Measuring and modeling mobile phone charger energy consumption ...

Measuring and modeling mobile phone charger energy consumption and environmental impact Mikko V. J. Heikkinen

Jukka K. Nurminen

School of Electrical Engineering Aalto University, Espoo, Finland Email: [email protected]

School of Science Aalto University, Espoo, Finland Email: [email protected]

Abstract—This paper studies the electricity consumption of mobile phone chargers. The charger’s electricity consumption varies depending on its state. We measured the electricity consumption of various phone and charger models in different states. We also did panel studies on the recharging behavior of smartphone users. Based on these and other sources, we are able to estimate mobile phone recharging electricity consumption, cost, and CO2 emissions both in Europe and in the USA. Our analysis shows that the actual recharging of batteries consumes only 40% of the total energy; the rest is wasted mainly by unnecessarily plugged-in chargers consuming 55% of the total energy.

I. I NTRODUCTION The number of mobile phones in the world is rising rapidly. Therefore the environmental impact of the mobile phones starts to be a relevant concern. Although recharging a single mobile phone is not consuming much electrical energy, the extremely large user base means that the aggregate energy consumption of mobile phones is significant. In this paper, we analyze the electrical energy consumption of mobile phone recharging. We want to understand how much energy is spent on recharging the mobile phones, how large impact the recharging behavior of consumers has on energy consumption, and how much savings could be achieved if the users changed their behavior. We also measure if there are fundamental differences between the recharging energy consumption between different mobile phone vendors. We can divide the recharging process into a number of states. When both phone and charger are disconnected from the power grid, no external energy is consumed. The phone consumes the power stored in its battery. The battery capacity, user actions, and the energy-efficiency of the phone hardware and software determine how fast the charge in the battery declines. At some point the user decides to recharge the phone. This decision can arise from alert of low battery, suitable recharging opportunity, or regular habit [1]. While the energy consumption of operating on battery power does not have a direct influence on the electrical energy consumption from the power grid, the indirect effect is important. High power consumption leads to frequent recharging needs, and thus to higher electrical energy consumption from the power grid. In the recharging state, both phone and charger are connected to the power grid; the user has plugged in the phone for

recharging. The energy and power management logic monitors the battery charge and the battery is actively recharged. When battery charge has reached full capacity, the system enters the idle state, where it either allows some battery consumption before restarting active recharging, or draws power from the grid instead of the battery. The detailed mechanisms governing these decisions are vendor specific. In the no-load state, the phone is disconnected but the charger is left plugged into the grid. The charger may still consume some energy in the form of current leakage. Active R&D work aims minimizing the wasted energy and is strongly encouraged by regulators (see [2]). As an additional approach, some phone vendors try to influence user behavior with alerts suggesting the user to disconnect the charger when recharging is complete. From this analysis we can conclude that there are three main factors that influence the recharging electrical energy consumption of a mobile phone: 1) The usage pattern of the phone: how actively the user uses different applications, how mobility and signal strength variations influence power consumption, etc. Draining the battery faster results into frequent need to recharge and via it to increased energy consumption. 2) The recharging behavior of the phone user: when do users typically recharge their phones, how long the phone is connected to the charger, etc. 3) The charger power consumption in different states (recharging, idle, and no-load). The first two points depend a lot of the phone user’s behavior while the last one is beyond the control of the phone user. In this paper we study these points in detail. We measure the electrical energy consumption in different states with various phone and charger models and complement the measurements with studies on users’ recharging behavior to estimate the aggregate electricity consumption in macro scale in Europe and in the USA. This paper is structured as follows. In Section II we briefly review related studies. Section III reports our measurements, Section IV describes our model and the results, and Section V discusses the results.

II. R ELATED W ORK We found two studies quantifying the energy consumption of mobile phones and networks. Etoh et al. [3] studied the total energy consumption of the largest mobile telecommunications operator in Japan in 2006 and estimated that an average customer’s phone consumes 0.83 Wh per day whereas the mobile network consumes 120 Wh per day per customer. Schaefer et al. [4] estimated the German mobile telecommunications sector to consume 79.5 Wh per day per subscriber in 2000, phones being responsible to 36 Wh of the consumption. The discrepancy between these two studies illustrates the challenge of estimating the power consumption of mobile telecommunications technologies. Research on recharging behavior of mobile phone users is rather limited. Ferreira et al. [5] assessed the recharging behavior of a large sample of Android users, but reported only limited details on their behavior. (For example, they did not report numbers and durations of recharging periods.) Heikkinen et al. [6] had a smaller sample but reported more details. Some previous work has quantified the energy efficiency of chargers for mobile phones [7]–[9]. As mobile services are increasingly using resources residing in the Internet, the energy consumption of the Internet and the data centers connected to it is becoming increasingly relevant to understand the overall picture of the energy consumption of mobile services [10]– [12]. III. M EASUREMENTS We used two types of measurements for our analysis: recharging behavior measurements with a usage logging software in two panel studies, and exploratory recharging power measurements of several phone and charger models. A. Phone recharging behavior measurements The first usage logging panel study (P07) started in November 2007 and ended in February 2008. The invitation to participate was sent by SMS to ca. 13,500 customers of three Finnish mobile network operators. The panel consisted of 253 participants with mean 50 active days in the panel. A typical panelist was male (81.7%), employed (68.0%), and 20-39 years of age (73.3%). The second usage logging panel study (P08) started in October 2008 and ended in December 2008. This time the invitation to participate was sent by SMS to ca. 10,000 customers of three Finnish mobile network operators. The panel consisted of 105 participants with mean 31 active days in the panel. Again, a typical panelist was male (73.3%), employed (64.8%) with mean 35 years of age (25th percentile 28 years, and 75th percentile 40 years). The demographics of both panels are biased compared to the general population in Finland, but not necessarily compared to the Finnish users of advanced phones at the time. Participation in the panels required a phone having a Nokia Symbian S60 Third Edition operating system and installation of the usage

TABLE I D URATIONS OF RECHARGING AND IDLE PERIODS PER PANELIST ( H / DAY )

Panel P07 Recharging Idle Panel P08 Recharging Idle

10th percentile

90th percentile

Mean

0.05 0.01

2.01 4.92

0.67 1.10

0.08 0.02

1.67 4.22

0.60 1.14

TABLE II E XPLORATORY RECHARGING POWER MEASUREMENTS RESULTS Phone

OS

Recharging (W)

Idle (W)

Apple iPhone 4 Ericsson U20i Ericsson X10i Motorola Milestone Nokia 6020 Nokia 6120c Nokia 6300 Nokia E71 Samsung Omnia 7 ZTE Blade

iOS Android v2.1 Android v2.1 Android v2.2 S40 S60 v3.1 S40 S60 v3.1 Windows 7 Android v2.1

3–4 4–5 4–5 4–5 3–6 1–6 1–6 1–6 2–5 4–5

1–2 0–1 0–2 0–1 0–0 0–0 0–0 0–0 0–1 2–3

logging software. Heikkinen et al. [6] discuss the sampling for the panels and the results from them in detail. For the purposes of this paper, we are interested in the recharging behavior of the panelists. Table I reports the daily durations of recharging and idle periods per panelist. B. Exploratory recharging power measurements We also did exploratory measurements to quantify the power consumption of different phones and their chargers. We measured the minimum and maximum power values during recharging, idle and no-load periods with a measurement device, which was placed between the charger and the wall socket. The device has an accuracy of 5% ± 0,5 W [13]. Table II reports minimum and maximum power consumption values for different phones and corresponding charger models. (Charger model information is omitted to save space.) In all cases the no-load consumption was measured to be 0 W but is likely to be higher because the accuracy of our metering device was only 0.5W. Our power consumption measurements should be considered exploratory for various reasons. First, we were able to reach only a coarse-grained accuracy with our measurement setup. Second, we report values from recharging cycles assumed to be stable over time, but we did not do any systematic analysis on the recharging cycles. Third, the batteries of the phones were of varying condition in terms of age and remaining capacity. Fourth, we measured a limited number of phones and chargers (for example, we were unable to measure any BlackBerry models). However, we consider our measurements sufficient for the purposes of this paper, as our

intent is to estimate total energy consumption, not to conduct definitive measurements. Based on the measurements, it appears all phone and charger combinations consume little power during no-load periods. Nokia phones and the Windows phone consumed very little power also during idle periods, whereas Android phones and the Apple iPhone power consumption varied as a function of user actions during idle periods. We believe the difference is due to a design choice whether to let the phone draw power from the battery or from the electricity grid while plugged into the charger. Also, we believe the differences during the recharging period are due to different recharging cycle designs. IV. M ODEL

TABLE III M ODEL INPUT VALUES Variable

Min

Max

Average

Precharging (W) Pidle,g (W) Pidle,b (W) Pno−load (W) trecharging (h/day) tidle (h/day) rplugged Europe radvanced rAndroid riOS nsubs penergy (EUR/kWh) rCO2−e (kg/kWh) USA radvanced rAndroid riOS nsubs penergy (USD/kWh) rCO2−e (kg/kWh)

3.50 1.00 0.03 0.03 0.06 0.02 0.90

4.50 2.00 0.30 0.30 1.84 4.57 0.99

4.00 1.50 0.17 0.17 0.63 1.12 0.95

0.25 0.10 0.15 520,000,000

0.35 0.40 0.25 630,000,000

0.30 0.25 0.20 575,000,000 0.1645 0.578

0.25 0.25 0.20 285,646,200

0.35 0.55 0.30 302,900,000

0.30 0.40 0.25 295,000,000 0.115 0.592

In this section we present our model for estimating the energy consumption of mobile phone recharging. We use power values in watts to estimate the energy consumption of chargers. We make a distinction between power consumption while recharging Precharging , while idle on electrical grid power Pidle,g , while idle on battery power Pidle,b , and while under no load (i.e., no phone plugged into a charger in the grid) Pno−load . We also need to consider the share of chargers plugged into the grid all the time rplugged . Many people, especially in developed economies, have a tendency of leaving the charger plugged in all the time. It is also increasingly common that people have multiple chargers, e.g. one at home and one in the office. In the extreme case more than one charger for each phone could be plugged in. Corresponding to the power values, we estimate the daily duration a charger is recharging trecharging , idle tidle , and has no load but is plugged into the wall outlet tno−load :

the ratio of carbon dioxide equivalent emissions released per electrical energy consumed rCO2−e . Using the above variables, calculating the total energy consumption E of chargers during a period of time is straightforward by adding up the energy consumption during recharging Erecharging , idle Eidle and no load Eno−load phases:

tno−load = (tmax − trecharging − tidle ) ∗ rplugged

E = Erecharging + Eidle + Eno−load

(1)

where tmax is 24 hours. Next, we introduce the variables varying based on the region under study. Share of advanced phones radvanced is the market share of so-called “smartphones” (i.e., phones with the Android, Apple iOS, BlackBerry, Symbian, Windows Mobile, or other advanced operating system). As Android and iOS phones are currently the most prominent class of advanced phones consuming grid power while not recharging but connected to a charger (i.e., idle on grid power), we use the combined share of Android rAndroid and iOS riOS phones of the share of advanced phones radvanced to estimate the share of phones idle on grid power ridle,g : ridle,g = (rAndroid + riOS ) ∗ radvanced

(2)

Then, we can calculate an estimate for power consumption while idle connected into charger Pidle : Pidle = ridle,g ∗ Pidle,g + (rmax − ridle,g ) ∗ Pidle,b

(3)

where rmax is 1 (i.e., 100%). Other region-dependent variables are the number of active subscriptions nsubs , the price of electrical energy penergy , and

(4)

where Erecharging = nsubs ∗ Precharging ∗ trecharging Eidle = nsubs ∗ Pidle ∗ tidle Eno−load = nsubs ∗ Pno−load ∗ tno−load The price of the consumed energy p = E ∗ penergy and the corresponding emission of carbon dioxide e = E ∗ rCO2−e . A. Input values We calculate scenarios with input values, which we consider to be the minimum, maximum, or average values. The input values are given in Table III. They can be classified into region-independent and -dependent values. The region-independent values are derived as follows: Recharging and idle on grid power values Precharging and Pidle,g are averaged over our power measurements, see Table II. Idle on battery power and under no-load values Pidle,b and Pno−load are based on a technical specification by Nokia [14]. (Other vendors do not publish detailed information about their chargers.) The daily durations for recharging trecharging and idle tidle are averaged over our behavior measurements, see

Table I. We did not find any source for the share of chargers plugged in all the time rplugged , so we estimated it. The region-dependent values are derived as follows: We only consider Europe and the USA, because we have no usage data generalizable to other markets. We define Europe to consist of the EU, EFTA, and CEFTA countries (see [15]). The shares of advanced phones radvanced and phones with the Android and the iOS operating systems, rAndroid and riOS , respectively, are based on analyst estimates [16]–[20]. The numbers of active subscriptions nsubs are based on data from [21]–[24]. The European energy price penergy and the ratio of carbon dioxide equivalent emissions released per electrical energy consumed rCO2−e are based on data from EU-27 countries [25]–[27]; the US numbers are based on data from [28], [29]. B. Results Table IV contains the results from our model for the minimum, maximum, and average scenarios for both Europe and the USA. They are given annually per subscriber. For each recharging state, the impact of consumption is quantified in kilowatt-hours (kWh), currency (EUR or USD), and kilograms of carbon dioxide equivalent emissions (CO2 -e kg). The average annual energy consumption of a subscriber (2.34 kWh) approximately equals to 39 hours of operation of a 60 W light bulb. Assuming 175 g/km CO2 -e emissions for a car, the average annual emissions of a subscriber (1.4 kg) equals to about 8 km of driving a car.

TABLE IV R ESULTS FROM THE MODEL : ANNUALLY PER SUBSCRIBER

Europe Recharging Idle No-load Total Recharging Idle No-load Total Recharging Idle No-load Total USA Recharging Idle No-load Total Recharging Idle No-load Total Recharging Idle No-load Total

Min

Max

Average

Unit

0.08 0.00 0.17 0.25 0.01 0.00 0.03 0.04 0.05 0.00 0.10 0.15

3.02 1.15 2.59 6.76 0.50 0.19 0.43 1.11 1.75 0.66 1.50 3.91

0.92 0.14 1.27 2.34 0.15 0.02 0.21 0.38 0.53 0.08 0.74 1.35

kWh kWh kWh kWh EUR EUR EUR EUR CO2 -e kg CO2 -e kg CO2 -e kg CO2 -e kg

0.08 0.00 0.17 0.25 0.01 0.00 0.02 0.03 0.05 0.00 0.10 0.15

3.02 1.34 2.59 6.96 0.35 0.15 0.30 0.80 1.79 0.80 1.54 4.12

0.92 0.17 1.27 2.37 0.11 0.02 0.15 0.27 0.55 0.10 0.75 1.40

kWh kWh kWh kWh USD USD USD USD CO2 -e kg CO2 -e kg CO2 -e kg CO2 -e kg

C. Comparison of results to other estimates Table V contains estimates derived from other sources. Our average estimate of mobile phone energy consumption in the USA is within the same range as estimates of mobile phone energy consumption in Germany in 2000 [4], but an order of magnitude higher than an estimate of mobile phone energy consumption in Japan in 2006 [3], [30]. Our average estimate of mobile phone energy consumption in Europe is within the same range as an estimate of mobile phone energy consumption in the USA in 2004 [7], [21], but much lower than the estimate on the energy consumption of mobile networks in Japan in 2006 [3], [30]. For comparison to other types of energy consumption, our average estimates are lower than estimates of Google’s energy consumption in 2010 [12] and an order of magnitude lower than estimates on the energy consumption of data centers in Western Europe or the USA in 2005 [11] or the total standby power consumed by consumer appliances in the USA in 1998 [31]. Finally, our average estimates were ca. 0.02% and 0.04% of the estimates on total net electricity generation in the USA and in Europe [32], respectively. D. Sensitivity analysis We also did elementary sensitivity analysis to the model, where we varied a single input parameter within its theoretical minimum and maximum range, and observed its effect on output in the average scenario. Table VI contains the range

TABLE V C OMPARISON TO OTHER ESTIMATES (1 E 9 K W H / YEAR ) Mobile phones Japan 2006 Mobile phones Germany 2000 Mobile networks Germany 2000 Mobile phones USA Mobile phones USA 2004 Mobile phones Europe Google total energy consumption 2010 Mobile networks Japan 2006 Data centers Western Europe 2005 USA standby power 1998 Data centers USA 2005 Total Electricity Net Generation Europe 2008 Total Electricity Net Generation USA 2009

0.03 0.6 0.7 0.7 1.3 1.3 2.3 4.6 41.3 44.3 56.0 3,421 3,953

of both input values and output values. The range of output is reported as a comparison to the average scenario. Our model is most sensitive to recharging and idle durations, and to power consumption under no-load; moderately sensitive to the share of chargers plugged in all the time, and to the share of phones idle on grid power; and least sensitive to power consumption while idle or recharging.

TABLE VI S ENSITIVITY ANALYSIS

Pidle,g (W) Pidle,b (W) Precharging (W) ridle,g rplugged Pno−load (W) tidle (h/day) trecharging (h/day)

Input Min 1.0 0.03 3.5 0.01 0.01 0.03 0.01 0.01

range Max 2.0 0.3 4.5 0.99 0.99 0.3 24.00 2.00

Output Min -1.7% -2.0% -4.9% -4.3% -53.9% -44.6% -4.6% -37.3%

range Max 1.7% 2.0% 4.9% 19.9% 2.3% 44.6% 94.7% 82.2%

example, the phone could consume battery power while idle, and grid power while performing energy-intensive functions. Further work is warranted at least on the following topics: defining the share of mobile phone chargers plugged in all the time, and conducting more accurate measurements on the power consumption of mobile phone chargers (including analysis of recharging cycles and batteries of various condition). Gaining a holistic understanding including the energy usage of mobile networks and service infrastructure is also of interest. ACKNOWLEDGMENT

V. D ISCUSSION As with any model, the accuracy of input values to our model defines the accuracy of the output. The uncertainty in our input values prevents us from giving a conclusive answer for the energy consumption and environmental impact of mobile phone chargers, but we can characterize the possible ranges of both input variables and resulting scenarios. We did only exploratory measurements with different mobile phone and charger models, but as we demonstrated in Sec. IV-D, our model is not very sensitive to power consumption related input values. We believe we have very reasonable behavior estimates based on two subsequent panel studies, but we admit the challenges of generalizing results from panel studies to large populations. According to our model, recharging constitutes ca. 40% of the total mobile phone energy usage. Having chargers unnecessarily plugged in the grid wastes ca. 55% of the energy usage, whereas having idle phones plugged into chargers wastes ca. 5% of the energy usage. Our model suggests a major energy saving could be gained by unplugging chargers after recharging a mobile phone. However, a typical Western consumer probably considers the cost savings and environmental benefits not worth the hassle of unplugging the charger. Therefore, charger manufacturers should intensify their efforts to produce chargers wasting minimum amount of energy while under no load (see [2]). Vendor-specific solutions trying to influence user behavior (e.g., prompting users to unplug chargers) are likely to be less efficient as technical improvements. However, while the operating time with a single recharge is clearly a competing factor improving the user experience, the electricity consumption of the recharging process is not immediately relevant to the end user and therefore of smaller importance to mobile phone vendors. Our exploratory analysis suggests phone vendors have two different approaches towards maintaining charge while phone is plugged into a charger: the phone either takes energy from battery or from the power grid. Consuming power from the grid may waste energy, but guarantees that the phone remains charged if it is used for high-energy-consuming functions (e.g., acting as a mobile modem) while plugged in. One approach would be to make the phone intelligently decide which approach to utilize when plugged into a charger: for

MH’s work was supported by the MoMIE project, TEKES and industrial partners. R EFERENCES [1] M. V. J. Heikkinen and J. K. Nurminen, “Consumer attitudes towards energy consumption of mobile phones and services,” in VTC 2010-Fall, Sep. 2010. [2] Nokia, “Energy efficiency: Chargers.” [3] M. Etoh, T. Ohya, and Y. Nakayama, “Energy consumption issues on mobile network systems,” in SAINT, Jul. 2008, pp. 365–368. [4] C. Schaefer, C. Weber, and A. Voss, “Energy usage of mobile telephone services in Germany,” Energy, vol. 28, no. 5, pp. 411–420, Apr. 2003. [5] D. Ferreira, A. K. Dey, and V. Kostakos, “Understanding humansmartphone concerns: A study of battery life,” in Pervasive, Jun. 2011, pp. 19–33. [6] M. V. J. Heikkinen, J. K. Nurminen, T. Smura, and H. H¨amm¨ainen, “Energy efficiency of mobile handsets: measuring user attitudes and behavior,” Submitted for publication. [7] P. Ostendorp, S. Foster, and C. Calwell, “Cellular phones: Advancements in energy efficiency and opportunities for energy saving,” Natural Resources Defense Council, Tech. Rep., Oct. 2004. [8] S. Weier and R. A. McMahon, “Calorimetry for power conversion in mobile telephone chargers,” in UPEC, Sep. 2007, pp. 986–991. [9] J. Ruutu, J. K. Nurminen, and K. Rissanen, “Energy efficiency of recharging a mobile device,” in NGMAST, Sep. 2011. [10] J. Baliga, K. Hinton, and R. S. Tucker, “Energy consumption of the Internet,” in COINACOFT, Jun. 2007. [11] J. G. Koomey, “Worldwide electricity used in data centers,” Environmental Research Letters, vol. 3, no. 3, Jul. 2008. [12] Google, “Google green the big picture FAQs.” [13] Plugwise, “Getting started with Plugwise: Installation manual,” 2010. [14] Nokia, “Nokia fast charger AC-8 technical specifications.” [15] Wikipedia, “Free trade areas in Europe.” [16] Comscore, “Mobile year in review 2010,” Feb. 2011. [17] Nielsen, “Mobile snapshot: Smartphones now 28% of U.S. cellphone market,” Nov. 2010. [18] ——, “U.S. smartphone market: Who’s the most wanted?” Apr. 2011. [19] Gartner, “Gartner says Android to command nearly half of worldwide smartphone operating system market by year-end 2012,” Apr. 2011. [20] IDC, “IDC forecasts worldwide smartphone market to grow by nearly 50% in 2011,” Mar. 2011. [21] CTIA, “U.S. wireless quick facts.” [22] Eurostat, “Total population.” [23] ITU, “Mobile cellular subscriptions.” [24] U.S. Census Bureau, “Population estimates.” [25] Eurostat, “Electricity prices for second semester 2009,” 2010. [26] Covenant of Mayors, “Technical annex to the SEAP template instructions document: The emission factors.” [27] European Commission Joint Research Centre, “Electricity mix.” [28] U.S. Energy Information Administration, “Average retail price of electricity to ultimate consumers: Total by end-use sector.” [29] U.S. Environmental Protection Agency, “eGRID2010 version 1.1. year 2007 summary tables,” May 2011. [30] IDG News, “NTT DoCoMo market share drops under 50%,” Apr. 2008. [31] B. Lebot, A. Meier, and A. Anglade, “Global implications of standby power use,” Lawrence Berkeley National Laboratory, Tech. Rep., 2000. [32] U.S. Energy Information Administration, “International energy statistics.”