. By contrast, for the low-traffic case (energy saving case), the 5G BSs are switched off, and coverage is guaranteed by the LTE BSs; the maximum data rate at the edge of the LTE cell is 4.2 Mbps (R
= 500 m) at 1/8 QPSK with BW
= 10 MHz and two antennas used, as shown in Figure 8
. However, the 4.2 Mbps data rate during the low traffic case (11 PM to 10 AM) is acceptable because the number of subscribers that are using the network for downloading data is so low during that time. Meanwhile, the voice calls are secured with full coverage by the LTE BSs. Figure 13
shows the EE performance for LTE BSs versus the cell radii.
From Figure 13
, the EE at the edge of the LTE cell is 7.19 kb/j, which was computed based on Equation (18), i.e., a data rate of 4.2 Mbps at 1/8 QPSK with BW of 10 MHz and two antennas used, divided by the total power consumption of the 584.12 W.
For a 5G small cell, the EE at 1/8 QPSK with BW 500 MHz and eight antennas used is 25.06 Mb/j, as shown in Figure 14
, where the data rate of 1.167 Gbps is divided by the total power consumption of 46.57 W. Moreover, Figure 15
shows EE versus the number of antennas for the different BWs at the edge of the 5G small cell.
Finally, the energy savings achieved by the proposed method is briefly discussed. The simulation layout involved seven 5G BSs and covered a smaller area than the LTE BS (as shown in Figure 3
). The 5G BSs are active for only 13 h (10 a.m. to 11 p.m.) during peak time (as shown in Figure 2
) and are in sleep mode at other times. The LTE BSs are active all the time; during high-traffic periods, they ensure the coverage and support of the 5G BSs, whereas during low-traffic periods, LTE BSs have two functions, data delivery and ensuring maximum coverage and radio service.
Without the switch-off, all BSs are active for 24 h:
With the proposed switch-off:
The energy savings that can be achieved is 3.52 kWh per day in this case study, which meets the needs of mobile subscribers for high-speed data at peak time. Figure 16
summarizes the data rate versus network power consumption over time based on the network layout shown in Figure 1
. However, the energy savings in this approach depends on the number of BSs that will be turned off. If a large number of BSs are turned off, then the energy savings will be substantial.