Ian H. Rowlands, Tobi Reid and Paul Parker
University of Waterloo, Faculty of Environment, Waterloo, Ontario, Canada
Changes in electricity systems mean that more detailed information about demand levels for particular energy services in the home are now available to energy researchers. Accordingly, it is useful to determine how these data might be best used by energy researchers. To advance this discussion, 13 studies that use intrusive load‐monitoring techniques to generate, to present, and to make effective use of, disaggregated end‐use electricity data from households are identified. These studies are placed within a broader literature context (including studies using nonintrusive load‐monitoring techniques), are summarized briefly, and are cross‐compared in order to delineate emerging issues. These issues are as follows: methodological challenges, including monitoring equipment performance and participant recruitment; ways to present the time‐and space‐specific nature of the end‐use electricity data generated; advances with respect to end‐use electricity models that can be built; appliance‐specific insights; and future priorities for this kind of work, including energy conservation insights, relevant policy recommendations, and priority academic investigations. Finally, reflection upon these 13 studies, as well as the broader energy research agenda, generates a number of priority areas for work going forward: making effective use of additional data; broadening the focus to include electricity production and storage, as well as other energy, carbon, resource, and information flows; placing these data within broader social contexts and wider power system considerations; and encouraging effective use of these data to advance energy system sustainability, at both the household and community levels.
Changes in electricity systems mean that more detailed information about end uses are increasingly available to energy researchers. This chapter directs attention to one part of that broader set of transformations – specifically, it focuses upon fine‐resolution, disaggregated end‐use electricity data in households. The development of a variety of advanced sub‐metering technologies (e.g. smart electrical panels and smart plugs) has facilitated the collection of more electricity end‐use information. With this detailed information now more readily available (e.g. how much electricity was used by a refrigerator at different times of the day in a particular household), a whole new set of energy research questions can be investigated.
The purpose of this chapter is to determine the impact of the availability of disaggregated (divided according to time‐of‐use) end‐use electricity information from households upon energy research. The investigation unfolds in four parts. First, the area of study is elaborated both by identifying our focus as well as other contiguous areas of research – together, this forms the broader context. Second, key research articles that use fine‐resolution (i.e. high frequency), disaggregated end‐use electricity data in households are identified and briefly summarized. Third, these articles are compared with each other in order to delineate emerging themes. Fourth, reflection upon these themes and the potential to address policy debates is undertaken in order to sketch an agenda for energy researchers going forward.
It is critical that recent advances in information and communication technologies (which enable the collection of these data) are accompanied by discussion and reflection among energy researchers more broadly. This will allow these new data to be analyzed efficiently for decision makers who are working to make progress toward a sustainable energy future.
This investigation is timely because of the presence of three key trends.
First, there have been significant advances in information and communication technologies. Not only are devices able to retrieve and to present more electricity end‐use data in a faster manner, but they are able to perform to these higher standards at lower cost[1]. Greater attention to consumers' needs has also contributed to significant improvements in the ways in which data are presented – they are often now much more user‐friendly[2].
Second, electricity systems are being reorganized, presenting opportunities for more widespread participation. The traditional monopolistic structure has, in many locations, been transformed: additional players, with innovative applications and diverse interests, are now able to contribute to the improved performance of electricity systems[3]. This diversity of players is serving to create a broader community of electricity system stakeholders, with alternative management and social structures being proposed and investigated[4].
And third, levels of local engagement in many parts of the world have risen. Be it driven by frustration with international processes or disillusionment with central government politics more broadly, there is an increased sentiment that, to achieve shared objectives, communities, working locally and working collectively, have to be more involved[5]. The rise of electricity cooperatives in many parts of the world is one manifestation of this trend[6].
Together, these three trends have given rise to the following: that there are more electricity end‐use data technically obtainable; that, in turn, these data are accessible to those who are interested in analyzing them; and that, finally, there exists a demand for relevant, actionable information by communities. In sum, conditions are ripe to understand better how disaggregated, end‐use electricity information should be best used. Indeed, there appears to be great potential benefit, at both the level of the individual and the level of the system.
For individuals, current levels of electricity literacy are generally low[7]. People, however, want more information with respect to how they are using electricity in their homes, not only with respect to particular energy‐consuming devices but also with respect to sets of energy services. To illustrate, consider the following example: not only is information with regard to the television's end‐use electricity demand useful[8–10], but so too are data about the living room's or entertainment center's overall power[11] requirements.
A body of evidence is accumulating to suggest that more detailed electricity information provision can empower individuals to take “better” electricity‐use decisions[12, 13]. This feedback may, in itself, or with other enabling technologies, lead to actions that result in lower operating costs or increased levels of comfort[14–16]. Alternatively, enhanced tools may allow individuals to perform their own “what‐if” investigations, to determine what actions to take given their own priorities (Ref.[1, 17])
At the system level, disaggregated electricity data about individual households could help utilities better understand their customer base by identifying critical locations instead of relying upon representations of “the average household” within their service territory[18]. With this, the cost‐effectiveness of their conservation programs – given that they could better target households with personalized conservation and demand management messages – would likely increase[19]. Community‐wide benefits would include lower prices, higher reliability, and potentially other positives such as lower emissions. By engaging households more actively, a more constructive partnership between utility and customer could result[20], with greater appreciation for the fact that electricity demand is wrapped up in social practice[21].
These individual‐level and system‐level benefits are, of course, not guaranteed. Nor should we ignore the fact that potential drawbacks also exist. Concerns regarding invasions of privacy[22], information overload[23, 24], and high capital costs (Ref.[7]) exist and warrant attention. While these issues deserve to be investigated and addressed appropriately, consideration of how best to use fine‐resolution, disaggregated end‐use electricity data should continue nevertheless.
Our investigation searched for studies that used fine‐resolution, disaggregated end‐use electricity data in households in order to advance understanding of energy use in some way. More specifically, we looked for studies that focused on the monitoring, presentation, and analysis of measured electricity consumption from different end uses in households. In the literature, this is often referred to as either direct submetering[25] or intrusive load monitoring[26]. We used different methods to locate these studies: Scopus and Google Scholar searches combining the terms “disaggregated,” “appliances,” and/or “end use” with various combinations of “electricity,” “residential,” and “household”; broader web searches of the same sets of terms; and snowball strategies following references in sources we secured as well as oft‐cited reviews in the literature (e.g. Ref.[12]). We did not restrict our investigation to any particular part of the world, nor did we limit our search to peer‐reviewed research articles; gray literature was also probed. In the section Studies Under Investigation, we introduce and review the studies we selected.
To place our focus within its broader context, it is useful to review briefly contiguous areas of research. Those areas of research are broadly captured by the umbrella term, “nonintrusive load monitoring” – i.e. capturing end‐use data without measuring, directly, the particular end uses. A variety of nonintrusive load‐monitoring techniques have been used, drawing upon electricity data that record the household's whole‐house consumption values, aggregated data from multiple households, household sociodemographic data, property characteristics, or any combination of these datasets. While nonintrusive load‐monitoring techniques are attractive to researchers because they are usually quite inexpensive and they are relatively easy to pursue, their main drawback can be with respect to the quality of the data they generate[27]. Nevertheless, a brief review is useful.
Top‐down studies[28, 29] take jurisdictional‐level electricity consumption data and seek to derive individual end‐use estimates by considering a range of energy, economic and social indicators. Some may focus upon “econometric data,” while others direct attention to “technological data” (Ref.[28]). Still others may combine elements of both and use, for instance, aggregated electricity consumption values, demographic variables, and appliance penetration information[30].
By contrast, bottom‐up studies (Ref.[28]) use individual household‐level data from a representative set[29]. For example, Bedir et al. collected questionnaire data on appliance usage and energy consumption from 323 households in the Netherlands[31]. The researchers used this information to build three different regression models based on direct (e.g. number of appliances) and indirect (e.g. resident presence/occupancy) determinants; regression models of this kind have been built by a variety of analysts.
Conditional demand analysis is another bottom‐up approach. It relies on survey data regarding, as a minimum, total household energy‐use and appliance ownership (Ref.[25]). Newsham and Donnelly's review of such studies finds that there is no “standard set of variables used as predictors in the regression equation,” for each has its own unique approach (Ref.[25]). Their investigation using data from 9773 Canadian households builds, in particular, upon Aydinalp‐Koksal and Ugursal's earlier study[32].
In contrast to the statistical or statistical/regression approaches[29], the so‐called engineering estimates aim to isolate electronic signatures of particular end uses. More specifically, one sensor is deployed to collect aggregate power consumption signals for the household as a whole. Fluctuations in these signals are then analyzed in order to determine what end uses, within the home, are deployed at what times. Efforts to improve both the training process for appliance signature recognition and the quality of the results are ongoing[33, 34].
These studies are making important contributions to energy studies, proving particularly valuable in identifying trends over time and spatial anomalies across jurisdictions. Complementary studies reinforced by datasets of representative households within jurisdictions [for instance, Digital Home Energy Management System (Dehems)[7]] will enable analysts to promote energy sustainability in the near future. However, as technology continues to become more fungible, more reliable and more cost‐effective, the availability of widespread and comprehensive disaggregated electricity information for households will grow. Increasing the understanding of how those data should be best put to use will similarly become critical. These studies are examined next.
Our search for studies that collected and presented disaggregated electricity information in households to answer particular research questions yielded 13 studies. Studies that appeared to collect and analyze these kind of data, but that did not lay out their data‐collection methods and analysis techniques – let alone their results – to the extent of those below were not included in this review[23, 35–37]. Of course, some of these studies, as more of their results are published, may be appropriate to include in future reviews of this kind. Similarly, studies that report upon the ways in which disaggregated electricity data are used as part of broader “feedback” strategies – but that do not provide details of the actual end‐use consumption patterns – were not part of this investigation[38, 39]. While we do not review them here, we recognize that this is nevertheless an important area for continued research[40].
Table 31.1 provides much of the basic information about these studies. Notable here is that almost half of these studies (six) took place in Europe, with another three in the United States. In addition, sample size varied from 9 to 1300. The monitoring period was as short as 2 weeks or as long as just over 2 years, and studies were conducted at various times during the past 30 years.
Table 31.1 Basic information regarding studies under investigation.
Study | Overarching project | Location | Recruitment method | Number of participants | Length of monitoring period | Dates of monitoring period |
Bladh and Krantz[41] | Larger, ongoing monitoring study undertaken by the Swedish Energy Agency | Sweden | Different number of occupants; different age groups; different levels of consumption. “This assured us a certain variation in the types of households we studied,” (p. 3523) | 69 households (large sample); seven households (small sample) | 2 wk | 11–23 October 2005 |
Coleman et al.[42] | N/A | United Kingdom | “‘Snowball’ sampling strategy” (p. 62); only those households with a relatively “typical” range of appliances selected (p. 62) | 14 households | 2 wk | March 2008 to August 2009 |
de Almeida et al.[43, 44] | REMODECE; European Residential Electricity Consumption Database | 12 EU countries | Effort to secure sample representativeness, but not always realized (p. 1885) | 1300 (approximately 100 in each country) | 2 wk | 2006–2008 |
Hart and de Dear[45] | Residential Energy Study (RES) | Sydney, Australia | Aim to secure “typical households in [New South Wales]” on the basis of broad demographic and socioeconomic characteristics (P. 162) | 136 households | A few days to 18 mo | 1993–1994 |
Isaacs et al.[46] | Household Energy End‐use Study | New Zealand | Population‐weighted sample framework across different climate regions | 399 households | At least 11 months for hot water; less than 11 mo for some individually monitored end uses | 1999–2005 |
Parker[47] | Unnamed large utility company | Central Florida, USA | “The homes represent a statistically‐drawn sample …” (P‐863) | 169 households | 1 yr | 1999 |
Pratt et al.[48, 49] | End‐Use Load and Consumer Assessment Program (ELCAP) Residential Base Study | Five states (Idaho, Montana, Oregon, Washington, Wyoming), USA | Attempted to replicate regional characteristics across five variables: climate zone, utility type, house age, household income, and presence of wood‐heating equipment; while claiming to be “roughly representative” (p. 180), there were still some “self‐selection biases” (p. 181) | 288 households | Unclear | October 1983 to May 1989 |
Puckett et al.[50] | Grid Smart Appliance Demonstration Project | Barren County, Kentucky, USA | “Due to the numerous requirements established and self‐selection, the 20 households are viewed as a large ‘focus group’,” (p. 1880) | 20 households | 5–6.5 mo | March 2012 to September 2012 |
Saldanha and Beausoleil‐Morrison[51] | N/A | Ottawa, Canada | “The houses were selected to span the range of consumption patterns typical of Canadian houses,” (p. 520) | 12 households | 12–18 mo | April 2009 to September 2010 |
Sidler[52, 53] | EURECO project | Four European countries (Denmark, Greece, Italy, Portugal) | No particular selection criteria because “it was difficult to find 100 voluntary households in each country” (p. 2.168) | 397 households | 1 mo | Between January 2000 and June 2001 |
Sidler et al.[54] | ECUEL project | Central France | “Expected to give results within approximately ±5% of the national average” (p. 1.293) | 98 households | 1 mo | January to July 1998 |
Ueno et al.[55] | N/A | Kyoto, Japan | Unclear | Nine households | Just over 2 yr | February 2000 to March 2002 |
Zimmermann[56] | Swedish Energy Agency: “Improved Energy Statistics in Buildings and Industry” project | Sweden | Efforts at representativeness: “a good picture of the different type of households present in Sweden” (p. 8) | 400 households | 1 month (360 households); 1 yr (40 households) | August 2005 to December 2008 |
Table 31.2 provides some more detailed information about the monitoring of the end uses. A variety of equipment was used, alternatively collecting consumption data from the end‐use itself (by means of some kind of data‐logger or plug‐in load monitor), from the household's electricity panel or from some combination of the two. Monitoring periods – the time between readings – varied from 1 to 30 minutes, with 10 minutes being the most commonly used time interval. Finally, end uses monitored ranged in number from one to dozens, with particular regional priorities often driving the selection (e.g. lighting in Sweden[41], heating in New Zealand,[46] and rice cookers in Japan[55]).
Table 31.2 Monitoring information regarding studies under investigation.
Study | Equipment used | Time interval for monitoring | End‐uses monitored |
Bladh and Krantz[41] | Watt/10 min | 10 min | Lighting |
Coleman et al.[42] | Plug‐in meters | 5 min | Consumer electronics and information and communication technologies equipment |
de Almeida et al.[43, 44] | Serial data‐loggers, Watt meters, standby energy monitors, lamp meter loggers | 10 min | Differed across countries, but main categories were “domestic computers and peripherals, new domestic entertainment, other standby loads and additional loads perceived to be changing fast, such as lighting and air conditioning” (p. 1885) |
Hart and de Dear[45] | Channel data‐loggers | 30 min | Room air conditioners, room heaters, refrigerators, stand‐alone freezers, and domestic hot water systems |
Isaacs et al.(final report)[46] | Data‐loggers | 10 min | Hot water was the main end‐use monitored; additional main categories were entertainment, heating and cooling, large miscellaneous, lighting, refrigeration, other climate control, other cooking, small miscellaneous |
Parker[47] | Incoming electrical service and end‐use metering | 15 min | Space heating, cooling, domestic hot water and either pool, dryer, or range |
Pratt et al.[48, 49] | Data‐logger | 1 h | Order of priority: heat, hot water, air conditioning, oven, refrigerator, dryer, lights and conveniences, special major appliances, freezer, washing machine, dishwasher; eight per household, on average |
Puckett et al.[50] | Key appliances on dedicated circuits | 15 min | Refrigerator, dishwasher, clothes washer, clothes dryer, range/oven, and water heater |
Saldanha and Beausoleil‐Morrison[51] | Monitoring equipment on electrical panel | 1 min | Air conditioning and furnace; additionally on a subset of households: range, dryer, dishwasher, and hot water heater |
Sidler[52, 53] | Enertech Lamp‐Meters and meters plugged in series with the appliance | 10 min | In all households: cold appliances, light sources, audiovisual site, washing machine; in some households: circulation pump of individual boilers, computer site, dishwasher, air conditioner, water heater |
Sidler et al.[54] | DIACE monitoring system | 10 min | 32 electrical appliances |
Ueno et al.[55] | Load‐survey meter and end‐use meter | 30 min | Multiple end uses, including space heating, television sets, and refrigerators |
Zimmermann[56] | Lampmeters, Wattmeters (wall sockets), from the main panel | 10 min | Heating, water heater, oven, washing machine, dishwasher, dryer, TV, audiovisual site, computer site, cold appliances, microwave, car heater, ventilation system, lighting |
Researchers are careful to consider the limitations of their conclusions, arising from the challenges associated with completing monitoring investigations of this kind. This recognition focused around participant recruitment methods and the performance of the monitoring technology.
Across the studies, efforts were usually made to secure a sample that was representative of the broader population (usually taken to be households in the jurisdiction under consideration). This was sometimes driven by socioeconomic characteristics[45], consumption characteristics[51], or some combination of the two[41]. While this was often the intention, the goal of a comprehensive and representative sample was never achieved. Even those with relatively considered reflection upon their recruitment methods found that there were ultimately still some biases in the recruitment protocol they followed[48].
Electricity end‐use consumption datasets were often identified as inaccurate and/or incomplete. This was sometimes a function of “anticipated problems” – for instance, the challenges associated with measuring particularly low levels of electricity consumption[42, 51, 55] or the implications of having multiple end uses on one monitored channel[48]. Other times, it was the result of “unanticipated problems” – in other words, the equipment did not capture the readings as planned. Hart and de Dear, for example, report that “Of the appliances monitored only 8% had complete records without any apparent errors or missing half‐hourly readings” (Ref.[45]). The cause of such failures may be technological [e.g. “equipment malfunctions” (Ref.[51])]; alternatively, nontechnological factors can play a role. With respect to the latter, Isaacs et al., for instance, report that a few loggers were “melted or drowned,” and that “one set of monitoring equipment was taken over by a cockroach infestation” (Ref.[46]). Using real‐world communities as “living laboratories” (as these 13 studies did) means that these kinds of situations are often encountered.
All of the studies selected present data regarding end‐use consumption. By virtue of our selection criteria, each presents information about the energy demand for at least one particular appliance (or other energy service) over the course of some period of time. Beyond this, however, there are variations with respect to how the data are presented. Most intriguing are the differing timescales upon which data are presented – that is, investigating whether particular times are more‐ or less‐intensive, in terms of relative electricity consumption levels for a particular end use.
In these studies, this issue is usually explored by means of a simple “24‐hour reflection” in a graph: the x‐axis is one full day (usually in terms of an “average” day), and the quantity plotted along the y‐axis is the hourly electricity consumption for that particular end use. The ways in which television use peaks in the evening (Ref.[43]), as does demand for the dryer (Ref.[50]), are examples of this kind of presentation. Indeed, 12 of the 13 studies reviewed in this chapter present data in this way (the exception is Isaacs et al.[46]).
Seasonality is the next most‐common temporal method of presentation. While studies that have monitoring periods less than one year are not able to investigate differences across four seasons (thus excluding six of the studies), two of the remaining seven studies provide a single visual across a full year in one graph (thus revealing “time‐of‐year” impacts) (Refs.[48, 56]). The other studies provide information about seasonal variations in different ways: for instance, Parker calculates daily averages for different months and then puts them together on one graph (Ref.[47]), while Hart and deDear use outdoor temperature as a proxy for time‐of‐year (Ref.[45]). In both instances, the relative importance of air conditioning in different seasons (critical in the summer) is a key insight provided. Other electricity end uses also show seasonal variations: lower consumption for the electric range/oven during the summer months in Florida, for instance, may reflect a greater propensity to eat out, to barbeque or to opt for a different (cooler) diet during those months (Ref.[47]; see also, Ref.[54]); the seasonality effects regarding dishwasher use in Midwestern US states may follow from the same kinds of household patterns during the summer (Ref.[48]).
The only other temporal approaches uncovered consisted of differences between “workdays” and “weekends/holidays,” and this was undertaken by two of the studies[50, 56]. The relatively “flat” end‐use figures associated with refrigerators found across all days of the week (Ref.[50]) contrasts with the relatively more intensive use of washing machines on weekends (Ref.[56]).
When hourly data are presented in these studies, they are usually “average” values – not only averaged across all days in the month or year but also averaged across all participants in the particular study. While such average values are the norm, it is still worthwhile noting that a number of studies explicitly note the range of values within their sample. Zimmermann, for instance, provides a number of results per household while also superimposing mean values on top: as an example, all refrigerators' mean electricity consumption value of 218 kWh per year is set against a range, for individual households, of just over 100 kWh yr−1 to close to 1200 kWh yr−1 (Ref.[56]). Moreover, Saldhana and Beausoleil‐Morrison find differences in air conditioning consumption values of almost 50‐fold in absolute terms, and that is across only 12 households (Ref.[51]). [Even in terms of air conditioning's relative contribution to overall household electricity consumption, they find a difference that is 25‐fold (15% for one house, 0.6% for another) (Ref.[51]).] Pratt et al. provide box plots to reveal not only median values and the entire range but also the 25th and the 75th percentile values (Ref.[48]). Puckett et al. give details across all participants: not surprisingly, refrigerators showed more modest variability across the 20 households, when compared with dishwashers and dryers (Ref.[50]).
In addition to variations across time (either days or months) and variations across participants, studies also present variations across space. De Almeida et al.'s investigation across 12 European countries allows them to present a striking graph revealing the importance of lighting in northern European countries (particularly Norway) and of air conditioning in southern European countries (particularly Greece) (Ref.[43]). Meanwhile, Isaacs et al. examine different regions within one country (New Zealand)[46]. Temperature difference is an important distinguishing factor between the country's North Island and its South Island.
Many studies build electricity consumption models, striving to explain either whole‐house or individual end‐use consumption levels. In some instances, additional information has been collected and subsequently used. Saldanha and Beausoleil‐Morrison, for instance, look at the relationship between livable floor area and, in turn, furnace circuit electricity consumption and air conditioner electricity consumption. They find a noteworthy (and positive) relationship for the former, though it still had “some significant scatter”; they do not find a significant relationship for the latter (Ref.[51]). Parker examines a variety of physical characteristics of the property and finds that the presence of, among other things, higher levels of ceiling insulation and light colored roofs, tile roofs, and shading lowered air‐conditioning energy and demand (Ref.[47]). Still others, as already noted, look at the relationship between outdoor temperature and air conditioning consumption values[45, 47].
Others investigate the predictive strength of the household's sociodemographic characteristics. Bladh and Krantz, for instance, find that as the number of people in a household rises, per capita lighting consumption falls; they also find that older people use less lighting than younger people (Ref.[41]). Pratt et al. also investigate a couple of potential explanatory factors, focusing upon lights and convenience loads: “[This] shows several expected trends. There is a strong relationship with number of occupants and house size, and to a lesser extent with income.” (Ref.[48]) Sidler et al. also focus on lights, but ask “where” in the household the lighting load is highest: their discovery is that it is in the lounge‐living room‐day room areas (Ref.[52]). In an earlier study, Sidler et al. find another “where” explanatory factor: “… that simply by keeping a cold appliance in a nonheated store room rather than a kitchen, average annual energy savings of 36% were achieved” (Ref.[54]).
As study authors reflect upon what their load monitoring revealed, they often note whether particular end uses are surprisingly low or surprisingly high.[Their expectations are formed by other data sources available, including internationally accepted estimates or modeled results that do not use intrusive load‐monitoring data (Refs.[48, 54]).] In more than one study, kitchen cooking appliances (alternatively called ovens, ranges, and/or stoves) are found to be surprisingly low consumers of electricity, relatively (Refs.[47, 48]). Sidler also notes that “refrigerators and freezers seem to be[on] average more energy efficient” (Ref.[52]). Alternatively, dryers (Refs.[50, 54]), electric kettles (Ref.[54]), and pool pumps (Ref.[47]) are singled out for their surprisingly high contributions to total household electricity consumption.
What receives considerable attention in a number of studies are entertainment devices, information, and communication technologies and standby loads (for instance, Ref.[52]). Coleman et al.'s investigation, for instance, concludes that information, communication, and entertainment (ICE) appliances significantly influence overall levels of household electricity consumption. According to their research, ICE appliances account for approximately 23% of total electricity use (38.3 kWh on average over a two‐week period). Much of this electricity consumption is found to have occurred in standby power modes: 30% of ICE power consumption and 7% of whole‐house consumption come from standby power draws[42].
Finally, a couple of studies use the results to investigate appliance performance – more specifically, electricity consumption per appliance cycle. Across all the studied households that have clothes dryers, for instance, Zimmermann finds that “80% of the [dryer] cycles consume less than 2000 Wh, 20% less than 500 Wh and only 7% more than 3 kWh” (Ref.[56]). In addition, Sidler et al. compare washing cycle electricity consumption across various appliance vintages – however, they do not find much difference by age (Ref.[52]).
A number of studies cast their eyes to the future, making suggestions as to what could be done with the insights generated by their investigations as well as what further research should be undertaken.
The potential for energy conservation is determined by a number of studies. To do this, the difference between the electricity consumption found in the particular study and the electricity consumption of the then‐currently available “best available technology” is calculated. This is done for particular end uses – for instance, Zimmermann finds that table‐top freezers hold the greatest potential for energy savings among cold appliances; if all existing ones were replaced with high‐efficiency models (consuming only 127 kWh yr−1), then average savings would be 362 kWh yr−1 (Ref.[56]). De Alemeida et al., furthermore, found that “the aggregate savings from switching from present state to Best Available Technology [across all appliances] were estimated to be about 1300 kWh/year for an average … household” (Ref.[43]). Others also make appliance‐specific recommendations[41, 54].
A number of studies make policy recommendations. For example, Coleman et al. highlight the importance of “better product design,” arguing that their results justify the implementation of minimum energy performance standards (Ref.[42]). De Almeida et al. complement their call for “demanding minimum efficiency standards” with an emphasis upon awareness‐raising campaigns (Ref.[43]). They also argue that the use of energy labels – informing potential purchasers of energy efficiency performance – could be enforced more systematically and expanded to include more products. Moreover, financial incentives could be used to catalyze increased deployment of energy‐efficient technologies in the home (Ref.[43]).
Further research priorities are also suggested in different studies. Hart and de Dear, for instance, have specific priorities as they work to refine their model: they would like demographic information regarding the occupants, additional data concerning the appliances being investigated (e.g. their type, age, and location) and more details regarding the property and its surroundings (e.g. shading characteristics and solar radiation data more broadly) (Ref.[45]). De Almeida et al. encourage researchers working in this area to be aware of changing parameters, including both technological developments (e.g. light‐emitting diodes, LEDs) and social trends (e.g. occupants choosing larger‐sized appliances) (Ref.[43]). Finally, Coleman et al. call for additional data regarding occupant behavior patterns so that electricity consumption could be curtailed in the short term (whereas the turnover of appliances may be a longer‐term strategy) (Ref.[42]). As they note, “Energy monitoring accurately details patterns of electricity use, but to convert these data into more useful information there is a need to gain insights into the behavior of the people causing the consumption” (Ref.[42]).
With continuing advances in technology, the prospects for even greater access to disaggregated household electricity data in the future look bright. In this section, therefore, we turn our attention to what should be on energy researchers' agendas as they move forward, ensuring that their activities are in step with technological progress. To construct this agenda, we reflect upon both what has been included in work to date (and reviewed above), as well as that which has yet to be explored.
The aforementioned advances in technology mean that more households will be able to be monitored for end‐use electricity consumption more easily.1Consequently, larger sample sizes will be within reach. Closer attention, therefore, should be paid to recruitment strategies. While eager “technophiles” have usually provided the vast majority of volunteers for these kinds of investigations, there will soon be the opportunity for stratified samples that better represent wider populations. Accordingly, resultant insights and lessons will overcome many of the recruitment challenges we found in the studies reviewed and will thus have more policy relevance. Literature that examines social experiments generally[57], and residential energy studies in particular[58], offer useful guidance in this regard. Attention to privacy considerations, in light of global trends, should also be part of design considerations[59].
This trend will also allow for more end uses within individual households to be monitored. As a number of our studies have shown, the “other” category – the generic name that is applied to all of those electricity end uses that are not submetered – can be large (in terms of the share of total household consumption)[43, 50]. Consequently, insights that may have been offered by the discovery of unexpectedly large loads could have been buried within these aggregate numbers. With more end uses available to monitor, ex ante choices about “significant loads” will not have to be made. Consequently, a more complete picture may be gleaned.
With parallel advances in on‐site power generation and energy storage technologies, studies of the kind investigated in this chapter will, in the future, need to have a wider swathe: more specifically, they will need not only to monitor and to report electricity consumption data, but they will also have to integrate these data with fine‐resolution, production, and storage data (which are similarly disaggregated by device). Using solar‐photovoltaic and wind turbine generation data, Lund et al. provide insight into the kind of investigation that will be more frequently forthcoming[60].
In the future, to optimize more fully on‐site management of power supply and demand, detailed information about all aspects of the electricity supply chain will be vital. Broader discussions about so‐called “DR 2.0” (the next generation of demand response strategies) – and the more general emergence of “prosumers” (those who both produce and consume electricity) – gives further impetus to this consideration[61]. Ways in which appliance‐specific DR “offers” may facilitate system‐wide goals (e.g. lights will be dimmed when prices surpass a certain point, but the entertainment unit will not be cut‐out until prices are twice that level) could be catalyzed by greater understanding of individual end uses.
In this spirit of increasing the scope of consideration, focus should extend beyond solely electricity supply chains and systems in the home to include all kinds of energy supply chains and systems therein. While this, for many homes in industrialized countries, would turn attention to natural gas use, extensive investigation would also consider other liquid and solid fuels and energy carriers (e.g. kerosene, gasoline, and wood pellets). Given the potential substitutability across fuels with respect to some energy services in the home (e.g. a plug‐in electric heater when compared with a basement furnace powered by natural gas or a rooftop solar‐thermal system), it would appear important to provide detailed information – and to conduct detailed analyzes – across all means of providing energy services. The literature on the so‐called “rebound effect” – i.e. savings in one area may be offset by increased consumption in another area – would suggest that this is critical (Ref.[7]). Indeed, one of our studies – namely the one in Japan – already foreshadowed these kinds of impacts:
However, the amount of increase in city gas and kerosene consumptions for space heating after the installation of an [electricity consumption information system] was unknown, and it is therefore desirable to display simultaneously not only electric power but also other energy sources such as city gas when supplying energy information.
(Ref.[60])
The scope can broaden further to include other resources (e.g. water), as well as other services in the home (e.g. entertainment, health) and even the internet of things (Ref.[1, 62, 63].)
Collection of these data would appear to offer outstanding opportunities to understand better how householders make use of energy services within different social contexts. For one, it would allow more detailed consideration of practices within households – for instance, the cultural and social elements associated with the desire for, and the use of, energy services in the home. The result would be a more appropriately balanced analysis of the ways in which advanced energy technologies to monitor and to control energy service provision in the home are used by occupants. More effective engagement – and more movement toward a sustainable energy future – could result[20].
More detailed investigation into the impact of energy service demand upon society more broadly, and how this information can be “fed back” to households also becomes possible. More specifically, the extent to which particular end uses put pressure on the reliability of the energy system (by, for instance, making a disproportionate share of their contribution during peak demand periods) would be one important insight to have. This would allow system operators to target these particular end uses for conservation and load‐shifting messages. The extent to which particular end uses are disproportionately used during periods in which the grid is most carbon‐intensive would be another important learning. Those advancing decarbonization goals would similarly know where their efforts should be placed.
Finally, it is critical to continue to think about how this information can be acted upon so that movement toward a sustainable future can be encouraged. Potential benefits for both the individual householder and the responsible system operator have been identified in this chapter. Here, we summarize them and add others.
With more detailed information about their electricity consumption (and potentially electricity production and storage, as well as flows of other household resources), householders will be able to better meet their goals. Those may include cost minimization, profit maximization, carbon minimization, environmental impact reduction, improved reliability, or any combination of these. Fine‐resolution, real‐time data – perhaps analyzed by computer applications, with decisions acted on by in‐home information and communication technologies – have the potential to identify appliance faults, to motivate curtailment of discretionary end uses during peak carbon periods and to encourage participation in energy arbitrage markets. Cross‐comparison of insights from the kinds of studies investigated in this review and contributions from feedback investigations that use disaggregated electricity information (Ref.[55] is one such example) could help in this regard.
The detailed information provided to system operators – and policy makers more generally – would allow them to target messages to their customers in a systematic manner. The potential benefits of appliance replacement campaigns, for instance, could be better anticipated. More generally, utility communications would soon come to be known to “matter” and consequently receive greater attention from householders. For planning purposes, a variety of “what‐ifs” could be completed, to anticipate better outcomes of particular policy decisions (and nondecisions) (Ref.[25]). Finally, to advance the aforementioned DR 2.0 world – particularly one with a lot of variable wind and solar power generation – detailed understanding of appliance use would be critical.
The purpose of this chapter was to investigate the ways in which increased availability of disaggregated (divided according to end use, as well as according to time of use) end‐use electricity information from households has affected, and will continue to affect, energy research. After reviewing the area of investigation with a relatively wide lens, the focus turned to those studies that have collected their data by various means of intrusive load monitoring. In all, 13 studies were identified and briefly summarized. They were analyzed to identify three key themes in this emerging literature. As a first theme, outstanding methodological challenges in their execution – with an emphasis on the quality of the monitoring equipment and the difficulties in participant recruitment – were highlighted.
In a second theme, a number of key areas were noted: the time‐ and space‐specific nature of the consumption information provided (and the insights they could provide), the end‐use consumption models that can thus be built, and the appliance‐specific insights (particularly low‐users or high‐users, as well as defective operation).
Finally, we discussed – in a third theme – the areas for future research suggested in the 13 studies. They are the potential for energy conservation identified by the data acquired in these studies, the policy recommendations that flow from the insights, and the further areas for academic research that are highlighted.
Indeed, the final section of this chapter builds upon these forward‐looking insights by sketching priority areas for work going forward. These include the following:
Developments in technology, energy markets, and community engagement have the potential to create a path toward energy sustainability. It is thus incumbent upon energy researchers to work to ensure that movement along that path occurs.
This chapter was completed under the auspices of the Energy Hub Management System project, which is supported by the Ontario Centres of Excellence, Hydro One Networks Incorporated, Milton Hydro Distribution Incorporated, and Energent Incorporated. The authors also received contributions from Merih Aydinalp‐Koksal, Alina Rehkopf, and Julia Shulist; valuable work on related topics by graduate students and other members of the Sustainable Energy Policy group at the University of Waterloo also helped to inform thinking in this area. The authors are grateful for this support; they, however, remain solely responsible for the contents of this chapter.
18.117.196.184