The construction of a truly global personal energy meter applicable to anyone, anywhere, would be an engineering challenge well beyond the scope of a dissertation such as this.
Instead, it addresses mainly the first two challenges set out in Section 1.3 of data collection and processing; since the HCI aspects and exact form in which feedback is presented makes a significant difference to its effectiveness (see Section 2.1) only example interfaces are developed. It focuses on the principles and concepts that are transferable from proofs of concept and case studies centred on the author and colleagues at the Computer Laboratory of the University of Cambridge. It identifies and draws together existing work which might help account for significant sources of energy consumption such as transport, and tackles only those areas which are lacking, with a particular focus on the apportionment of the energy costs of shared resources.
Chapter 2
Related work
Contents
2.1 The importance of feedback . . . 34 2.2 Persuasive technologies . . . 35 2.3 Metering electricity consumption . . . 42 2.4 Metering other forms of consumption . . . 50 2.5 Reducing deployment costs . . . 56 2.6 Occupancy detection . . . 60 2.7 Purpose-built location systems . . . 62 2.8 Opportunistic location systems . . . 70 2.9 Syndication of sensor data . . . 83 2.10 Summary . . . 84
Overview
This chapter provides a detailed survey of related work. The idea of a personal energy me-ter cuts across a broad range of research areas, from energy monitoring through location and identity sensing systems to human-computer interaction and social questions. This chapter first motivates and places in context the personal energy meter through a review of studies showing the effectiveness of feedback on reducing energy consumption (Sec-tion 2.1), then discusses existing technologies designed to promote behavioural change (Section 2.2). It surveys systems for measuring or inferring energy consumption that might provide useful input to a personal energy meter (Sections 2.3 and 2.4), highlights the problems with requirements for extensive additional infrastructure and reviews po-tential solutions in the form of user-deployed sensing or crowd-sourced data (Section 2.5).
Location will provide important context information to help apportion consumption, and 33
Sections 2.7 and 2.8 provide a detailed survey of existing systems, focussing on the es-sential properties for energy metering. Finally, Section 2.9 discusses potential methods for aggregating data from many disparate sensor systems which will be necessary for any heterogeneous personal energy meter. This chapter therefore identifies where further research is needed and guides the remainder of this dissertation.
2.1 The importance of feedback
Behavioural and environmental psychology studies have demonstrated many times the impact that feedback can have on encouraging people to reduce their energy consumption;
this provides strong motivation for the creation of a personal energy meter which provides continuous fine-grained feedback across all the aspects of a person’s life rather than on specific places or types of consumption. The first known study of ‘eco-feedback’ was in 1976, when Kohlenberg et al. found that even a light bulb which illuminated when households were within 90% of their peak energy levels changed energy usage behaviour, reducing time spent above a predetermined peak power level by up to 50% [121]. Real-time web-based feedback has been shown to produce an overall 32% reduction in electricity use by dormitory residents, high resolution feedback proving more effective than low resolution [167]. Feedback is not only useful in its own right, as a self-teaching tool, but it also improves the effectiveness of other information in achieving better understanding and control of energy use. This section highlights some of the more significant reviews to obtain aggregate measures of the effects of feedback.
In a review of 38 feedback studies carried out over a period of 25 years, Darby found typical energy savings of 10–15% and showed that improved feedback may reduce consumption by up to 20%; she concluded that “clear feedback is a necessary element in learning how to control fuel use more effectively over a long period of time” [37].
Fisher surveyed 26 projects from 1987 onward on the effects of feedback of electricity consumption and on consumers’ reactions, attitudes and wishes concerning such feed-back [53]. She found that typical energy savings were between 5% and 12% and that feedback is most successful when it “is based on actual consumption, given frequently and over a long time, provides an appliance-specific breakdown, is presented in a clear and appealing way and uses computerised and interactive tools.” Although only three of the studies reviewed used computerised feedback, these were the ones that resulted in the greatest change in consumption. She hypothesised that successful feedback has to draw a close link between specific actions and their effects; this is one of the core aims of the personal energy meter. Winett et al. also showed the importance of specificity, with more specific signs resulting in a 60% reduction in days lights were left on compared with more general feedback [204].
Abrahamse et al. compared feedback to a number of other intervention strategies, such as
goal-setting and information, through a review of 38 studies from social and environmental psychology [1]. They conclude that “feedback has proven its merits, particularly when given frequently” and that it can also increase the effectiveness of antecedent interventions.
They argue that an important first step in any intervention aimed at reducing energy is a ‘thorough problem diagnosis’ to identify the behaviours that significantly contribute to environmental problems; this is one of the key aims of a personal energy meter.
Froehlich et al. published a comprehensive survey of examples of what they call ‘eco-feedback technology’, including 89 papers from environmental psychology and a further 44 from ubiquitous computing and human-computer interaction literature [63]. They pointed out that although these two fields are closely related they have tended to remain wholly separate and argued that Computer Science researchers have not yet focussed on evaluating the potential strengths of their designs with respect to their ability to change behaviour.
Finally, Fitzpatrick and Smith reviewed the methodology of previous studies of feedback technologies and set out design concerns and questions that they hope will influence future work, focussing on how feedback should be presented [54]. They suggested that current work is too utility-centric and a more holistic view is necessary, imagining that “multiple resources could be monitored, not just electricity and gas, but also water, garbage, chem-icals, food, and so on; feedback on light use could be juxtaposed against occupancy and activity monitoring.” The personal energy meter attempts to take just such a view.