3.1. Estimation of Mobile Operators’ Energy Consumption and Energy Intensity of Data Transfer in Finland
The development of the total energy consumption of the main mobile network operators in Finland between 2010 and 2017 are presented in Figure 1
. An estimate of the overall energy consumption of all operators in Finland was around 0.6 TWh/a in 2017. This corresponds to 0.7% of the total annual electricity consumption in Finland in 2017 (85.5 TWh/a). Rapidly increasing mobile data consumption has placed pressures to mobile operators, nonetheless, it seems that these pressures have not yet led to a corresponding growth in total energy consumption related to mobile networks. However, as it is anticipated that the growth in data usage will continue (see Figure 2
), this is a topic that requires attention in future.
The constructed growth trend for the transmitted mobile data in Finland is presented in Figure 2
. The growth trend was prepared based on the data published by FICORA in 2017 [6
]. The polynomial and exponential trends (y = y(x)) presented in Figure 2
were estimated by means of least squares fit using the data in the histogram. X in these equations refers to numbers from 1 to 11 corresponding to the years 2010 through to 2020. The statistic R2
(R-squared, i.e., coefficient of determination) describes goodness of the fit. Values close to one refer to a good fit between the estimated equations and the historical data.
During the second half of the last reported year (2017), the growth in data transmission volumes was 40%, in comparison to the consumption figures from the previous year and the total amount of transferred data in 2017 reached 1,500,000 terabytes [6
]. After a period of fast exponential growth, the pace of growth slowed down, from an exponential equation to a polynomial trend estimated by a means equation. However, the growth rate remained significant. If the polynomial growth trend continues, data consumption could reach 4.5 million terabytes per year by 2020. In addition to exponential and polynomial trends, a linear trend line based on 2015–2017 data was constructed as a lower estimate for 2020. Even if future growth is linear, the current levels of data consumption double by 2020. In this context, it is important to note that previous consumption levels do not predict future consumption, as any change could affect future development. Instead, these developments indicate potential future developments in the event the pace of growth follows earlier development trends.
During the assessed period, the total number of mobile subscriptions in Finland remained rather stable, being 9.5 million in 2017. According to the statistics provided by FICORA, the majority (two-thirds) of the subscriptions were mobile subscriptions (including voice and data), while one fifth were mobile broadband subscriptions (including only data) and 10% were voice-only subscriptions (no data transmission included). In Finland, the number of mobile subscriptions per capita was more than 1.7, in comparison to 1.2–1.6 in other Nordic and Baltic countries, whereas the growth rate in the number of fixed broadband subscriptions was lower in comparison to other countries in the region [8
]. According to OECD statistics, the average monthly mobile data consumption per mobile broadband subscription in Finland in 2016 was the highest in the world (11 GB/month/subscription) and the second highest mobile data consumption level was in Latvia (8.2 GB/month/subscription) [7
By combining the overall electricity consumption estimate for production networks (80% of operators’ overall consumption) (Figure 1
) with previous estimates of overall data usage (Figure 2
), an indicative trend of electricity consumption (kWh) per transferred gigabyte for the years 2010–2017 was created (Figure 3
), together with an estimate for the coming years. The results presented in Figure 3
indicate that the specific electricity consumption per transmitted gigabyte has decreased substantially.
Based on data usage and electricity consumption data, it can be estimated that, during 2016 the specific electricity consumption per one gigabyte decreased below the level 0.5 kWh/gigabyte and will continue to decrease when the capacity utilisation factor of existing 4G LTE networks improves. Based on the equation (presented in Figure 3
) the 0.1 kWh/gigabyte level could be achievable by around 2020 even though annual electricity consumption might increase in comparison to the current level. Thus, new efficiency improvement measures should be introduced and implemented to restrict and control increasing operational costs of electricity.
The results presented here can be considered as preliminary and a direct comparison with other available studies is not possible. However, a brief analysis of the available literature is included in order to evaluate the accuracy of our results. It is difficult to compare our findings to those of earlier studies because the rather rapid technical development and increase in data consumption entails that available estimates quickly become outdated. In addition, life cycle analyses that include energy consumption figures are often case-specific and largely dependent on the assumed network technologies, end-user devices and assumptions related to the use phase.
In previous studies, several estimates about the electricity intensity of Internet data transmission have been presented and thoroughly analysed (See: [18
]). Electricity intensity in this context has most often been defined as energy consumed per amount of transmitted data [19
] and is commonly expressed as a ratio of kilowatt hours per gigabyte (kWh/GB). While existing studies typically use different approaches and system boundaries, most of the existing estimates consider only fixed-line access networks and less information about the mobile access networks seems to be available. This is most likely due to the previously dominating role of fixed access networks in several countries. However, the latest developments related to increasing mobile data usage and wider coverage of advanced mobile network technologies (LTE, 4G) indicate that more information about the environmental impacts related to mobile networks and mobile data transfer will be needed in the future.
Using data volumes from the year 2010 Malmodin and colleagues estimated electricity consumption per data volume as follows: 0.08 kWh/gigabyte for averaged fixed broadband access network, compared to 2.9 kWh/gigabyte for average 3G mobile broadband access network and 37 kWh/gigabyte for average 2G mobile communication [10
]. Within the same study, the energy consumption of data transmission and IP core network was estimated to be 0.08 kWh/gigabyte. Different from our study, the estimates presented by [10
] were prepared using a bottom-up approach, in which the starting point was the measured energy consumption of dedicated network devices.
A modelling and scenario study by Andrae and Edler [3
] included estimates of the electricity consumption of fixed and wired access networks together with estimates of global ICT related energy consumption for the years 2010, 2020 and 2030. Their analysis of the electricity consumption of fixed and wired access networks and data centres revealed that wireless access networks (WAN) are currently considered more energy consuming per kWh/GB than fixed access networks [3
However, due to the increasing data use and improved efficiency of radio network technologies, the situation should change by the year 2030, when 5G network technologies are predicted to become dominant. To achieve this goal and to keep the total electricity consumption of mobile data networks at a reasonable level, an energy efficiency improvement of more than 99 percent per kWh/GB for the WAN is required [3
]. The energy consumption estimates presented in [3
] are relatively close to those generated in our study. In the current study, it was estimated that the specific energy consumption of mobile data transfer in Finland could be under the level of 0.1 kWh/GB by 2020. This is in line with the estimated global consumption levels between 0.047–1.04 kWh/GB by 2020 presented in [3
]. Similarly, our estimate of energy consumption (kWh/GB) in 2010 was 12.3, which is within the range estimated by Andrae and Edler within their study (6–15 kWh/GB) [3
3.2. Energy Efficiency Related Developments in Mobile Networks
Energy efficiency in wireless communications is usually defined as the number of successfully received bits divided by the energy consumed for transmitting and receiving those bits. An alternative definition is the throughput divided by the consumed power, which also results in the same bit/J quantity. (Note that these definitions are different from those used in the previous section, where the focus was more on consumed energy than transmitted bits). For the earlier mobile access generations until 4G, there have not been any requirements for energy efficiency. However, the energy efficiency has been continually improving mostly due to the considerable cell throughput improvements. This behaviour is also visible at national level in Finland, as seen in Figure 3
More recently for 5G systems, network energy efficiency has been defined as one of the main requirements targeting 100-fold improvement over the 4G systems [20
]. The same requirements also stipulate that the area traffic capacity (bit/s/m2
) should improve by a factor of 100. From these requirements, it can be concluded that the power consumption of the 5G network should remain at the same level as the current 4G network power consumption. Very roughly speaking, if the traffic volume in 5G networks increases to more than 100-times that of the traffic volume in 4G networks, it is reasonable to expect that the power consumption in the future mobile access networks will increase.
Increasing mobile data usage might lead to a growing total energy consumption within the mobile access networks in the near future. The background for this is that traffic volume is currently increasing at a rapid rate [6
] driven by the flat rate pricing strategy of the Finnish operators. Currently, there are three operational mobile access network generations: 2G/GSM, 3G/UMTS and 4G/LTE. As the 4G coverage is almost nationwide and the penetration of 4G smart phones and mobile routers is increasing, the traffic volume is shifting from 3G to 4G networks. This has forced the operators to add more capacity and install more 4G equipment to fulfil the data rate expectations of their customers, which, in turn, increases the energy consumption. On the other hand, even though the traffic volumes of the legacy 2G and 3G networks are decreasing, they still consume a significant amount of power. For example, according to the EU Code of Conduct for network equipment, 2G and 3G base stations consume approx. 70% of the busy hour power when they are at the low-load state [21
]. Finnish operators have not yet published any schedules for closing down their 2G and 3G networks, which would obviously bring power consumption savings. Currently, the penetration of voice-over-LTE (VoLTE) service over 4G networks is low, which keeps the 2G or 3G coverage mandatory in order to make conventional phone calls. Another reason for maintaining legacy mobile access networks is that many machine-to-machine (M2M) mobile data subscriptions are still based on 2G connections. However, this is expected to change in the coming years as LTE IoT extensions narrowband-IoT (NB-IoT) and Cat-M become operational [22
In the future, 5G networks will introduce several new concepts, which will have a clear impact on the power consumption and energy efficiency of the mobile access. A conceptual illustration of mobile access via the 5G networks is shown in Figure 4
(Figure adapted from [23
]). Massive multiple-input multiple-output (MIMO) [24
] has been proposed to improve the area traffic capacity and spectral efficiency [25
] of these networks. In massive MIMO, a large number of antennas are used to serve a set of single-antenna user equipment (UE) over the same physical resource blocks. When the number of antennas is much larger than the number of UEs, the signal quality of the UEs can be clearly improved and the transmitted power can be reduced. It is expected that massive MIMO should provide energy efficiency gain over conventional 4G base stations mostly due to the possibility of using low-power RF components [26
]. Increasing the number of antennas not only increases the cell throughput but also significantly increases the circuit power consumption. Thus, for a given number of UEs, there is an energy efficiency-optimal number of antennas to be used for transmission and reception [27
]. If it is possible to adapt the number of active antennas in a massive MIMO base station to the number of served UEs, significant energy efficiency gain can be achieved [28
Dense networks that consist of small cells are essential to the ability of 5G to reach the ambitious ubiquitous user-experienced data rate targets [18
]. Small cells can potentially also significantly improve energy efficiency, especially when the sleep modes are enabled during low or no traffic periods [30
]. An efficient way to enable the deep sleep modes of small cells is so called dual connectivity where large macro cells and overlapping small cells provide control and data plane services, correspondingly [31
]. In this case, small cells can be safely put into deep sleep mode when they are not serving any UEs because the macro cell provides downlink broadcast signalling that enables UEs to connect to the network.
Millimetre wave access is the key to reach really high data rates in small cells through the use of very wide bandwidths (up to 1–2 GHz). Massive MIMO has the potential to fit millimetre wave communications well. The shorter wavelengths at higher frequencies reduce the sizes of large antenna arrays so that they are more practical. Additionally, the high beamforming gain of massive MIMO can partially compensate for the increased path losses that result from millimetre wave propagation [32
]. However, even if the energy efficiency associated with millimetre wave transmitters is expected to improve, the power consumption in the RF circuits will increase without careful design. This can be explained by the increased power consumption in the RF circuits, which is dominated by the maximum operational frequency [33
]. Cloud-access network architecture has been proposed for 5G to improve scalability and resource efficiency. In a cloud-access network, a baseband unit pool is shared among a large number of cells. This improves the utilisation rate of processing resources and decreases cooling power consumption by cutting down the number of site equipment shelters that require air conditioning [34
]. In addition, the centralised control of remote radio units makes it easier to implement cell sleep modes and shaping according to geographical traffic fluctuations [35
In general, mobile communication is evolving from transmission power domination to computing power intense operation as communication distances are becoming shorter. The related challenges need to be considered carefully both at the base station [36
] and mobile terminal side [37
]. Due to the increasing role of ICT and mobile communication within our societies, it is important to consider the potential implications of these technological changes from both the point of view of the end-user (consumer) and the environment. And, instead of developing single technologies, services or products, we should think about how these changes affect the whole system in which they operate.