Content uploaded by Jan Krämer
Author content
All content in this area was uploaded by Jan Krämer on Feb 19, 2019
Content may be subject to copyright.
Towards a user-centered feedback design for smart meter
interfaces to support efficient energy-use choices: A design
science approach
Dr. Anders Dal´en
Viktoria Swedish ICT
Lindholmspiren 3A, 41756 G¨oteborg, Sweden
Tel.: +46 730 654 750
Fax: +46 31 772 8963
anders.dalen@viktoria.se
Prof. Dr. Jan Kr¨amer
University of Passau
Dr. Hans-Kapfinger-Str. 12, 94032 Passau, Germany
Tel.: +49 851 509 2580
Fax: +49 851 509 2582
jan.kraemer@uni-passau.de
September 2016
Teaser
Based on user feedback from a field experiment, we develop requirements and principles
for the design of smart meter interfaces. In particular, we propose to provide users
with personalized information at the appliance level on the monetary effect of a change
in usage behavior or the replacement of the appliance. Based on these requirements a
prototypical smart meter design is implemented to demonstrate the technical feasibility.
Abstract
Based on interviews of users’ experience with current smart-meter technologies
we propose, implement and evaluate a user-centered design of an energy-use informa-
tion system that assists private households in making efficient energy consumption
decisions. Instead of providing disaggregated data, the envisioned system automati-
cally calculates the monetary savings from replacing an appliance or by changing the
operational behavior of an appliance. The information provided is personalized with
respect to appliance use and also comprises information from external databases. A
prototype is implemented and evaluated in a use case with white goods household
appliances. The study concludes with directions for further interactivity improve-
ments and research into the structures of an openly shared appliance database.
Keywords: Smart metering, Design science, User-centered design, GreenIS
1 Introduction
The use of smart-meter technology greatly facilitates the collection and exchange of in-
formation about private households’ energy consumption. In principle, this information
could be used to make energy users aware of their household’s electricity usage and,
thereby, induce more sustainable energy consumption choices (Poortinga et al, 2003;
Abrahamse et al, 2005; Mattle et al, 2011; Han et al, 2013). In contrast to the indus-
trial sector, household energy usage continues to increase and comprises a large share
of the total energy use. For example, in 2010 German households made up 27.7% of
the domestic energy demand at 141 terawatt-hours (TWh) (AG Energiebilanzen, 2014).
For the EU, it is estimated that up to 27% of the housholds’ energy use can be saved
through more efficient energy use (European Commission, 2006). According to the Ger-
man Energy Agency, 50% of the electricity costs of German households are due to the
energy usage of white goods, such as washing machines, dryers, refrigerators, freezers
and dishwashers (Langgassner, 2001). Furthermore, in the past years, both electricity
use and the number of household appliances have increased almost constantly (Umwelt-
bundesamt, 2012). Thus, the household sector offers a large potential for improvements
in energy efficiency that is to date largely untapped.
Field studies have confirmed that providing direct feedback information on energy
use alone can potentially lead to savings of up to 15%-20% (Abrahamse et al, 2005,
2007; Dietz et al, 2009; Grønhøj and Thøgersen, 2011; Vassileva et al, 2013; Pullinger
et al, 2014; Lossin et al, 2016; Zhou and Yang, 2016). However, in practice unfamiliarity
with the provided technical information (Abrahamse et al, 2005), information overload,
(Loock et al, 2013), and lack of means to interpret current electricity consumption make
it difficult for the end user to turn the information into action (Fischer, 2008; Simmhan
et al, 2011; Darby, 2010). For example, a consumer survey issued by the German Federal
Ministry of the Environment regarding environmental awareness in Germany revealed
that 20 to 30% of the representative sample (n=2034) felt that a lack of transparency
prevented even larger changes, e.g., making more sustainable energy use choices (Kuckart
et al, 2006). Social economic factors, such as age, gender, education level and income,
showed no impact on the survey results, however the participants all agreed that the
consumer must save energy in everyday life. Moreover, recent field experiments sup-
plying access to energy-use information through web-based or dedicated displays have
shown that the realized energy savings are usually short lived as they disappear once
the feedback information has been discontinued (Hargreaves et al, 2013; Pullinger et al,
2014; Van Dam et al, 2010, 2012), and that the feedback did ultimately not lead to an
altered appliance usage (Thuvander et al, 2012; Hargreaves et al, 2013; Van Dam et al,
2010).
While generalized feedback has failed to evoke interest and make sense to energy
users, McMakin et al (2002, p. 860) found conclusive evidence that “an effective in-
tervention must be customized to the population and situation being targeted.” To
this end tailoring (providing customized energy-use feedback) and modelling (provid-
ing examples for altered behavior) as well as goal-setting are recommended means of
1
feedback intervention that have provided to improve upon generalized feedback in field
studies (Abrahamse et al, 2005; Benders et al, 2006; Abrahamse et al, 2007; Allcott and
Mullainathan, 2010; Vassileva et al, 2012; Loock et al, 2013). However, among others
(Fischer, 2008; Simmhan et al, 2011; Van Dam et al, 2012, e.g.,), Darby (2010, p. 455)
concluded in her review on smart-meter interfaces and energy-use information that the
provided feedback has yet to be further developed to provide more “appropriate forms
of feedback, narrative and support”. In particular, current smart-meter interfaces lack
the ability to influence behavioral energy-use practices by determining “how much en-
ergy each practice uses and derive options for change” (Pullinger et al, 2014, p. 1150).
This would require to measure “energy use per appliance and tailoring advice to the
households’ specific appliances” (Pullinger et al, 2014, p. 1150).
The objective of this study is to overcome this shortcoming of existing energy-use
feedback systems and to design an artifact that is able to provide private households
with tailored energy-use decision support on the impact of behavior change and appli-
ance exchange. In particular, under the design science research paradigm we follow a
user-centered approach to derive five requirements and eventually four distinct design
principles for a future energy-use feedback system. More precisely, we propose to pro-
vide users with continuous energy-use data at the appliance level (Design Principle 1),
which is then processed to automatically identify appliance states and settings (Design
Principle 2). This allows to determine a personalized usage profile of the appliances in
a household as well as the actual energy consumption of appliances, given the personal
usage profile. The internal information gathered in this manner can then be compared
to external information from appliances databases to provide decision support on behav-
ioral changes as well as the replacement of an appliance (Design Principle 3). Finally,
we suggest to provide decision support based on monetary savings or, for a long-term
perspective, derivatives thereof, such as amortisation periods (Design Principle 4).
After having identified and motivated the problem as well as our design objective, the
remainder of this article is structured as follows: Next, we discuss our methodological
approach in the context of design science research. Then, based on this approach, we
identify the solution objectives, i.e., the user-centered design requirements. Next, design
principles are derived that incorporate these requirements. Subsequently, the feasibility
of the energy feedback system based on these design principles is demonstrated in a
prototypical implementation. Finally, the artifact is evaluated and discussed, including
limitations and directions for future research.
2 Methodological Approach
To ensure a rigorous development of the design we adopt the design science research
(DSR) methodology, which is deemed suitable to “create and evaluate IT artifacts in-
tended to solve identified organizational problems” (Hevner et al, 2004, p.77). This
methodology framework assists research of design theory that is prescriptive, practical
and a basis for action (Baskerville and Pries-Heje, 2010). Additionally, the well-defined
structure of design science also strengthens the potential for cumulative development
2
Possible points of entry
Problem(
centered(
Objec/ve(
centered(
Design5(and(
Development(
centered(
Client(
centered(
Iden/fy(problem(
and(mo/vate(
Define(objec/ves(
of(solu/on(
Design(and(
development( Demonstra/on( Evalua/on( Communica/on(
Nominal process
Process iteration
Figure 1: Structure proposed for a design science methodology (Peffers et al, 2007),
highlighting the design- and development centered approach entry point and the nominal
process adhered to in this study.
of the artifact (Gregor and Jones, 2007). In particular, we follow the process model
developed by Peffers et al (2007, p. 64) and pursue what the authors call a “design- and
development-centered approach”. That is, we build on a previous routine design instan-
tiation of an energy-feedback artifact (a state-of-the-art smart metering device with web
interface) that is evaluated through a field experiment and from which requirements and
design principles for an improved artifact are derived. See Figure 1 for an overview of the
design science process pursued here. The entry point for this research is highlighted and
the nominal design science process shows the sequential structure for both, the proposed
DSR methodology and for this article.
After an evaluation of the challenges that the experimental participants have with
adopting and using the current energy-use information design, Wallenborn et al (2011)
recommend researchers to integrate the users in the design process. Embracing a user-
centered design has been suggested as critical for “the migration of electricity users to
the demand response world” (Honebein et al, 2009, p. 39). Therefore, the requirements
of the design artifact indentified in this study are based on the user experience of a
previous instantiation of the artifact. The six core principles of user-centered design
are that: 1) the design is based upon an explicit understanding of users, tasks and
environments, 2) users are involved throughout the process, 3) the design is driven and
refined by user-centered evaluation, 4) the process is iterative, 5) the design addresses
the whole user experience and 6) the design team includes multidisciplinary skills and
perspectives (ISO, 2010).
Design development is a continuous process and this study focuses on the first iter-
ative step. We will not focus on developing the user-centered design theory; instead, we
3
agree with Goes (2014, p.vi) that the contribution of DSR is not restricted to theory
and that research rigour can be “pursued in the methods employed in the development
of the artifact”. Thus, by following Peffers et al (2007)’s DSR methodology our con-
tribution is the establishment of design principles for the stated objective as well as
the development of a prototype. As we propose a new design to a known problem, our
design artifact qualifies as an “improvement” in the terminology of Gregor and Hevner
(2013). In particular, our design science research artifact differs from routine design as
we propose and envision the use of non-standard solutions, which are technically feasible
but, in combination, currently not readily available for routine design efforts.
3 Objectives of the Solution
Following Peffers et al (2007), the objectives or (meta) requirements (Walls et al, 1992;
Eekels and Roozenburg, 1991) are derived for the proposed solution artifact. Follow-
ing a user-centered approach, the design requirements are based on experiments and
qualitative interviews of households’ experiences with a routine design artifact, i.e., a
state-of-the-art smart metering device with a web interface that provides immediate
and historic feedback on the household’s energy consumption. In particular, the main
source of user input is based on our own semi-structured interview with 21 participants
who took part in a three-month field experiment with this type of frequent energy-use
information.
The interview questions covered the topics of energy awareness, flexible energy usage
and privacy. The questions were asked open-ended and in an order that came natural
during the interview to allow the users to freely explain their experience with the infor-
mation provided. The recorded interviews were coded by labeling specific phenomena in
accordance with grounded theory (Corbin and Strauss, 2008). To build theory around
the utility of energy-use information the phenomena were analyzed for common themes
(Urquhart et al, 2009). By writing and sorting memos of these diverse subjects seven
categories were established, which defined the participants’ processes in regard to the
energy-use information. These seven categories were i) general energy use practices, ii)
energy use awareness, iii) perceived energy use savings, iv) practices with energy use in-
formation, v) display limitations, vi) ability to shift energy use in time and vii) privacy
concerns.
What remains from the grounded theory process is to analyze the interview results
in light of related literature and experiments to find commonalities and differences that
can extend the knowledge base (Corbin and Strauss, 2008). In this study this amounts
to studying related experimental experiences with frequent energy-use information and
comparing our qualitative findings to them.
The first insight derived from the interviews was that users wish to learn or confirm
the levels of energy usage of different appliances. For example, one user in our experi-
ment explained his evaluation process where he turned off the electricity supply to all
rooms except one from the fusebox and then tested the appliances of interest in this
temporary laboratory. Furthermore, three of the 21 interviewed participants reported
4
that the power use information had been used to support decisions to exchange appli-
ances. Similar accounts of active user analysis of individual appliances were reported
by Hargreaves et al (2013) who interviewed 11 participants from the UK who got access
to energy-use information over 18 months through dedicated displays. Vassileva et al
(2012) and Schwartz et al (2013) also highlight the need for providing information at the
appliance level to improve energy awareness and knowledge based on experiences from
field and living-lab studies. Based on this evidence, we propose:
Requirement 1: The energy-use feedback system for private households should pro-
vide information about the energy usage of individual appliances.
The design artifact used in the field experiment provided energy-use data at a high
frequency (every 8 seconds). It was found that this enabled a range of different informa-
tion uses, such as determining the household base load, the influence of standby power
and remotely monitoring roommates. Previous experiments also confirm that continuous
provisioning of energy-use information is desirable (Abrahamse et al, 2005; Ehrhardt-
Martinez et al, 2010; Hopf et al, 2016; Nilsson et al, 2014; Zhou and Yang, 2016) and
have led to a heightened awareness and knowledge level among household users (Darby
et al, 2011; Hargreaves et al, 2013; Thuvander et al, 2012). Thus, we propose:
Requirement 2: Immediate, high frequency energy-use data is a fundamental require-
ment for an energy-use feedback system for private households as it improves awareness
and allows for individual explorations as well as more advanced processing of the energy-
use data.
Furthermore, the participants in our study were also interested in the impact of
changes in operating behavior of individual appliances. Nine participants reported that
they had tried to use certain appliances less often, for example, by only running the
washing machine when full or by cooking food collectively. Although the operating be-
haviors were often evaluated, many commented that the impacts of certain changes were
difficult to estimate when only aggregate energy-use data is provided. Hargreaves et al
(2013) similarly found that users would only go so far in changing their energy-use pat-
terns since the impact was considered to be negligible or outside the preferred comfort
zone. However, activity level energy-use information has been found to improve the
understanding of personal energy usage (Costanza et al, 2012). Correspondingly, many
participants were surprised to learn the significant power use of some of the stand-by
enabled appliances, which influenced them to disconnect these appliances from the elec-
tricity source when not in use. Disconnecting and gathering appliances on power outlet
strips were also found in the living-lab study by Schwartz et al (2013). Hargreaves et al
(2013) had a slightly different experience, as their users reported that the base energy
use was quickly understood as the norm, and it is not reported whether it prompted any
changes in appliance operation.
Behavior change has statistically been more important for improving efficiency than
what the rising cost of energy could accomplish in the same time (Frieden and Baker,
1983). Supporting the evaluation of the value of energy usage behavior is therefore
important (Grønhøj and Thøgersen, 2011; Pullinger et al, 2014). It has also been shown
that changing behaviors to improve efficiency is more successful for saving energy, both
5
in the short and long term, and is easier to implement on a large scale than curtailing
behaviors (Ritchie and McDougall, 1985; Benders et al, 2006). In summary, we identify
participants’ desire for an energy-use feedback system that facilitates the analysis of how
different appliance operating-modes impact the energy usage. Thus, we propose:
Requirement 3: An energy-use feedback system for private households should provide
information on the energy usage of different operating behaviors of individual appliances.
In our field experiment the feedback format was limited to current power load, mea-
sured in kW. In the interviews it became apparent that the participants mainly evaluated
and discussed their energy usage in monetary units. One user directly criticized the cur-
rent choice of interface unit as too technical. Another user went further and reported
that the costs reported on the energy bill had a greater motivating effect to save en-
ergy than the current experimental feedback. This argument is strengthened by Kamb
et al (1998), who found that a monetary incentive outperformed other units of equal
value. In contrast, altruistic feedback, directed at the goodwill of users, show little or no
effect on direct behavior change (McMakin et al, 2002; Ritchie and McDougall, 1985).
The importance of comprehensible feedback is also echoed by other experiments, where a
more personal language is recommended to design future energy-use information systems
(Thuvander et al, 2012; Schwartz et al, 2013). Thus, in an effort to present energy-use
information in a motivating and understandable way, we propose:
Requirement 4: An energy-use feedback system for private households should provide
feedback information based on non-technical units that are easily comprehensible to
private household users, such as monetary units.
However, just providing energy-use information through comprehensible units may
not be sufficient. For example, monetary units were criticized for being “unimpressive”
(Wallenborn et al, 2011, p. 151), because the small short-term gains that were displayed
with the live feedback did not motivate the users to change their energy usage (Wallen-
born et al, 2011; Grønhøj and Thøgersen, 2011). A related informational issue, which
was voiced in our interviews, was the uncertainty of the impact of exchanging an ap-
pliance or operating behavior for another. For example, users expressed uncertainty on
whether LED lights would be profitable over the current compact fluorescent or in how
many years a more efficient washing machine would pay back the investment over the
old one. This type of relevant decision support information cannot be satisfactorily an-
swered with unprocessed energy-use data, irrespective of the unit used. Similar accounts
were also evident in the study by Hargreaves et al (2013), where the energy-use infor-
mation also failed to provide a convincing argument of an action’s value over a longer
term. Generally, households seem to prefer technical energy-savings measures over be-
havioral measures (Poortinga et al, 2003), but especially the combination of feedback
on behavioral and appliance alternatives is considered to be promising and has largely
been untapped in previous design instantiations (Grønhøj and Thøgersen, 2011). Thus,
we propose:
Requirement 5: An energy-use feedback system for private households should provide
users with decision support with respect to exchanging individual appliances as well as
with respect to changing the operating behavior of individual appliances.
6
Finally, and for completeness, we note that our interviews also identified some contra-
dictory requirements with respect to display technology and visualization of energy-use
feedback, which precluded listing them as additional requirements. For example, partici-
pants were in disagreement whether they would prefer a dedicated display. Some argued
that this would better remind them of the current energy usage, while others saw a po-
tential risk of conflict if this information was always visible. Both the potential reminder
and the risk of conflict from using a dedicated display was also voiced by the partici-
pants in Wallenborn et al (2011) and Hargreaves et al (2013) studies, who used different
dedicated display technologies. On the one hand the information was backgrounded but
continued to remind users of their actions passively, however on the other hand, this
reminder was, in some cases, experienced as “nagging”. Similarly, participants had con-
trary views with respect to the use of social comparisons and competitions in energy-use
feedback. Although in some studies social comparisons are found to engage users and
help motivate energy savings (Petersen et al, 2007), others only report negligible results
(Abrahamse et al, 2005).
4 Design and Development
4.1 Suitability of Existing Energy Feedback Solutions
Routine solutions to provide private households with energy-use feedback information
range from media campaigns over labeling schemes to home audits and smart metering
devices (see Abrahamse et al, 2005, for a comparative review). However, it is evident
that the aforementioned requirements can only be fulfilled by a technical instantiation
of an information system, and thus we will focus on a presentation of feedback provided
by currently available instantiations of smart metering devices: A comprehensive list of
commercial and non-commercial designs are reviewed by Weiss (2010) and Pullinger et al
(2014). Most devices are generally able to provide feedback not only on the amount of
kWh used, but also in terms of costs, which is in line with Requirement 4. Furthermore,
Weiss (2010) distinguishes broadly between those devices that report the households
total energy consumption and those that report the energy consumption of individual
appliances be means of outlet sensors, which can also be combined to a mesh network
that allows to monitor several appliances at once. In both categories, there exist devices
that report energy use data at high frequency and are thus compatible with Require-
ment 2, but obviously only the devices in the latter category would be compatible with
Requirement 1. The device proposed by Weiss (2010) is even able to detect on-off-states
of individual appliances and thus provides functionality in the spirit of Requirement 3.
However, none of the devices reviewed by (Weiss, 2010), including his own design
effort, offers contextual feedback on behavior change or appliance alternatives, which
is demanded by Requirement 5. Similarly, Pullinger et al (2014) identifies only one
device in the UK market that is able to disaggregate total electricity use (fulfilling
Requirement 1,2,4, and arguably 3), but highlights that “the provision of practice-based
advice tailored to the household” (Pullinger et al, 2014, p.1151 and Table 1) is entirely
7
Figure 2: Schematic design of the proposed energy-use feedback system
absent, thus violating Requirement 5. This leads the authors to conclude that future
versions of smart meters should include “specific disaggregation algorithms, types of
feedback and criteria for tailoring their delivery to households based on their specific
[energy use] practices” (Pullinger et al, 2014, p.1159). This is very much in line with the
requirements and design that we propose here.
4.2 Design Principles
Hevner et al (2004) proposed that designing a solution can be thought of as a search
process. First the search will identify existing systems and related prototypes similar
to what has been presented in the preceding sections. The aim of this search process
is a set of design principles that will guide the development and ensure that all of the
identified requirements can be met. Our design is based on several sources of data,
which, through processing, are combined to advise private households about the current
appliance and operation behavior options. Figure 2 shows an overview of the planned
energy-use feedback system’s modules and information flows, which we will now present
in detail.
4.2.1 Load monitoring and state differentiation
The necessary appliance-level information can either be measured at the supply of the en-
ergy flow network by distributed sensors (distributed sensing) or by disaggregating load
data at a junction point (central sensing). When the central measurement technique
is successful, this approach could save the need for sensitizing single objects, however,
it must be noted that such central disaggregation techniques are currently not reliably
generalizable beyond laboratory environments (Liang et al, 2010). Since Requirement
3 specifically calls for behavior level information, reliable measurements and categoriza-
tions of individual appliances and their settings are necessary. This level of detail makes
us opt for outlet sensing in this iteration.
Design Principle 1: Collect continuous energy-use data at the appliance level through
electrical outlet sensors.
Furthermore, an automatic post-processing of the operational behavior data is pro-
posed in this study. In this vein, only an initial manual setup is necessary, after which
8
different settings will automatically be recognized for this particular appliance. The op-
eration mode disaggregation algorithm that was devised to produce this data is based on
a simple form of the Kirchhoffs law state model that was envisioned by Hart (1992). The
basic premise is that every appliance returns to the original state through a program of
a finite number of states. This model allowed us to analyze when a program finished
(i.e., the appliances power consumption went back to the starting point) and to compare
the duration of specific states of different programs to distinguish between them. To
demonstrate this post-processing we focused on the identification of on and off states of
the appliances. Specific signature sections of the different appliance programs were then
extracted and labeled manually. However, this part could also be automated in future
instantiations be means of machine learning (see, e.g. Zufferey et al, 2012; Hopf et al,
2016; Zhou and Yang, 2016). The subsequent matching process followed automatically
by comparing the duration of the signature sections. The process of disaggregating states
and appliance settings are more fully explained in the demonstration (Section 5).
Design Principle 2: Provide information on appliances states (e.g., on or off) and
settings (e.g., wash program) through automatic post-processing of the disaggregated
energy use data.
4.2.2 Integration of public appliance databases
An external source of data that provides information that is both useful for users and spe-
cific to each appliance are publicly available appliance databases (e.g., www.ecotopten.de).
These databases, which contain energy-use data for a wide range of appliances, are well
suited to perform general comparisons within and between appliances types. For exam-
ple, the appliance data supplied from these sources could be used in a purchase situation
to compare some legacy devices and new appliances available on the market. As users
change their appliances or behavior this can be reflected in the feedback.
Unfortunately, to date these databases lack the depth necessary for the state and
program matching that were explained above. Wiki-type projects for gathering and
sharing information and appliance data like the PowerPedia (Weiss et al, 2012) are
promising approaches in this regard and might eventually be the only feasible way of
creating a global database of appliances and their operating modes. By introducing such
a database, general appliance learning is reduced to the instance when the appliance is
first measured. In order to demonstrate the value of appliance and behavior change,
a prototype database with appliances that included the necessary appliance state and
operational mode signatures information was designed in this study.
Design Principle 3: The appliance exchange and behavior change information is
based on comparisons between internal information (appliance-level measurements) and
external information (appliance databases).
4.2.3 Using comprehensible units
In order to provide a causal relationship between actions and their effects on energy
efficiency, we propose to give decision support in the form of monetary savings. There
9
are several advantages to this approach. First, in line with Requirement 4, money is a
comprehensible unit and therefore has been reported to be accepted and well understood
by a wide range of users (Fitzpatrick and Smith, 2009). Monetary feedback can also
be presented as investments over the long-term. This alleviates the concern of only
negligible savings in the moment (Wallenborn et al, 2011). Moreover, it does not require
technical knowledge and thus, it will also help to reduce the cognitive burden on the
decision maker. Second, by using money as the feedback unit it is possible to compare
different types of appliances between each other. This comparison is, e.g., not possible
with the current form of energy labels as they are tied to a type of appliance. As the
overall energy saving is the main aim of exchanging a certain appliance, being able
to compare different appliance types would allow private households to exchange the
appliance with the highest potential effect, irrespective of the appliance type.
Design Principle 4: Provide energy-use information support in the form of monetary
savings, both in the short term due to behavior changes and in the long term due to
appliance exchange.
5 Demonstration
In this section, we describe a proof-of-concept artifact that exemplifies how the design
principles can be executed. It presents the outcome of the first iterative step in creating
an improved, user-centered energy-use feedback system. To this end, we focus on well
known household appliances to exemplify the data gathering, processing and potential
presentation approaches.
5.1 Setup
In order to allow for detailed appliance level information, in line with Design Principle 1,
a power outlet energy-logging device was used to gather data. The current and voltage
is calculated continuously at the relative high frequency of 1 Hz by the microcontroller
into real power. In accordance with Design Principle 2 the information post-processing
of operating modes, comparison and visualization was then handled by a custom-made
Java program.
Altogether, a washing machine, a dryer and two refrigerators were equipped with
the energy-logging device. These appliances were chosen based on their common occur-
rence and relative large energy usage in households. To demonstrate the use of external
information of alternative appliances, as predicated by Design Principle 3, related appli-
ance energy-use data for the comparison database was gathered from the publicly avail-
able appliance benchmark information portal EcoTopTen (http://www.ecotopten.de).
This website provides information on the most efficient household appliances currently
available in the German market, and provides a relevant comparison for the analyzed
appliances in this study.
10
Figure 3: Disaggregated states for a washing machine, showing the heating power surges
used for determining the washing cycle used. The area below the graph represents the
energy used during a certain time-span.
5.2 Processing of appliance and operating mode information
To provide the necessary data for calculating monetary savings and amortisation (De-
sign Principle 4), the logged energy-use data was processed for three parameters: (i)
the number of completed cycles Ncycle, (ii) type of operating mode Xand (iii) the ap-
pliance’s energy consumption per cycle Eappliance. By combining the number of cycles
and type of operating modes (e.g., 60◦C and 40◦C for a washing machine) with the
energy-use measurements, the energy usage of a certain operating behavior, Ebehavior ,
was determined. More specifically, the cycle counter identifies the return to the initial off
state after a minimum predetermined length of time as a finished program cycle. In the
case of the washing machine, a five Watt power use change from the initial steady-state
was used as the threshold to register the start and stop of the appliance. The cycle start
and stop naturally also framed the total amount of energy used (see Figure 3).
In the operating behavior analysis, the washing machine program load signature was
determined to be the initial heating period in the beginning of the program, which is the
initial energy spike period marked in Figure 3. This duration parameter could success-
fully sort all the 60◦C and 40◦C programs used in this study. For the demonstration of
feedback on behavioral changes we will focus on the monitored washing machine. There
are two reasons for this choice. First, washing machines motors and heating blocks have
a direct relation to the energy consumed. This is not necessarily true for all household
appliances. Refrigerators, for example, are dependent on the temperature setting, ambi-
ent temperature, frequency and duration of door openings and closings and the thermal
capacity of the content, which can only be read from the electricity usage indirectly.
Measuring in- and out-side temperatures could alleviate this specific problem but is out
of the scope of this design demonstration. Second, the operating modes of washing ma-
chines allow for a straightforward evaluation of behavioral changes. An example of a
possible alternative operating behavior was collected by reviewing research on washing
machine and detergent technology. A switch from 60◦C (140 F) to 40◦C (104 F) has,
11
for example, shown to have little impact on the resulting cleanliness of clothes but will
impact the energy used of about 125.58 kJ/kg water or about 230 Wh energy for ten
liters of water (R¨udenauer et al, 2006).1This result provides a host of possible oper-
ational changes that do not necessarily have an impact on the quality of the service
provided. Finally, washing machines are also often targeted by current energy efficiency
support campaigns and provide a good basis for evaluating the defined requirements for
this problem type in comparison to the existing energy feedback solutions.
5.3 Calculating Feedback
The parameters from the data processing were then combined to calculate the monetary
savings associates with appliance exchange and behavior change (Design Principle 4).
The known unit and amortisation requirements were catered to by processing the energy
use from a unit of service (e.g., a full washing cycle) in monetary terms to appreciate
the expected yearly gain.
5.3.1 Feedback on the impact of exchanging appliances
With respect to the replacement of an appliance, the annual savings potential (Mappliance
y)
can be calculated by the difference between the amount of energy for the current appli-
ance (Ecur) and the alternative appliance (Ealt ) in performing a specific unit of service
(e.g., one cycle or one hour of time), multiplied by the number of the units of service
per year (Ny) and the current price of electric energy (CkW h), as shown in Equation 1:
Mappliance
y=CkW h ·(Eappliance
cur −Eappliance
alt )·Ny(1)
Some appliances involve a greater investment and are predicted to be running for
several years. The simple payback method is the most common indicator for evaluating
the profitability of investments. Although it only provides a rough indication of the
financial prospect, it is an estimate of how long the money will be tied up in an investment
(White, 1993). Equation 2 details that the yearly amortisation (Ay) is the quotient from
dividing the purchase cost of the alternative (Cappliance
alt ) by the annual savings potential
of the alternative appliance (Mappliance
alt ).
Ay=Cappliance
alt
Mappliance
y
=Cappliance
alt
CkW h ·(Eappliance
cur −Eappliance
alt )·Ny
(2)
This information further improves the basis for exchanging an appliance without
demanding more sources of data to be collected. The problem formulation’s simplicity
also promotes a general understanding of the feedback given from the data.
Figure 4 demonstrates the feedback on the value of appliance exchange that was
calculated based on data gathered by our prototypical implementation. The figure shows
the different appliances tested and the monetary value in exchanging appliances for the
1Specific heat capacity of water (4186J/(kg ·deg) ) ·temperature difference (60deg −40deg = 20deg)
·water mass (10liter ≈10kg)·Wh per joule (1/3600(W h)/J ).
12
Figure 4: Results showing the value of exchanging a washing machine a dryer and a
refrigerator to a current top alternative of comparable size and program setting.
current, most efficient ones listed in the EcoTopTen-database and the amortization rate
according to the simple payback method. Energy prices (0.26e/kWh) where taken from
October 2012 from the German “Bundesverband der Energie- und Wasserwirtschaft”
and the estimated purchase prices of the appliances from the online retailer Amazon.
The operational live was assumed to be 12.2 years for the washing machine and dryer
and 14.6 years for the refrigerator (Gutberlet, 2008).
5.3.2 Feedback on the impact of behavioral changes
The monetary value of changing operating behavior within the same appliance (i.e., the
washing machine) was calculated for the current measured appliance and two newer
alternatives. The number of cycles was normalized to be comparable to the yearly
consumption base line of the analyzed appliances in the EcoTopTen database. The
calculation to evaluate the behavior change in terms of monetary savings is shown in
Equation 3. The annual savings of behavior change(Mbehavior
y) is calculated, similar to
the savings from appliance change (Equation 1), by multiplying the current electricity
price (CkW h) with the number of yearly cycles (Ny) and the change in electricity usage
due to the behavior change (Ebehavior
cur −Ebehavior
alt ). The factor Xis a factor to vary the
grade of operating behavior change between 0 and 100%. This variable was implemented
to allow an evaluation of partial behavior changes.
Mappliance
y=CkW h ·(Ebehavior
cur −Ebehavior
alt )·Ny·X(3)
Figure 5 demonstrates the feedback that results from our prototype.
13
Figure 5: Results showing the value of changing the washing machine operating behavior
from 60◦C (140 F) to 40◦C (104 F).
6 Evaluation and Discussion
6.1 Design Contribution
The objective of our design effort was to create a user-centered energy-use feedback
system to promote effective energy-use choices in private households. This objective is
pursued by involving users who have experience with energy-use feedback of a state-
of-the-art routine design artifact. Based on qualitative interviews, design requirements
were identified and operationalized through specific design principles following the DSR
methodology.
Our main design contribution is that we have proposed and implemented an energy-
use feedback information system for private households, which, in contrast to previous
IS energy feedback solutions (cf. Section 4.1), provides tailored decision support on the
impact of changes in behavioral practices or appliance exchange, i.e., which is based on
the actual energy usage behavior and energy consumption of the individual appliances in
a household. These new aspects of our design are codified in Design Principles 3 and 4.
Such tailored modelling feedback is deemed valuable, because it is more clearly under-
stood and may have the potential to yield sustainable changes in energy-use practices
towards more efficient behavior (Steg, 2008; Grønhøj and Thøgersen, 2011; Van Dam
et al, 2012; Pullinger et al, 2014).
6.2 Limitations and Future Research
The devised design principles for smart-meter interfaces are defined and implemented in
a prototypical artifact that lends itself well to continuous iterations. In particular, we
wish to highlight three implications for future design improvements that can build on
14
our design.
First, with respect to providing feedback on appliance exchange, this study has shown
the potential that can arise from implementing a shared database for appliance energy-
use data. Current publicly available databases support decisions between larger house-
hold appliances. However, it is still not possible to compare the current situation and
most often not even the currently owned appliances with newer ones. For example,
when comparing the expected operational life of common household appliances (Gut-
berlet, 2008) to the calculated payback time, it is immediately apparent that replacing
an appliance is, in many cases, not cost effective. Based on our demonstration, there is
only a marginal chance of getting a return on investment when exchanging the Medion
refrigerator or the Miele W 844 washing machine (see Figure 4). With a shared ap-
pliance database with appliance setting differentiation, current appliances operated in
specific ways can be used as the benchmark as was demonstrated in this study. The
appliance efficiency could then be followed over its operational life, in contrast to the
initial evaluation done today.
Although much of the necessary information can be gathered through several publicly
available sources, such an appliance database is currently not available. It requires a
community effort to accumulate the load profiles for various household appliances and
their associated cycles and states. However, once this information has been added, it will
benefit all other users with the same appliance. By uploading more usage parameters to
the database, the accuracy of the expected mean consumption of the appliances can be
improved which will extend the usability of the platform for the whole community. A
potential avenue for research would thus be to investigate how the development of such
a database can be established. This entails research on how appliance signatures can be
standardized as well as how households can be incentivized to provide this information.
Second, our demonstration has revealed that, in the context of white goods appli-
ances, a higher potential for savings is achievable by changing operating habits, as com-
pared the savings associated with exchanging appliances net of replacement costs. This
result confirms that being able to compare appliances based on the same (monetary)
unit is key for making effective decisions both in terms of energy and economic aspects.
Evidently, behavioral changes are generally more relevant for those appliances that are
less energy efficient. Thus, the effect of changing the behavior is directly dependent on
what appliance is currently in use.
Third, it is important to remember that, “feedback does not have to be complex to
be effective” (Darby, 2008, p. 506). There is a clear risk that the ability to add more
information to a system might finally make it more complex. Therefore, before imple-
menting the proof-of-concept design for another quantitative and qualitative evaluation,
the more fundamental concern of information overload should be evaluated. Due to the
potential informational richness of an energy information system based on Green IS,
research exploring how to balance the information for accurate and timely decisions is
becoming more important. A strict separation and measurement of information design
and interface design is necessary to build cumulative knowledge of how the adoptions
process is influenced (Bhattacherjee and Sanford, 2006). By combining the result from
15
how an appropriate informational load should be designed with results from energy-use
display design research (Anderson and White, 2009), another set of interface require-
ments can be tested in the next design iteration.
Evidently, our study also bears several limitations. In order to focus on a clear pre-
sentation of the conceptualization and design of the envisioned energy-use information
system, we deliberately chose a simple approach to conduct the underlying economic
evaluation of different appliances and behavioral alternatives. Obviously, several im-
provements are feasible here. For example, the payback period could incorporate an
appropriate discount factor, and possibly also a forecast on future energy costs. More-
over, it would also be feasible to provide monetary information on the outcomes that can
be achieved by replacing appliances and changing the usage behavior. In this context, it
might be worthwhile to extend the comparison engine in our system by a collaborative-
filtering based recommender system that could disseminate best practices of similar
households. It is important to note that the decision support provided only takes into
consideration the electricity used, while other efficiency improvements - such as less water
consumption for the washing machine - might also be an important reason to exchange
appliances.
Finally, it is emphasized that the feedback system proposed in this study does not
claim to be superior to all other forms of energy-use feedback. For example, the face-to-
face interaction of home audits is arguably more personal than other forms of communi-
cation, and mass-media’s ability to provide compelling and comprehensible explanations
can be more powerful for introducing new services. However, the proposed system’s
ability to cover a broad range of key user-centered requirements that go beyond the
current instances with the support of Green IS research is what could establish transfor-
mational power (Brocke et al, 2013). Naturally, the possibilities for innovation in Green
IS go well beyond only using the information proposed in this study. For example, in-
fluencing attitudes of dissolution, i.e., that each individual has a very small perceived
impact (Strengers, 2011), could be targeted by integrating information about the power
of many. Furthermore, normative beliefs could be targeted by providing comparisons
between similar user and appliance groups.
Acknowledgements
The authors would like to express their gratitude to Michael Schilling and Henning
Quellenberg for their research assistance during the development of this study.
References
Abrahamse W, Steg L, Vlek C, Rothengatter T (2005) A review of intervention stud-
ies aimed at household energy conservation. Journal of Environmental Psychology
25(3):273–291
Abrahamse W, Steg L, Vlek C, Rothengatter T (2007) The effect of tailored information,
16
goal setting, and tailored feedback on household energy use, energy-related behaviors,
and behavioral antecedents. Journal of Environmental Psychology 27(4):265–276
AG Energiebilanzen (2014) Auswertungstabellen zur Energiebilanz f¨ur
die Bundesrepublik Deutschland 1990 bis 2011. URL http://www.ag-
energiebilanzen.de/index.php?article id=10
Allcott H, Mullainathan S (2010) Behavior and energy policy. Science 327(5970):1204–
1205
Anderson W, White V (2009) Exploring consumer preferences for home energy display
functionality. Tech. rep., Centre for Sustainable Energy, Report to the Energy Savings
Trust
Baskerville R, Pries-Heje J (2010) Explanatory Design Theory. Business & Information
Systems Engineering 2(5):271–282, DOI 10.1007/s12599-010-0118-4
Benders RM, Kok R, Moll HC, Wiersma G, Noorman KJ (2006) New approaches for
household energy conservationin search of personal household energy budgets and
energy reduction options. Energy Policy 34(18):3612–3622
Bhattacherjee A, Sanford C (2006) Influence processes for information technology ac-
ceptance: an elaboration likelihood model. MIS Quarterly 30(4):805–825
Brocke J, Loos P, Seidel S, Watson RT (2013) Green IS. Business & Information Systems
Engineering 5(5):295–297, DOI 10.1007/s12599-013-0288-y
Corbin J, Strauss A (2008) Basics of qualitative research: Techniques and procedures
for developing grounded theory. Sage, Newbury Park, CA
Costanza E, Ramchurn SD, Jennings NR (2012) Understanding domestic energy con-
sumption through interactive visualisation. In: Proceedings of the 2012 ACM Con-
ference on Ubiquitous Computing - UbiComp ’12, ACM Press, New York, New York,
USA, p 216, DOI 10.1145/2370216.2370251
Darby S (2008) Energy feedback in buildings: improving the infrastructure
for demand reduction. Building Research & Information 36(5):499–508, DOI
10.1080/09613210802028428
Darby S (2010) Smart metering: what potential for householder engagement? Building
Research & Information 38(5):442–457
Darby S, Anderson W, White V (2011) Large-scale testing of new technology: some
lessons from the UK smart metering and feedback trials. In: European Council for an
Energy-Efficient Economy, Summer Study, pp 1–7
Dietz T, Gardner GT, Gilligan J, Stern PC, Vandenbergh MP (2009) Household actions
can provide a behavioral wedge to rapidly reduce US carbon emissions. Proceedings
of the National Academy of Sciences 106(44):18,452–18,456
17
Eekels J, Roozenburg N (1991) A methodological comparison of the structures of scien-
tific research and engineering design: their similarities and differences. Design Studies
12(4):197–203, DOI 10.1016/0142-694X(91)90031-Q
Ehrhardt-Martinez K, Donnelly KA, Laitner S (2010) Advanced metering initiatives
and residential feedback programs: A meta-review for household electricity-saving
opportunities. In: American Council for an Energy-Efficient Economy, Washington,
DC, US
European Commission (2006) Action plan for energy efficiency: Raising the potential.
COM(2006)545 final
Fischer C (2008) Feedback on household electricity consumption: a tool for saving en-
ergy? Energy Efficiency 1(1):79–104
Fitzpatrick G, Smith G (2009) Technology-Enabled Feedback on Domestic Energy Con-
sumption: Articulating a Set of Design Concerns. IEEE Pervasive Computing 8(1):37–
44, DOI 10.1109/MPRV.2009.17
Frieden BJ, Baker K (1983) The market needs help: The disappointing record of home
energy conservation. Journal of Policy Analysis and Management 2(3):432–448
Goes PB (2014) Editor’s comments: design science research in top
information systems journals. MIS Quarterly 38(1):iii–viii, URL
http://dl.acm.org/citation.cfm?id=2600518.2600519
Gregor S, Hevner AR (2013) Positioning and presenting design sci-
ence research for maximum impact. MIS Quarterly 37(2):337–356, URL
http://dl.acm.org/citation.cfm?id=2535658.2535660
Gregor S, Jones D (2007) The anatomy of a design theory. Journal of the Association
of Information Systems 8(5):312–335
Grønhøj A, Thøgersen J (2011) Feedback on household electricity consumption: learning
and social influence processes. International Journal of Consumer Studies 35(2):138–
145
Gutberlet KL (2008) Energieeffizienz im Haushalt. URL http://www.gfk-
verein.org/veranstaltungen/gfk-tagung-2008/vortrag-2, GFK-Tagung, Nuremberg,
4.7.2008
Han Q, Nieuwenhijsen I, de Vries B, Blokhuis E, Schaefer W (2013) Intervention strategy
to stimulate energy-saving behavior of local residents. Energy Policy 52:706–715
Hargreaves T, Nye M, Burgess J (2013) Keeping energy visible? Exploring how house-
holders interact with feedback from smart energy monitors in the longer term. Energy
Policy 52:126–134, DOI 10.1016/j.enpol.2012.03.027
18
Hart G (1992) Nonintrusive appliance load monitoring. Proceedings of the IEEE
80(12):1870–1891, DOI 10.1109/5.192069
Hevner AR, March ST, Park J, Ram S (2004) Design science in information systems
research. MIS Quarterly 28(1):75–105
Honebein PC, Cammarano RF, Donnelly KA (2009) Will smart meters ripen or rot?
Five first principles for embracing customers as co-creators of value. The Electricity
Journal 22(5):39–44
Hopf K, Sodenkamp M, Kozlovkiy I, Staake T (2016) Feature extraction and filtering
for household classification based on smart electricity meter data. Computer Science-
Research and Development 31(3):141–148
ISO (2010) ISO 9241-210:2010: Ergonomics of human-system interaction – Part 210:
Human-centred design for interactive systems. Tech. rep., International Organization
for Standardization, Geneva, Switzerland
Kamb ML, Rhodes F, Hoxworth T, Rogers J, Lentz A, Kent C, MacGowen R, Peterman
TA (1998) What about money? Effect of small monetary incentives on enrollment,
retention, and motivation to change behaviour in an HIV/STD prevention counselling
intervention. The Project RESPECT Study Group. Sexually transmitted infections
74(4):253–255
Kuckart U, R¨adiker S, Rheingans-Heintze A (2006) Umweltbewusstsein in Deutschland
2006 - Ergebnisse einer repr¨asentativen Bev¨olkerungsumfrage. Tech. rep., Bundesmin-
isterium f¨ur Umwelt, Naturschutz und Reaktorsicherheit (BMU)
Langgassner W (2001) Energieeffizienz elektrischer Antriebe in Haushaltsger¨aten. E und
M, Energie-und-Management-Verl.-Ges, Herrsching, Germany, IFE Schriftenreihe
Liang J, Ng SKK, Kendall G, Cheng JWM (2010) Load Signature StudyPart II: Dis-
aggregation Framework, Simulation, and Applications. IEEE Transactions on Power
Delivery 25(2):561–569, DOI 10.1109/TPWRD.2009.2033800
Loock CM, Staake T, Thiesse F (2013) Motivating energy-efficient behavior with green
is: An investigation of goal setting and the role of defaults. MIS Quarterly 37(4):1313–
1332
Lossin F, Loder A, Staake T (2016) Energy informatics for behavioral change. Computer
Science - Research and Development 31(3):149–155, DOI 10.1007/s00450-014-0295-3
Mattle P, Aigner M, Schmautzer E, Fickert L (2011) Smart Plug - Konzept f¨ur ein intel-
ligentes Lastmanagementsystem f¨ur Einzelverbraucher. In: Energieversorgung 2011:
M¨arkte um des Marktes Willen?, TU Wien, Institut f¨ur Energiesysteme und Elek-
trische Antriebe, Vienna, pp 96–97, 7. Internationale Energiewirtschaftstagung, 16.-
18. Februar 2011
19
McMakin AH, Malone EL, Lundgren RE (2002) Motivating residents to conserve energy
without financial incentives. Environment and Behavior 34(6):848–863
Nilsson A, Bergstad CJ, Thuvander L, Andersson D, Andersson K, Meiling P (2014)
Effects of continuous feedback on households electricity consumption: Potentials and
barriers. Applied Energy 122:17–23
Peffers K, Tuunanen T, Rothenberger MA, Chatterjee S (2007) A design science research
methodology for information systems research. Journal of Management Information
Systems 24(3):45–77
Petersen JE, Shunturov V, Janda K, Platt G, Weinberger K (2007) Dormitory resi-
dents reduce electricity consumption when exposed to real-time visual feedback and
incentives. International Journal of Sustainability in Higher Education 8(1):16–33,
DOI 10.1108/14676370710717562
Poortinga W, Steg L, Vlek C, Wiersma G (2003) Household preferences for energy-saving
measures: A conjoint analysis. Journal of Economic Psychology 24(1):49–64
Pullinger M, Lovell H, Webb J (2014) Influencing household energy practices: a critical
review of uk smart metering standards and commercial feedback devices. Technology
Analysis & Strategic Management 26(10):1144–1162
Ritchie JRB, McDougall GHG (1985) Designing and marketing consumer energy con-
servation policies and programs: Implications from a decade of research. Journal of
Public Policy & Marketing 4(1):14–32
R¨udenauer I, Eberle U, Grieß hammer R (2006) ¨
Okobilanz und Lebenszykluskosten-
rechnung W¨aschewaschen. Tech. rep., ¨
Oko-Institut e.V., Freiburg/Hamburg, URL
http://www.oeko.de/oekodoc/289/2006-008-de.pdf
Schwartz T, Denef S, Stevens G, Ramirez L, Wulf V (2013) Cultivating energy lit-
eracy. In: Proceedings of the SIGCHI Conference on Human Factors in Comput-
ing Systems - CHI ’13, ACM Press, New York, New York, USA, p 1193, DOI
10.1145/2470654.2466154
Simmhan Y, Aman S, Cao B, Giakkoupis M, Kumbhare A, Zhou Q, Paul D, Fern C,
Sharma A, Prasanna V (2011) An informatics approach to demand response optimiza-
tion in smart grids. NATURAL GAS 31:60
Steg L (2008) Promoting household energy conservation. Energy Policy 36(12):4449–
4453
Strengers Y (2011) Designing eco-feedback systems for everyday life. In: Proceedings of
the 2011 annual conference on Human factors in computing systems - CHI ’11, ACM
Press, New York, NY, US, pp 2135–2144, DOI 10.1145/1978942.1979252
20
Thuvander L, Meiling P, Andersson K (2012) Energivisualisering via display: F¨or¨andras
beteendet n¨ar hyresg¨asterna har m¨ojlighet att f¨olja sin elf¨orbrukning? Tech. rep.,
Chalmers University of Technology, G¨oteborg, Sweden
Umweltbundesamt (2012) Ausstattung privater Haushalte mit langlebigen Ge-
brauchsg¨utern. Tech. rep., Daten zur Umwelt, URL http://www.umweltbundesamt-
daten-zur-umwelt.de/umweltdaten/public/theme.do?nodeIdent=3535, last accessed:
03/15/2013
Urquhart C, Lehmann H, Myers MD (2009) Putting the theory back into grounded
theory: guidelines for grounded theory studies in information systems. Information
Systems Journal 20(4):357–381, DOI 10.1111/j.1365-2575.2009.00328.x
Van Dam S, Bakker C, Van Hal J (2010) Home energy monitors: impact over the
medium-term. Building Research & Information 38(5):458–469
Van Dam S, Bakker C, Van Hal J (2012) Insights into the design, use and implementation
of home energy management systems. Journal of Design Research 10(1-2):86–101
Vassileva I, Odlare M, Wallin F, Dahlquist E (2012) The impact of consumers feedback
preferences on domestic electricity consumption. Applied Energy 93:575–582
Vassileva I, Dahlquist E, Wallin F, Campillo J (2013) Energy consumption feedback
devices impact evaluation on domestic energy use. Applied energy 106:314–320
Wallenborn G, Orsini M, Vanhaverbeke J (2011) Household appropriation of elec-
tricity monitors. International Journal of Consumer Studies 35(2):146–152, DOI
10.1111/j.1470-6431.2010.00985.x
Walls JG, Widmeyer GR, El Sawy OA (1992) Building an Information System De-
sign Theory for Vigilant EIS. Information Systems Research 3(1):36–59, DOI
10.1287/isre.3.1.36
Weiss M (2010) emeter: Stromverbrauchsfeedback auf basis eines pervasive en-
ergy monitoring systems, Doktoranden-Workshop Energieinformatik 2010. Available
at https://www.vs.inf.ethz.ch/publ/papers/weismark-emeter-2010.pdf. Last accessed:
16.7.2016
Weiss M, Staake T, Mattern F, Fleisch E (2012) PowerPedia: changing energy usage
with the help of a community-based smartphone application. Personal and Ubiquitous
Computing 16(6):655–664
White AL (1993) Accounting for pollution prevention. EPA Journal 19(3):23–25
Zhou K, Yang S (2016) Understanding household energy consumption behavior: The
contribution of energy big data analytics. Renewable and Sustainable Energy Reviews
56:810–819
21
Zufferey D, Gisler C, Khaled OA, Hennebert J (2012) Machine learning approaches
for electric appliance classification. In: 11th International Conference on Informa-
tion Science, Signal Processing and their Applications (ISSPA), IEEE, 2-5 July 2012,
Montreal, QC, pp 740–745
22
