Exploring Implications of Capacity-Based Electricity Pricing for Peak Demand Reduction Dan Schien¶∗ , Colin Nolden†k , Caroline Bird‡∗ , Denise Wilkins§∗∗ , Kopo M. Ramokapane††∗ , x xi, ORCiD: 0000-0001-6293-3445, ∗ Chris Preist‡‡∗ , Phani Chitti ∗ , Ruzanna Chitchyan ∗ Department of Computer Science, k School of Law ∗∗ Life and Environmental Sciences University of Bristol, UK University of Exeter, UK Email: { ¶ daniel.schien, † colin.nolden, ‡ caroline.bird, †† marvin.ramokapane, ‡‡ chris.preist, x xi sc18092, r.chitchyan } @bristol.ac.uk, § d.j.wilkins@exeter.ac.uk, Abstract—In the face of climate change, the UK has set increasing flexibility, defined as “modifying generation and/or concrete goals to decarbonise energy systems. Associated strate- consumption patterns in reaction to an external signal (such gies include increasing electrification of residential heating and as a change in price) to provide a service within the energy transport and the substitution of fossil fuel energy sources with renewables. These entail expensive infrastructure reinforcements system”, are being considered [5], [7]. to support increased peak loads. Energy demand management In the context of UK electricity networks, large elec- can help mitigate this anticipated load increase. Capacity-based tricity consumers have been charged by capacity for some pricing is a mechanism to incentivise energy demand manage- time [8]. Residential households, on the other hand, are ment by charging for a maximum power draw as opposed to the volume of consumed energy. In this text we use a persona typically charged via a model that combines standing charges technique to study how user needs drive the typical energy (measured per day) and volume-based charges in proportion services that compose the aggregate power draw in a family to consumption (measured per kWh). Capacity-based pricing home. We use this to reflect on user feedback (obtained through has the potential to incentivise consumers to be more flexible a demand shifting workshop with 10 participants) to discuss the in their service consumption and reduce or defer consumption factors that would allow excessive power demand to be shifted away from peak demand times, thus reducing the maximum at peak times. In this text, we study for the first time the draw. Subsequently, the role of ICT as an enabler of energy implications of such a capacity-based alternative charging demand flexibility is discussed which provides the basis for a model for residential households from the ICT4S perspective. quantitative evaluation of demand flexibility scenarios.We find In this paper we explore the following questions: that household power draw is dominated by a few services that require significant changes to increase flexibility. 1) What energy services constitute peak time electricity demand for residential households? I. I NTRODUCTION 2) What energy services currently offer flexibility to ac- Global energy demand has grown steadily at around commodate capacity constraints? 2.4%/year (± 0.08%) since 1850. In 2017, it grew by 2.1% 3) How can ICT augment energy services to increase after 5 years of 0.9% average growth, with over 70% derived flexibility or substitution potential? from fossil fuels. Energy-related carbon dioxide emissions In order to investigate the implications of a capacity-based increased concurrently by 1.4% [1]. Scenarios by the UKs pricing approach and identify new strategies that help defer National Grid suggest that annual electricity demand will peak time electricity consumption, we have drawn on the increase from 297TWh in 2017 to 373-441TWh in 2050. notion of personas [9]. In our context this involved the creation Peak demand is expected to increase accordingly from 59GW of a representative UK household undertaking its activities in 2017 to 79-87GW in 2050 [2]. In the face of climate during a normal peak-demand time. We use this personification change, the UK, like many other countries, has set targets to identify energy consuming activities and services, and to decarbonise the energy system [3]. These plans require the model the energy use profile of our representative household substitutions of fossil-fuel-based heating and transport systems and explore implications of demand shifting on its power with renewable electricity-based systems. profile. We then integrate the feedback from an energy demand One of the challenges associated with the growing need shifting workshop carried out with 10 participants to discuss for (clean) electricity is the need for significant reinforce- how energy consumption at peak demand time can be deferred ment of the electricity grid to accommodate this projected and reduced. Following from this we discuss current barriers increase in unregulated demand, particularly peak demand at to greater flexibility and identify strategies to overcome these times of heavy usage (typically between 5-9pm) [4]–[6]. To barriers based on ICT innovations and policy interventions, avoid costly additional capacity, alternative approaches such as and consider the social implications. II. BACKGROUND load control (DLC) of appliances by third parties [13], [14]. Increasing penetration of variable renewable energy sources Research suggest that households switching their use of white places increasing strain on energy systems. Aside from the good appliances such as washing machines, tumble dryers total levelized cost of electricity, their integration into en- and dishwashers from evening peak to non-peak periods can ergy systems also imposes integration costs upon specific transfer at least 8% of peak demand (on average 57W per market participants. These costs depend on the extent to household) [15]. which demand-side and supply-side flexibility can counteract Such DSR depends on the nature of the signal used to variable and uncertain renewable energy generation. The most influence consumption patterns, including price signals (e.g. cost-efficient approaches to flexibility are technologies (and cost-effective or dynamic pricing), volume signals (e.g., load behaviours) that provide flexibility as a by-product [10]. capping) and direct signals (e.g., DLC of appliances) [13], Capacity peaks can be reduced through positive flexibil- [16]. Cost-reflective or dynamic pricing includes time-based ity (ramping up generation and/or releasing stored energy) pricing [17] critical pricing [18] and real time pricing [19]. while excess generation by variable renewable energy sources Dutta and Mitra [20] differentiate between between the fol- require negative flexibility (ramping down generation and/or lowing electricity pricing policies: storing energy). Although the increasing penetration of vari- • Flat tariffs able renewable energy sources increases market opportunities • Block rate tariffs for both ‘positive’ and ‘negative’ flexibility assets, market • Seasonal tariffs uncertainty in the UK limits the viability both of storage and • Time-of-use (TOU) tariffs additional generation capacity. At the same time, increasing • Superpeak TOU specialization and diversification in the UK’s flexibility market • Critical peak pricing (CPP) is actually decreasing margins. This constellation lends itself • Variable peak pricing (VPP) to the exploration of alternative pricing models for energy • Real-time pricing (RTP) demand. • Peak time rebates (PTR) Innovative approaches to incentivise DSR combine pricing A. Volume and Capacity-based Pricing signals with volume signals through abovementioned capacity- Infrastructure system services - of which the energy system based electricity pricing [21]. is one - can be charged for in a variety of ways. Usage- Capacity-based pricing is expected to flatten load curves based charges provide an important mechanism to incentivise by encouraging both optimised load scheduling (i.e. shifting the consumption of services in certain ways. Usage-based consumption) as well as total energy demand reduction [16], charge models include volume and/or capacity-based models [22]. and combinations of both. In principle, volume based charges Electricity pricing is based on complex models that aim to increase by the cost per consumed unit of a finite resource. balance the need to fund the cost of the network effectively, For example, residential water services are frequently charged guarantee supply, and encourage sustainability while keeping by volume of consumed water. the cost for consumers low [23]. For example, the charge Capacity-based pricing is motivated by the cost of providing model for extra high voltage (EHV) customers combines infrastructure via a “big enough pipe” to satisfy demand. To standing charges (per day), capacity charges (per kVA per use another example, wired internet access networks (DSL, day), exceeded capacity charge (per kVA per day) and volume Cable, etc) are typically priced based on maximum speed based peak time penalty charges (per kWh) for the consump- (i.e. data transfer capacity) they can provide. Here, the cost tion as well as generation of electricity. Notably, the charge for the service use does not increase in proportion with the model for EHV customers does not include a charge that is amount of data consumed per month. On the other hand, proportional to the amount of electricity consumed outside of in the same domain of internet access networks, charges for peak times [8]. cellular mobile network data are typically based on transferred Generally, capacity-based pricing is based on demand (mea- volume of data. Consumers have adapted to those alternative sured in kilowatt (kW) to represent power) instead of con- pricing approaches through a variety of strategies, depending, sumption (measured kilowatt hours (kWh) to represent en- among others, on the price elasticity of broadband [11]. ergy). It requires households to pay for the maximum amount Such strategies to substitute use of mobile broadband include of electric power they draw from the electricity grid at any proactively downloading media content onto the phone before moment, as opposed to how much electric power they use leaving home, or deferring consumption of content [12]. over the billing period. Capacity-based pricing involves the With domestic energy demand expected to grow as a result allocation of a particular load ceiling (load capping) on which of increasing electrification of heating and transport, capacity- the monthly charge is based. If the household exceeds the based pricing charges might play a role in energy demand ceiling, it will be moved to a higher band of service at an reduction. Compared to increases in generation and storage increased charge [16], [21]. capacities, domestic demand side response (DSR) is cheap, To date, most studies on innovative pricing models have quick and easy to implement. DSR can also enhance the analysed differential pricing such as time-of-use or Economy reliability of electricity systems through pricing and direct 7 [13], while studies on automated DSR have focused on 2 dynamic scheduling algorithms, such as the scheduling of 30%. They also observed that those who were most concerned distributed energy resources [24], price-volume signals to af- about energy prices were least keen on DSM technologies fect consumption patterns by integrating personal preferences whilst they were still willing to think about and reduce energy and technical information and constraints [25], and predicting use themselves. The study suggests that this might reflect household behaviour to optimise appliance control schedules concerns over sharing data and feelings of powerlessness and (Moratori et al., 2013). More recent studies combine customer vulnerabile to exploitation. preferences with automated demand scheduling (Rahseed et Burchell et. al. [32] contrast the 2 approaches to changing al., 2016) and algorithms to optimise scheduling of demand- behaviours - described as literacy and know-how and char- side appliance management [22]. acterised respectively by factual knowledge and reasoning or Few studies, however, focus specifically on capacity-based by practical skills and experience. The former is more easily pricing and user comfort. Similar to Agnetis et al. [25], scaled up and tends to be used by policy makers but achieves we hypothesise a household energy consumption optimisation less than a know-how approach which, whilst more resource problem derived from a prototypical consumer scenario. Our intensive, recognises that personal and social contexts are focus, however, is on specific load caps and specific remote important. The challenge lies in finding ways of combining demand-side interventions [13], [16]. these approaches. Variable capacity-based pricing relies on capacity caps to determine costs for individual households, B. Understanding energy behaviours, social practices and and the extent to which consumers can respond to these caps socio-technical systems will determine overall costs. It might be argued that consumers In order to explore capacity-based pricing and its impli- for whom overall cost is a less significant driver are less cations for consumers, it is necessary to understand energy likely to respond whilst those at lower income levels will usage and behaviours. A number of studies have attempted be forced to change their consumption patterns in order to to categorise energy consumption behaviours and contexts to maintain or reduce overall costs, thus placing the burden of assess importance and flexibility. There are two different areas response unfairly on lower income households. The issue of to explore here, one is overall consumption and the other load justice in the context of energy is an emerging and significant profiles - the more relevant area for our study. area for exploration [33] and needs to be aligned alongside For example, Boomsma et al. [26] categorise consumption the fast-changing application of technological approaches to by contexts: morning, evening, regular, important, most en- addressing energy demand saving and shifting. ergy consuming, summer or winter. Jones et al. [27], in an extensive literature review, identified 62 factors that affect III. M ETHODOLOGY AND S TUDY D ESIGN energy consumption of households. These factors were related A. Methodology Overview to three key areas: socio-economic (e.g., number of occupants, presence of teenagers, higher disposable income), building- related (e.g., number of rooms, floor area, electric space and water heating, air conditioning) and appliance-related (e.g., Persona number of appliances, rate of use of appliances, desktop PC, Workshop electric cooking, tumble dryer). Kavousian et al. in their US- Household based study [28] came up with similar areas with the addition of external conditions (weather / location). Whilst the factors above are important, energy use is not an end in itself but rather the result of ‘a complex series of interlinked and interacting socio-economic, dwelling and appliance related factors.’ [27]. Shove and Walker [29] suggest Capacity that ‘energy supply and demand are realized through artefacts Flexibility and infrastructures that constitute and that are in turn woven Scenarios into bundles and complexes of social practice’ and thus a part of the ongoing transformation of society. It is important that Fig. 1. Overview of Study Methodology we approach our study with an understanding of this range of practices and factors affecting energy usage. In Fig. 1 we present the four elements of the overall methodology used in this explorative study. This study started C. Socio-economic implications of variable pricing with a user co-design workshop where 10 user-researchers A substantial study of over 2400 households in 2012 (Dem- got together in a workshop to discuss demand response and ski et al. [30] and Spence et al. [31]) revealed variations in identify what issues they consider relevant for energy demand the acceptability of different demand-management scenarios: shifting away from peak time. The workshop started with for example switching appliances off standby after a certain the participants reflecting on their individual behaviours while time was acceptable to 78% of respondents but a fridge-freezer using various appliances, and noting factors that encourage being turned off for short periods was only acceptable to (or prevent) demand reduction or shifting to other times. The 3 individual reflections were then shared and discussed with the band. These changes were categorised as easy, moderate, and group and compiled in a list of behaviours and their change- difficult. The scenario, changes, capacity bands and the related encouraging/preventing factors. The workshop re-convened a pricing and implications are discussed in the subsequent week later to discuss the household-persona construction for subsections. analysis of consumption and demand shifting behaviours. The workshop participants fully acknowledge that be- Personas are archetypal users that exemplify the characteris- haviours and factors outlined in the resultant lists and the tics of actual users in an understandable and amiable form [9]. household-persona are biased towards the perceptions and First developed as a tool to aid software development [34], experiences of the participants. Yet, as this is an explorative personas have become a commonly used method for product study, both the lists and the sample household-personal are and system design. They can be used to consider the needs and considered to be adequate tools for the present exploration. goals of the users that the system is being designed for [9]. Thus, personas enable users to be a key component of the B. Scenario Design design process, a tool that is fundamental for the success of The exploratory scenario is set between 5 and 7 pm GMT product or system design (e.g., Maness et al. [35]). on a weekday in the UK. Personas are increasingly used to examine novel energy sys- The persona-household, which consists of a mother, tems. For example, they have been used to explore behaviours, father, a teenage son and a younger daughter, have attitudes and motivations towards domestic energy retrofitting all just come home. They are not even aware that in owner-occupied homes [9]. Likewise, personas provided an they left the light in the front garden on, as they understanding about how households make decisions about used it when unlocking the door. The fridge is electricity pricing [36]. Similarly, Dodge et al. [37] used running, as normal, without anyone’s interference, personas to examine how electricity consumption feedback as is the router (they always have the router on). affects consumption behaviour. Due to their ability to make de- The father has started preparing dinner. He is in the signers’ assumptions explicit, provide meaningful constraints kitchen, with lights on, listening to music on the to the problem space, and represent users in an engaging stereo, while using an induction hob to cook some manner, personas are particularly relevant for building a shared vegetables and the oven to prepare fish. He has also understanding of the different user groups that might engage switched on the kitchen extractor fan to remove with a new system [9], [38], [39]. In the present case, we cooking smells and fumes. The mother comes into expect a persona-driven study to provide fruitful insights for the kitchen and boils the kettle to make herself and understanding the implications of capacity-based pricing for her husband a cup of tea. She notices that all the peak-demand shaving in one household type. cups, some cutlery and most dishes are used and Although personas are frequently based on qualitative re- have been placed in the dishwasher, and there will search [40], they can also be assumption-based [41], [42]. Ad- not be enough clean dishes for serving dinner. So hoc personas are particularly effective in the early stages of she also switches dishwasher on a fast cycle. Before a project to formalise what developers know or infer about coming into the kitchen to make a cup of tea, the users [9]. Thus, they represent a resource-efficient method for mother has been working on her laptop, and has exploring the needs of different users [42]. Moreover, while left it switched on while charging because she will personas can be used as a distinct method to support design, return to it with her tea, when ready. they can also be used in combination with other (qualitative The teenage son has been playing football at school, and quantitative) methods to amplify their effectiveness [43]. and is now taking an electric shower to wash away We used our persona in combination with data on power draw the mud. As he switches on the shower, the towel of some common household devices and average demograph- rail starts to warm up, and the bathroom extractor ics of the UK households to outline scenarios and explore the fan turns on to remove the steam. He will need feasibility of a capacity-based pricing model. his sports kit again for school tomorrow, so his During the workshop the participants defined occupancy father has started the washing machine to get the energy service profiles for their own personal households. sports kit washed and ready for the morning (he The occupancy profile that was most substantially considered has two sets of sports kit, but the other one was during the workshop was that of a family household with used over the weekend and needed a wash too). two working adults and two children. This profile was then Meanwhile, the daughter is playing a game on the selected for this current scenario analysis in which the persona- games console hooked to an LED TV. She has also household undertakes various energy-demand activities on connected her smartphone to a quick charger, as an average evening peak demand time. This scenario then she is going to have a long call with a friend later provides a quantitative basis to explore the flexibility of a on, and the phone was running low. variety of energy services. Having outlined the scenario, the participants then discussed C. Power Draw and Energy Consumption changes that could be incorporated into the given scenario For each device category providing some energy consuming to reduce the power draw and retain it within a set capacity service we sourced a representative maximum and average 4 power draw during use at grid peak (5 to 7pm GMT). We the most energy intense households routinely drawing more also estimated how long a device would be in use during the than 18kW (on average) while the least energy intense draw peak time interval. We estimated the duration of use of the just over 2kW from the grid. According to the same survey, the individual energy services based on the participant personal mean average maximum power demand for households with experience. We then calculate the total energy consumption per primary electric heating was 9.3kW. device category over the entire peak time duration. Our power In the following section will analyse different interventions consumption estimates are derived from real-time measure- that can reduce the demand in the persona household. We ment, user guides and a wide range of both lab and consumers ground this analysis in the responses from the previously websites such as CSE [44] and British Gas [45]. The power mentioned demand shifting workshop which we describe next. consumption values for the device categories are supposed to be representative only. The power consumption by actual E. Demand Shifting Scenarios devices may vary from our values. The goal of the values As noted above, to explore the potential for shifting energy is not to predict average household power consumption but demand away from the peak time, ten researcher-users carried to provide a quantitative view on the implications of capacity- out reflection and analysis activities at a co-design workshop. based pricing that can help rank and sanity check conclusions. During the workshop participants brainstormed the set of For many device categories, power consumption varies energy services they typically used during peak time. throughout their use. For example, while a typical washing For each service, they considered the motivating factors machine draws on average 440W while heating water, it draws that drove their use; these are here called “contexts”. The significantly less energy at other times during the wash cycle workshop participants then thought of interventions that could with average power consumption of about 50W. Similarly, a be undertaken to defer or substitute the energy service for fridge/freezer draws an average of 40W but it peaks when each context. These interventions were then rated according the freezer door is left open for too long at 400W. The oven to their ease of shifting to result in a reduction of peak shows a steady high power draw and the washing machine and energy consumption. The workshop participants also identified dishwasher peak at the beginning and end of their usage cycle enabling and constraining factors for each context specific respectively. Other appliances have an continuous power draw intervention or change. such as light bulbs and chargers. IV. R ESULTS AND I MPLICATIONS D. Capacity Bands A. Peak Time Power Consumption Variability can also be found in the peak demand for The individual energy services together with their power electricity from the grid; affected by several factors, including consumption values are listed in Table I. The overall energy chiefly the weather and working/non-working days. consumption in the scenario is 12.7kWh which equates to an Additionally, statistical variability of demand in households average peak power draw of about 6.4kW (when accounting aggregates to a increasingly less variable load on the grid for the estimated duration of the device use). branching points to combine to the observed aggregate levels of demand in the grid [46]. At the local transformer station B. Workshop Results - Demand Shifting Potential that aggregates the connections of the individual households In this subsection we present the ease of shifting various the instantaneous power draw of each household is relevant as household energy services as it had been identified during the they are installed with a given peak capacity (either single or workshop. The participants discussed shifting the peak time three phase 100A with a maximum power of 24kW/41kW). In use by dishwashers, washing machines, power showers, hobs, the result section we estimate this instantaneous power draw ovens, microwaves and kettles. The participants produced a list by a persona household. For comparison, we also calculate the of contexts under which the appliances were used. For each effective household power consumption during the peak period context they then judged the ease of shifting the appliance (5-7pm) which is calculated from the total energy consumption use as either having low, medium or high potential to provide during the peak period divided by the duration of the peak effective shifting of energy consumption away from peak time. period. Assuming everything is equal any local reduction of The summary of the appliance use contexts, as well as possible demand will also result in a reduction of peak demand at the drivers and obstacles for shifting the appliance use time are aggregate level. Deferral of local demand spikes to a time summarised in Table II. outside of the peak period will translate to a marginal reduction The most common intervention to enable deferral of energy of aggregated peak demand at grid level. However, deferral of consumption was found to be additional planning of appliance local demand spikes within the peak period do not generally use to take place outside of the peak period. The greatest affect the aggregated peak period demand at grid level. flexibility and thus most probable changes were identified to According to the Household Energy Survey [47] the average be those for which the time of the appliance use is incon- peak demand for a household without electric heating was sequential. Conversely, participants found that consumption 7.5kW. Importantly, the mean (of the sampled population) is most difficult to shift in case when energy services are average (typical for the household) peak power demand varies needed to respond to the immediate context. This usually refers significantly within the population of survey households, with to activities that take place once household members return 5 TABLE I in the above scenario be motion sensitive, they would turn on AVERAGE POWER DRAW BY ENERGY SERVICES IN THE PERSONA and off for the period where the family members enter or leave HOUSEHOLD . the house. Device category Power Draw [W] Use Peak max / (avg) duration time C. Factors supporting behaviour change (h energy between [kWh] In all cases where the use displacement requires effort 5-7pm) from the consumers and behaviour change, the workshop Electric shower 10500 (10500) 0.25 2.625 participants noted that such efforts would be fostered through Induction Hob (3 3500 (3500) 1 3.5 plates) motivational factors such as: Kettle 3000 (3000) 0.3 0.9 • Direct and sufficient financial benefit: should the income Dish washer 1800 (1000) 2 2 Oven 2400 (2400) 1 2.4 gained from the consumption time displacement be sub- Washing machine 440 (50) 2 0.1 stantial, most of the consumers would likely be willing to (avg 50W) carry out the required planning and preparation activities. Fridge/Freezer (40W 400 (40) 2 0.08 avg) Thus, capacity-based pricing with substantial differences Towel rail 250 (250) 1 0.25 in the prices per capacity band could be a promising Games console 100 (100) 2 0.2 solution. LED TV 80 (80) 2 0.16 • Ownership of carbon savings, where the households that Kitchen lighting 80 (80) 2 0.16 (LED) reduce their consumption are able to monetise their re- Kitchen extractor fan 50 (50) 1 0.05 duced greenhouse gas emissions. Here, instead of relying Smart phone (quick 35 (35) 1 0.035 charge) on a utility provider to set capacity bands, the individual Laptop 25 (25) 2 0.05 households can directly monetise energy demand below a TV box 25 (25) 2 0.05 capacity band and their “saved” emissions vis-a-vis their Extractor fan 24 (24) 2 0.048 Stereo 20 (20) 2 0.04 personal average. Garden lightning CFL 20 (20) 2 0.04 • Environmental benefits are another powerful motivator: Internet router 12 (12) 2 0.024 in contrast to personal monetary benefits, this motivation Tablet (charge) 10 (10) 2 0.02 stems from the ethical/societal values and beliefs of the total 12.732 [kWh] households who are concerned with climate change and eff. 6.366 wish to contribute to healthy living environments. power • Visibility of power draw allows users to see, and so [kW] remind themselves about their ongoing energy use (not unlike how the dripping tap reminds about the water loss). • Legacy (non-smart) devices prevent the user from ca- home. For example, in our scenario the son has to take a pacity shifting as they do not support preparation and shower right after arriving at home as he needs to wash away planning (e.g., if the dishwasher does not allow start time the mud immediately after the football game, which was just delay and/or remote activation, its use cannot be shifted to completed; the family must have dinner after the work/school late night or midday off-peak times, as somebody would day. These activities are difficult to shift because even with have to be present to activate the device). planning and preparation the consumers would need to respond • Finally, connecting with other users to form a group to to the context that they find themselves in when they return share capacity and spread the load could help even-out the home. variability in individual household daily activities, e.g., Medium levels of complexity to increase energy flexibility when the arrival of an unexpected guest requires extra were attested by the workshop participants to cases where cooking and/or washing activities. planning and preparation allow to move the consumption activities away from the peak time. For example, if the family D. The role of ICT to enable capacity reduction in the above scenario had installed a pre-heating kettle, there The workshop participants identified ICT as an enabling would be no need to boil the water at the peak time. This technology that can support our household-persona in meeting change is relatively easy to make as it requires a “one-time” capacity constraints, and therefore hopefully making it both effort of new kettle installation. On the other hand, the par- easier and more acceptable to those involved. In this section, ticipants agreed that energy consuming behaviour associated we consider the role of ICT in supporting our alternative to established practices (“rituals”) is much more difficult to capacity-constraint driven scenarios more generally. We cat- change and thus has lower potential. For example, getting egorise six ICT interventions: dishes washed right before when they are used (as the mother 1) Awareness: Eco-feedback and smart meter technology did in the scenario). have long been advocated as a means of encouraging reduced In contrast, the workshop participants found that automation energy usage. They have demonstrated modest but non-trivial of activity that is possible without any behavioural change are results, leading to criticism from some quarters. However, such easiest to implement. For instance, should the garden lights techniques are likely to be essential in supporting households 6 TABLE II F LEXIBILITY FOR DEMAND SHIFTING PER ENERGY SERVICE . Devices Context Potential (Ease of Shift- Shifting enablers Constraints ing) Need for clean Low Planning dishes Audio / visual reminders of full machine / Dishwasher empty cupboard Understanding implications of particular Resistance to change Ritual of daily uses use even if not high Automation to run at more suitable time Noise if running at night full Digi-reminder that its not full Automation to run when full within user-set limits Capacity of ma- Larger machine could reduce number of Dried-on food necessitates more chine uses intensive wash Need particular Automation to have a ‘complete by’ setting Need to be ready by a particular medium clothes time Clothes not available to wash ear- lier eg work / sports kit Washing machine Awareness of implications of different use Availability of user to load / unload times Regular wash of Differential pricing for different times accumulated high Delay / timer settings User has to have loaded machine washing Not left too long in machine Need to hang up or move to dryer Noise of machine, rules in the building re use Wash person - medium Visual display of other power draws in the Small window for wash Power shower evening home Morning wash medium Shorter shower - using timer Low Display showing cost / capacity at time of hungry family, timing when people Cooker - hob dinner use are home Medium - suggested by 20-30 mins Automation to allow short duration switch- dinner medium Cooker - oven ing off when other short duration devices are used pre-heating baking medium Microwave Defrosting / heat- medium ing Hot drink and Pre-heated water, +insulation of device, +in- medium Kettle forgetting use / terface automation reboiling reminder that its just boiled Cooking meal medium Pre-heat water at more suitable time inter- speed up face with other devices / automation availability of hot water asked to operate within a capacity constraint. Besides the decision support would spot patterns of appliance use, and/or traditional smart meter approach of providing an overall power have knowledge of the schedule of the household members consumption figure, this could be augmented with: via calendar integration. It would then make suggestions such • Some visual warning of an approach to a capacity con- as the most appropriate time to load the washing machine. straint (eg green/amber/red) 3) Buffering in response to anticipated demand at peak • Feedback at point of use. Eg a kettle displaying a warning times, and/or spreading of power demand: Energy at off- that switching it on will exceed the constraint. peak times can be used to prepare systems to reduce energy 2) Decision support: ICT can provide information and use during peak periods. In our scenario, we can imagine an prompts to enable behaviour change. This could be carried out insulated kettle which automatically pre-boils when demand both in-the-moment and proactively. In-the-moment decision is lower (possibly in response to past historical data of likely support would make suggestions as to possible courses of use times) and then can be re-boiled using a lower power action now. For example, if you want to boil the kettle, you input at peak times. Similarly, a fridge/freezer can cool itself will need to either wait 5 minutes for the oven to finish to a lower temperature in anticipation of peak periods, and so heating, or switch off the oven until it is boiled. Proactive either avoid re-chilling or chill at a slower rate. One important 7 area of buffering that ICT can support is the use of batteries common and co-occurring — they are principal to the national to enable self-consumption of energy generated from solar PV grid peak demand. On a national level it is the large-scale cells at off-peak times. In particular in the UK, peak demand alignment of work patterns (which includes the alignment of occurs outside of the time when solar PV cells are effective. the school day) that in turn causes the synchronisation of Local battery storage that is instrumented to release energy at consumption of energy in households during the peak period. peak times can be an effective mechanism for some suitable The ICT-enabled increase of home office jobs has somewhat households to reduce peak demand. increased the flexibility in timetables. In particular in the 4) Time slicing, micro-delays and appliance coordination: ICT sector there are a number of organisations of which the With the necessary integration of appliances such as washing staff is globally distributed and working on loosely coupled machines, dishwashers, ovens, hobs and fridge/freezers, ICT timetables. The diurnal rhythm of the body however naturally can multiplex available electricity among devices. These all limits the flexibility of meal and rest times. draw a high power, but for short periods of time. By automat- Macro delay is one of the most important ways in which ically coordinating them, and so delaying one while another ICT can enable flexibility. One important intervention to avoid is in operation, the overall peak demand can be reduced. For reinforcements, that is already been trialled, is the delay example, the washing machine and the dishwasher can adjust of charging electric vehicles to periods outside of the peak their operation to prevent concurrent heating cycles. interval [48]. According the national grid, electric vehicles This may result in short delays in estimated completion could add between 5 to 8 GW peak electricity demand to the time. In addition, ICT has a role in making this automation electric grid. This is not surprising given that electric vehicle visible and appropriately controllable by the user. It is known battery chargers can draw between 3 and 90kW. that there is resistance to allowing energy companies to re- Despite the potential that ICT has to create flexibility of motely make such micro-delays in customers. How can control demand, behavioural changes seems to be required for a and authority be placed in a customers hands and improve significant reduction of peak electricity demand. One area of acceptability of a strategy such as this? further work is to describe the factors that support behavioural 5) Macro-delay: Smart appliances, such as dishwashers and changes through theoretical frameworks, for example from washing machines, already have the ability to delay activity to behavioural psychology. off peak times such as running overnight. Anecdotal evidence Although framed in the context of grid reinforcement and suggests these abilities have only modest uptake in usage the avoidance thereof, the characteristics of demand flexibility currently. Can this functionality be combined with decision and the role of ICT as enabler are relevant in order to support support to find acceptable changes in routine for a household, the decarbonisation of the energy system. and coach them through it? Our scenarios are based on a persona household that is 6) Efficiency: Finally, via sensing and control ICT can inspired by personal experience but statistically not significant. be used to provide energy services when they are needed However, the actual power consumption level is relatively and in exactly the right proportion. This can prevent waste. similar to the average power consumption by the households For example, an intelligent washing machine can reduce the measured as part of the Household Energy Survey (7.5kW - amount of water heated for small loads or intelligent hobs no electrical heating) and thus are representative of real loads and instrumented pots can use the right amount of heat for a by residential households. specific recipe. V. D ISCUSSION VI. C ONCLUSION We have identified the following limitations and further Our workshop participants have identified a number of flex- research questions, that need addressing in order to reduce ible energy services with high load on the electric grid. Given significant uncertainty around the effectiveness of capacity- these services capacity-based pricing is an important lever for based pricing to encourage demand management. reducing peak demand. We have identified a categorisation of With easy-medium changes in place, our representative the ways in which ICT can avoid or support hard behavioural household would reduce the power draw from 23kW peak changes to support the shifting of energy service use. to around 17.7kW peak and further to 12-12.6kW peak with Given the limitations of our research approach, these find- hard changes in place, which are still not enough to stay below ings require significant empirical substantiation for gener- capacity band C. alisable conclusions to be drawn. Nevertheless, the results Delaying the power shower would shave 10,500W, bringing point towards the need to combine capacity-based pricing the household down from 17.7kW down to 7.2kW. The shower with a wide range of demand shifting facilitation. From a has been identified as a hard change. policy perspective, it is evident that capacity-based pricing Our scenarios indicated that, apart from the electric shower, gives the regulator greater leaverage to encourage demand the most energy consuming activities (those related to cooking reducing behaviour. Capacity-based pricing also lends itself to and boiling water) are the least flexible as they require rising block tariffs where a certain capacity can be considered significant behaviour changes. This should not be surprising as ‘baseload’ for each household persona (e.g., a family of four it is their lack of flexibility that results in these activities being versus an couple) with step-wise increases for higher demand. 8 This allows more vulnerable customers to be granted with a [19] A. J. Roscoe and G. Ault, “Supporting high penetrations of renewable certain ‘baseload’ at a reduced rate to address fuel poverty. generation via implementation of real-time electricity pricing and de- mand response,” IET Renewable Power Generation, vol. 4, no. 4, pp. Volumetric pricing alters the price signal which might 369–382, 2010. encourage energy saving behaviour and increase demand for [20] G. Dutta and K. Mitra, “A literature review on dynamic pricing of energy efficiency services. Overall, this paper suggests that electricity,” Journal of the Operational Research Society, vol. 68, no. 10, pp. 1131–1145, 2017. more research is required into the factors that are most likely [21] K. Stenner, E. Frederiks, E. V. Hobman, and S. Meikle, “Australian to change energy demand and underlying practices through consumers likely response to cost-reflective electricity pricing,” 2015. a combination of altruism, price signals, nudging and ICT [22] K. Mitra and G. Dutta, “A two-part dynamic pricing policy for house- hold electricity consumption scheduling with minimized expenditure,” automation. International Journal of Electrical Power & Energy Systems, vol. 100, pp. 29–41, 2018. ACKNOWLEDGMENTS [23] R. G. Laura Sandys, Jeff Hardy, “Reshaping regulation: Powering from This work is partly funded by the UK EPSRC funding the future,” 2018, https://www.imperial.ac.uk/media/imperial-college/ grantham-institute/public/publications/collaborative-publications/ for Household Supplier Energy Market (EP/P031838/1) and Reshaping-Regulation-Powering-from-the-future.pdf. Refactoring Energy Systems(EP/R007373/1)projects. [24] M. A. A. Pedrasa, T. D. Spooner, and I. F. MacGill, “Coordinated scheduling of residential distributed energy resources to optimize smart R EFERENCES home energy services,” IEEE Transactions on Smart Grid, vol. 1, no. 2, [1] IEA, “Global energy & co2 status report 2017,” International Energy pp. 134–143, 2010. Agency, Tech. Rep., 2018. [25] A. Agnetis, G. Dellino, P. Detti, G. Innocenti, G. De Pascale, and [2] NationalGridHouse, “Future energy scenarios,” National grid, Tech. A. Vicino, “Appliance operation scheduling for electricity consumption Rep., 2018. optimization,” in Decision and Control and European Control Confer- [3] HouseofCommons, “2020 renewable heat and transport targets,” House ence (CDC-ECC), 2011 50th IEEE Conference on. IEEE, 2011, pp. of Commons, Tech. Rep., 2016, second Report of Session 201617, 5899–5904. together with formal minutes relating to the report. [26] C. Boomsma, S. Pahl, J. Goodhew, and A. Gardner, “Daily behaviour [4] BEIS and Ofgem, “Upgrading our energy system smart systems and and energy saving strategies,” 06 2015, in ECEEE SUMMER STUDY flexibility plan: Progress update,” 2017, the National Archives, London. PROCEEDINGS. [5] ofgem, “Upgrading our energy system smart systems and flexibility [27] R. V. Jones, A. Fuertes, and K. J. Lomas, “The socio-economic, plan,” 2017. dwelling and appliance related factors affecting electricity consumption [6] E. Catapult, “Preparing uk electricity networks for electric vehicles,” in domestic buildings,” Renewable and Sustainable Energy Reviews, 2018. vol. 43, pp. 901–917, 2015. [7] ofgem, “Electricity system flexibility,” Tech. Rep., [28] A. Kavousian, R. Rajagopal, and M. Fischer, “Determinants of residen- 2018 (accessed January 2, 2019), https://www.ofgem. tial electricity consumption: Using smart meter data to examine the effect gov.uk/electricity/retail-market/market-review-and-reform/ of climate, building characteristics, appliance stock, and occupants’ smarter-markets-programme/electricity-system-flexibility?page=1# behavior,” Energy, vol. 55, pp. 184–194, 2013. block-views-publications-and-updates-block. [29] E. Shove and G. Walker, “What is energy for? social practice and [8] BEIS, “Ehv distribution charging methodology (edcm),” 2011, accessed energy demand,” Theory, Culture & Society, vol. 31, no. 5, pp. 41–58, through ofgem.gov.uk January 2019. 2014. [Online]. Available: https://doi.org/10.1177/0263276414536746 [9] V. Haines and V. Mitchell, “A persona-based approach to domestic [30] K. Parkhill, C. Demski, C. Butler, A. Spence, and N. Pidgeon, “Trans- energy retrofit,” Building Research & Information, vol. 42, no. 4, pp. forming the uk energy system: public values, attitudes and acceptability: 462–476, 2014. synthesis report,” 2013. [10] H. Kondziella and T. Bruckner, “Flexibility requirements of renewable [31] A. Spence, C. Demski, C. Butler, K. Parkhill, and N. Pidgeon, “Public energy based electricity systems a review of research results and perceptions of demand-side management and a smarter energy future,” methodologies,” Renewable and Sustainable Energy Reviews, vol. 53, Nature Climate Change, vol. 5, no. 6, p. 550, 2015. pp. 10–22, 2016. [Online]. Available: http://dx.doi.org/10.1016/j.rser. [32] K. Burchell, R. Rettie, and T. C. Roberts, “What is energy know-how 2015.07.199 and how can it be shared and acquired by householders?” ECEEE, [11] L. Grzybowski and F. Verboven, “Substitution and Complementarity 2015. between Fixed-Line and Mobile Access,” 2013. [33] K. Jenkins, D. McCauley, R. Heffron, H. Stephan, and R. Rehner, “En- [12] A. Mathur, B. Schlotfeldt, and M. Chetty, “A mixed-methods study of ergy justice: a conceptual review,” Energy Research & Social Science, mobile users’ data usage practices in south africa,” in Proceedings of the vol. 11, pp. 174–182, 2016. 2015 ACM International Joint Conference on Pervasive and Ubiquitous [34] A. Cooper et al., “The inmates are running the asylum:[why high-tech Computing. ACM, 2015, pp. 1209–1220. products drive us crazy and how to restore the sanity](vol. 261),” Sams [13] M. J. Fell, D. Shipworth, G. M. Huebner, and C. A. Elwell, “Public Indianapolis, 1999. acceptability of domestic demand-side response in great britain: The [35] J. M. Maness, T. Miaskiewicz, and T. Sumner, “Using personas to role of automation and direct load control,” Energy research & social understand the needs and goals of institutional repository users,” D-Lib science, vol. 9, pp. 72–84, 2015. Magazine, vol. 14, no. 9/10, pp. 1082–9873, 2008. [14] F. Wolak, “Efficient pricing: The key to unlocking radical innovation [36] R. Russell-Bennett, R. Mulcahy, K. Letheren, R. McAndrew, T. Swinton, in the electricity sector,” OECD Working Party No., vol. 2, no. 20174, N. Horrocks, and R. Ossington, “Innovating energy markets: a hybrid 2017. persona/segment approach,” 2017. [15] T. K. Jason Palmer, Nicola Terry, “Further analysis of the household [37] J. Dodge, M. Hilton, R. A. Metoyer, J. Hunter, K. Smeltzer, C. Vijay, electricity survey - early findings: Demand side management.” Cam- and A. Atkinson, “Deriving age diverse personas from a participatory bridge Architectural Research Limited, Tech. Rep., 2013, reference: design study on home electricity feedback,” in Proceedings of the 2017 475/09/2012. CHI Conference Extended Abstracts on Human Factors in Computing [16] E. V. Hobman, E. R. Frederiks, K. Stenner, and S. Meikle, “Uptake and Systems. ACM, 2017, pp. 959–968. usage of cost-reflective electricity pricing: Insights from psychology and [38] K. Dorst and N. Cross, “Creativity in the design process: co-evolution behavioural economics,” Renewable and Sustainable Energy Reviews, of problem–solution,” Design studies, vol. 22, no. 5, pp. 425–437, 2001. vol. 57, pp. 455–467, 2016. [39] J. Pruitt and T. Adlin, “The Persona Lifecycle: Keeping People in Mind [17] C. Eid, E. Koliou, M. Valles, J. Reneses, and R. Hakvoort, “Time- Throughout Product Design,” Morgan Kaufmann Publishers, 2006. based pricing and electricity demand response: Existing barriers and [40] A. Cooper, The Inmates are Running the Asylum. Wiesbaden: next steps,” Utilities Policy, vol. 40, pp. 15–25, 2016. Vieweg+Teubner Verlag, 1999, pp. 17–17. [Online]. Available: [18] A. Faruqui, R. Hledik, and J. Tsoukalis, “The power of dynamic pricing,” https://doi.org/10.1007/978-3-322-99786-9 1 The Electricity Journal, vol. 22, no. 3, pp. 42–56, 2009. 9 [41] D. A. Norman, Ad hoc Personas and Emathetic Focus, ser. how-much-electricity-am-i-using#table/. Interactive Technologies Series. Amsterdam, Netherlands: Morgan [45] BritishGas, This is the actual energy cost of your Kaufmann Publishers, Elsevier, 2006, pp. 154–157. [Online]. Available: daily routine, 2017 (accessed January 2, 2019), https: https://books.google.co.uk/books?id=wWuBbTCcsCoC //www.redonline.co.uk/interiors/decorating-ideas/a529321/ [42] T. Adlin and J. Pruitt, The essential persona lifecycle: Your guide to this-is-the-actual-energy-cost-of-your-daily-routine/. building and using personas. Morgan Kaufmann, 2010. [46] Gridwatch, “Live monitoring of the UK electricity National Grid,” [43] J. Pruitt and J. Grudin, “Personas: practice and theory,” in Proceedings 2019. [Online]. Available: http://gridwatch.co.uk/ of the 2003 conference on Designing for user experiences. ACM, 2003, [47] N. Terry and J. Palmer, “Trends in home computing and energy demand,” pp. 1–15. Building Research and Information, vol. 44, no. 2, pp. 175–187, 2016. [44] CSE, How much electricity am I using?, Not dated (accessed [48] Electric Nation, “Electric Nation — Your electric vehicle - Your smart January 2, 2019), https://www.cse.org.uk/advice/advice-and-support/ charge,” 2019. [Online]. Available: http://www.electricnation.org.uk/ 10