1 Introduction
Within the current climate emergency, energy communities have emerged as a key element of many strategies for achieving transition to clean energy sources [
8,
36]. These communities harness and store renewable energy, like solar or wind power, and seek ways to support local use of the power generated to help tackle challenges like decarbonization, grid management, and energy affordability. Energy communities take diverse forms, from neighbors sharing solar panels, to groups collectively switching energy suppliers, and are defined by collective forms of action and self-understanding that incorporate energy. Various governments, including the UK, are now also creating regulatory frameworks and initiating pilot projects to facilitate the growth of energy communities [
19]. Despite the potential and growing popularity of such communities, research on energy collaboration remains relatively limited within the field of HCI, especially regarding collective demand-shifting. Demand-shifting refers to the practice of adjusting energy consumption patterns, so as to align with the availability of renewable energy resources. For instance, it might involve modifying the timing of energy-intensive activities, such as using household appliances or charging electric vehicles, to coincide with periods when renewable energy, like from solar or wind sources, is available. In the context of energy communities, demand-shifting requires additional coordination across households so as to ensure that the total communal consumption still remains within the available renewable energy.
For HCI this coordination offers a number of interesting challenges for supporting the technical as well as the social expectations of these groups. For instance, the interfaces and applications such communities may need to coordinate might vary depending on the diversity of renewable sources and storage solutions, the community’s size and social fabric as well as their energy and technical literacy. Preliminary, exploratory work on community demand-shifting has suggested that coordination can introduce complexities such as accountability, privacy, fairness, social harmony [
39], as well as the necessity for cooperation [
26,
27]. Nevertheless, such prior work has mostly examined coordination theoretically, outside the context of actual neighborhoods [
27,
39] and has used off-the-shelf technologies rather than develop them in a user-centered way [
26]. Since in-the-wild evaluation can help produce more ecologically valid insights on the use of domestic technology [
52,
53], we build on existing research to design SolarClub, an energy community coordination system, and then evaluate it with real households.
Inspired by existing community energy pilots [
35], we specifically designed for the situation in which a group of neighbors (which we refer to as a ‘club’) share the energy generated by a solar panel that is collectively owned. Following a user-centered process we propose SolarClub, an interactive tool through which households can indicate their intention to run high-consumption appliances. SolarClub serves both as an eco-feedback system to examine live group energy consumption, as well as a coordination tool for future energy use. We deployed SolarClub in-the-wild with four groups of neighboring households in the UK and investigated its reception. The solar generation from the shared panels was simulated using real live data.
Through a combination of interviews and interaction log data, our findings indicate that the design of SolarClub successfully enabled neighbors to demand-shift and coordinate their electricity consumption, even when some participating households were less flexible themselves. Nonetheless, despite feeling able to coordinate, participants expected a stronger feeling of community, which SolarClub did not offer. Finally, unlike previous studies, not conducted in-the-wild, we found that participants asked for less rather than more privacy within their clubs and saw how SolarClub encouraged a feeling of empathy to neighbors even if they didn’t actively engage in-person throughout the deployment. We unpack these findings in the context of prior work, and present a series of considerations for designing demand-shifting systems among energy communities. This study makes a novel empirical contribution to demonstrate how HCI can support energy communities, while also broadening our understanding of the forms energy communities can take. These findings thus can support governments and other stakeholders work towards a renewable energy transition.
5 Findings
We first present our quantitative analysis of data collected by the SolarClub systems, tracking patterns of booking and consumption. We then discuss four themes emerging from the semi-structured interviews, that relate to observed patterns of engagement with the system and interaction within groups. These themes explore different scalar understandings of coordination emerging from the study. Participants framed coordination as unfolding within the household, among neighbours, within an energy community, and through imagination and empathy.
5.1 Engagement
Overall, the intervention elicited a variety of reactions from the participants. Participants like P11 who were generally flexible and self-described as doing quite well in the intervention, mentioned how “it was lovely to dip a toe in the water of that [community energy], [.] it was nice to do something about that, because it’s something which I very much believe in.” (P11). Others similarly found it “empowering to participate" (P1) and a positive exercise to bring you closer to your neighbors (P3, P9, P14). P16 also made concrete action points for the future to bring alternative solutions to their local council. Yet, the deployment also made some participants feel ‘restricted’ in their actions (P4) or ‘odd’ (P2) for needing to plan things in advance.
In line with the findings of previous work on people’s experiences with annotating their household’s electricity usage e.g. [
11,
21,
39]. participants enjoyed and learned from the annotating process, making an average of 28.9 annotations (min:11,max:53,SD:13.2) over one week
5. Unlike these previous studies, however, in this case households were instructed to focus on annotating their larger appliances, potentially affecting the quantity of annotations and limiting comparability.
In total, each of the 14 active households made between 4 and 61 bookings (mean: 19, SD:15.8), for a total of 268. The amount of bookings, on average, seemed to be quite stable with a small peak on the second of the three weeks of coordination (Figure
5). This finding is in line with previous work that noticed that a drop in eco-feedback engagement only occurred after approximately four weeks [
3,
41]. The majority of bookings (38%) were made 1-2 hours in advance of the booking start time, but in general bookings were made from less than a minute to 5 days in advance (mean: 12.5, SD:22 hours). The amount of time between making the booking and running the activity changed across participants as well as throughout the deployment. Figure
6 summarizes the bookings by household indicating the variation in household patterns. For instance, some people booked last minute “Generally four minutes before I wanted to use it”(P1, H1). H6 who made the most bookings overall also seemed to make these mostly the hour before the actual activity. Others planned only in the morning for the same day:
P4 (H3): Our planning was more done in discussion between her and I, [.] overnight to plan the day.
P5 (H3): Yeah. Yeah. We tended to book the night before or early that morning. So we were planning ahead.
For P11 (H8) the booking behavior evolved with the realization that the forecast accuracy is lower the longer in advance it is consulted, so that early bookings would be more likely to overshoot the actual solar availability (at the time they were run).
Moreover, as shown in Table
2, all groups but the last one, were ‘successful’ in shifting their energy consumption towards hours when there was solar. The last group included two households which engaged in energy shifting, but the other two (H15 and H14) were not actively participating. Interestingly, when comparing the last two columns of Table
2 we find that the amount of bookings does not seem to imply higher shift percentage. H7 which consisted of P10 who worked until late during the week, only booked 4 times. Nevertheless they still seemed to shift their overall consumption to more solar hours. On the other hand H5 did not manage to shift (their energy was already 47% covered by solar), yet they nevertheless signaled their activities by making 40 bookings.
Table
3 reports the solar generation for each of the groups. This data suggests that for three groups (1, 2, and 4) the solar generation was higher during the coordination weeks, on average, than during the annotation week. However, there seems to be no pattern of higher solar consumption being explained by higher solar generation: for groups 1 and 3 the percentage of solar consumption (as a share of solar generation) went up during the coordination phase, while for the other two groups it went down. This makes sense since where availability of solar generation is relatively plentiful and doesn’t act as a limiter on consumption (i.e. where participants’ energy needs are already being met), we wouldn’t necessarily expect a strong association between the two.
In all cases but one (P15) participants booked only to indicate their intention of using the available solar energy (i.e. made bookings only for times when solar generation was expected) as shown in Figure
8. P15 instead also booked their dishwasher outside the available solar hours in four bookings (out of their 17 in total, perhaps due to their inflexibility as detailed below). Participants understood booking as a strategy for solar coordination hence making sense only during solar generation times. As Figure
7 demonstrates, households booked when they wanted to consume during the day (i.e. there is coherence between the blue peaks and the pink bookings), and still consumed without booking outside the solar hours when they needed to.
Overall, only seven messages to other club members were sent through the MessageClub feature, whereas almost 25% of bookings included a note (67 out of the 268 total). Finally, we analyzed the flexibility data for the 101 bookings that were made at least 6 hours in advance. We focused on this subset of bookings because participants mentioned that when they booked their slots last-minute they were not flexible at that point anymore. Of these 101 bookings, 18 were marked as ‘no’ flexibility, 23 as ‘somewhat’ flexible and the remaining 60 as ‘flexible’ (which was also the default, pre-selected option in the dropdown). We analyze these flexibility indicators qualitatively in Section
5.3.
5.2 Using and Coordinating with SolarClub as a Household
Participants mentioned that the SolarClub graph was ‘simple’ (P7), ‘clear’ (P9, P17), ‘intuitive’ (P10, P11) even ‘satisfying’ (P12) to read and interact with. “I think the display works well in terms of understanding the resource and understanding what other people are doing.”(P1). The annotation drop-down selection however was described as ‘clunky’(P1) and hard to understand (P16) even if all participants managed to use it. Further analysis revealed this to be the case because of a bug where the labels used for the annotations did not match the ones in the coordination interface (e.g. ‘washing and drying’ became ‘laundry’).
The interviews highlighted how, the coordination phase required substantial within household discussion and that, at times, could create friction. For instance, P9 and their daughter were engaged with coordinating whereas the other parent was not: “I [.] just liked doing the project, booking in the time, and – it’s about putting order into your life, and I quite liked that bit of it, and that’s what irritated my wife a lot, I think because she doesn’t want to have to deal with that.”(P9). P13 who is a working father of four, shared the responsibility of how to use the booking system with his son (and to an extent with the younger daughters), making him responsible for coordinating with family members on their bookings. Moreover, unexpectedly, P11 had guests and P8 a new tenant half way through the coordination phase. While this meant that the households would have increasing energy consumption, they also found how these new members also became engaged in the coordination themselves finding it easy to understand the SolarClub system.
It should be noted that our intervention and study were not explicitly designed to account for coordination within the households. So, for example, while where possible we onboarded all householders, this was not always the case due to participants’ availability or intrinsic motivation to take part. Moreover, we did not provide dedicated resources that could be used to onboard those not present. Therefore, serendipitous variations in how many household members were onboarded, brought to light different dynamics within the households. In H2 and H4, for instance, we onboarded both members of the household in the introductory session. These two households went on to share the tasks of planning and filling in the booking among themselves. On the contrary, in households H1 and H3 we only onboarded a single participant which meant that they had to then explain the system to the other family members which required understanding and verbalizing the aim of the study. In the case of household H3, this was a point of tension as P5 signed up for participating yet only P4 was present for the initial onboarding. Perhaps exactly because they were the ones onboarded, P4 took the research seriously, wanting to ‘do it properly’ and to communicate that buy-in to their partner (P5) who seemed to take a more casual approach. Indeed P5 told us about their partner P4: “P4 likes taking things very, very seriously. So when I hadn’t quite embraced the system, P4 would tell me off that, no, you haven’t booked to do that. You need to book to do that.”
On the other hand, P1 did not seem to get the same resistance in participation but still originally struggled in explaining the functionality and purpose to the rest of the family.
So [PARTNER’S NAME] would say to me, is it alright for me to put the washing machine on? Should I do that tonight? And I had to kind of keep wracking my brain about, okay, what is it we’re trying to achieve here? I’m trying to achieve using energy during daylight, and communicating that. So it doesn’t matter whether I booked it, if it’s sunny outside, you can do what you like. If I’m not around, we haven’t booked it. It doesn’t matter, but if it’s sunny outside, go for it. [.]
It sounds stupid now, but at the start of the project, that felt like a conversation that we kept having, and I kept failing to find the language to just say, if it’s sunny, put the washing machine on. Those were the words I needed, I think, I think three weeks to find them. (P1)
As described P1’s quote above, households tended to find simple rules, such as in this case just checking if there is sunshine or not, to understand how to interact with the SolarClub system and to explain this to those they lived with. Such rules demonstrated each household’s distinctive understanding of (simulated) solar energy in the context of the intervention and the actions required. For example, P9 and P12 similarly used the principle of “do stuff when there’s power available”(P9) or “look outside and see what the weather is” (P12). Interestingly, participants often used terms that reflected the visual language used in our UI: “the main focus was on, how can we change what we did to make it more in the middle?” (P4, referring to the visualization showing the distribution of solar throughout the day) or “And then, yes, [we] are not making the most of the sun at all, really. The gray should be more equal to the orange.”(P8).
5.3 Coordinating as Neighbors
When asked about the process of booking, participants mentioned that booking was not necessary to use solar energy, but was helpful to ensure that others are not using the power at the same time. In fact, some participants deemed the booking system entirely sufficient for the purpose of communication, avoiding the notes and messaging features. “I tried not to overutilize, so I just moved my own things, rather than go and talk to somebody else about doing something different.” (P9).
Participants explicitly mentioned they were not only booking their activities to match the solar generation, but even adjusting their plans to take into account the collective consumption e.g. “I had booked for 9 but could see the usage was high so moved to 13:00” (note on a booking by P4). Similarly, others expressed their availability to change plan through the booking system: “going to put on the dishwasher when I go out so let me know if you need the slot urgently [P12 NAME]” (signed note of P12). Overbooking, however, was not that common. Although the SolarClub system sent automated notification emails in the case that people did overbook, participants reported that they would only notice and read emails when it was too late to act. “And the trouble was that [.] didn’t check my emails over my phone. So the time had gone... [.] If I’d said, oh, yeah, of course I can be flexible. The time had gone by then. Anyway, so it made no difference.”(P4). The SolarClub design, however, was not sufficiently oriented towards immediacy; with reminders for the bookings being one of the common features mentioned for a future iteration. “Then ideally you’d have it on an app, and everybody would have the app. And then the app would ping to say, actually ‘P6 and P7 [are] doing a three-beans casserole’, and ‘could you turn your van off’, you know, or ‘it’s not going to be as sunny as we thought, turn it all down’, or whatever.” (P1).
Moreover, P15 who was working full time and taking care of grandchildren in the evenings and weekends mentioned how for them, the the SolarClub system does not seem to make sense when compared to the annotation interface as they do not have almost any flexibility in their daily life. “And because I was slightly defeatist, I suppose, because I thought, ‘Well, I can’t do my washing at 10 o’clock in the morning when there’s all that sunshine on a Monday. So what’s the point?’” (P15).
The interviews also provided more context for how participants signaled their flexibility: “Yes, again, initially I was always saying it’s flexible, and then as it went, as time went on, I kind of realized we were all doing the same thing, which was doing it [the booking] at the last minute.” (P11). For P16, making a booking implied that all the maneuvering that was possible has already been done. “I’ve already been flexible because I was choosing that time.”(P16) so that they would often (in 6 of their 10 bookings) indicate their bookings as ‘not flexible’. There seemed to be also a reciprocal relationship with indicating flexibility as participants did not seem to pay attention to other people’s stated flexibility (i.e. the booking shading). As P9 mentioned “I did a bit at the beginning, thinking that it would be more polite if I was more flexible, but actually, [.] – nobody came back to me and said, ‘Can you change this?’”. As most participants understood bookings as ‘fixed’ and immutable after being made, the indication of flexibility as a shading on the booking representation seemed to be superfluous information.
All participants but one (P10) claimed that they felt like they were collaborating rather than competing with the other club members. P10 mentioned that it felt a bit like ‘coopetition’ (competition alongside cooperation) and they argued that ‘a bit of gamification doesn’t hurt’ among neighbors to make them more motivated. P8 and P7 also mentioned experiencing some ‘self-competition’, in which they challenged themselves to keep their incentive (P8) or energy consumption (P7) at higher or lower levels accordingly.
As per our design approach, the SolarClub interface did not not reveal how much each individual household was consuming, nor did it by default include the author of messages, notes or bookings. As such, almost all participants felt the system was privacy preserving. “It felt, to me it felt very private in that if you, when you saw someone else’s booking, if you clicked on it, you didn’t see who it was or what it was.” (P11). While P11 and P17 felt that this was sufficient and did not want more information, all the others claimed that they would like more transparency even when that meant that they would show more data about their household. “If anybody wants to know when I’m doing my laundry, they’re really welcome.” (P10). This feeling was supported by the observation that one participant also explicitly added their house number to their note and three more added their name, thus breaking anonymity.
Perhaps because of this lack of transparency, the graph provoked inferences about what the other Club members were doing although to a lesser extent than we anticipated. “There were a couple of early morning big power surges, actually, we were wondering what they were doing. ”(P5). In other instances this kind of guesswork was triggered by the messaging and notes function:
Because I just know P12 and I just felt it was her. “Oh, I can do that,” or, “I’m going to try and do that,” or, “I’m going to try and engage with that.” I know that P13 or his son wouldn’t talk like that – little messages – these few messages that came through. Maybe I’m wrong. (P15)
Moreover, discussions about accountability were quite minimal as participants mentioned they did not go back to check if people had actually followed through with their booking (P2). In fact, participants suggested that this kind of feature would be relevant to have in future iterations.
5.4 Building an Energy Community
A core recurring theme across interviews was the notion of community and collectivity. Participants had some preconceived ideas of what constitutes a community that went, in this case, far beyond the idea of sharing renewable solar energy. For P10 for instance, “a community communicates” either in person or through tools like WhatsApp. For P17 a community was the council estate in which they lived on and less a group of people sharing energy. Community was also associated with feelings of motivation and accountability as “[communal living] probably would be quite a good basis for chat and improvement and getting together on things, and saying, ‘Now look, come on Number 54, you can do a whole lot better than this. What’s the problem? Why can’t you use the stuff when it’s daylight? Can we help?’”(P12).
Despite their ‘success’ (as summarized in Table
2), participants mentioned that it was hard to judge how well their energy community was performing. “So I don’t know. I mean, it’s difficult when you’re one household sitting in front of a terminal, to know – to judge how well the whole thing has done. You’re probably in a better position to judge that than we are.” (P7). This lack of intuition was also related to the amount of communication that took place among households.
I’m not sure we did very well on the trading, because maybe we could’ve done a whole lot of messages. [.] I think individual people made choices. We did. I’m assuming others did. But I don’t know to what degree, if at all, we corporately made choices. (P7)
As mentioned in the quote, participants overall felt like they were working as individual households while keeping in mind what other households are doing somewhat redefining what collaboration could look like. “I think we probably approached it in an individual way, knowing that other people were doing – assuming that other people were doing the same thing.” (P9). “Well, I knew there were others because I looked on the scheduling tool, but we never spoke.” (P10). “I was a bit selfish about it. I tended to look at what I wanted and whether other people have booked it rather than looking at the whole team. I kind of thought oh wow this is blocked I won’t use it but I suppose that’s still working as a team in a way.” (P17).
For most, these experiences contrasted with their existing understanding of what ‘community’ entails, and even with what they were expecting from an energy community experiment such as the one they participated in. “It wasn’t too much about the community. We were very autonomous and self-sufficient” (P8). This lack of friction during the household coordination even left some feeling as if the other participants were non-existent. “If I book those times and somebody else also wanted to put those time then there would need to be an agreement between us, but I would move to a different time or a different day but that never happened so I was able to do everything I wanted to on the day I wanted at the time I wanted. I didn’t see anybody trying to book anything at the same time. [.] There was no friction at all, because I couldn’t see that anybody was trying to book any.” (P16).
This notion of community as interpersonal connection was also reflected in how participants used the communication features within the SolarClub coordination interface. Booking notes were the most popular option, seemingly related to their personal nature (“I like the notes. I found that... That made it quite personal, you know.” (P5)) and they ranged from brief explanations of the activity to longer argumentation of why it will take place (e.g. “I can watch the TV Later in the day but my favorite show is on can’t miss that”(H10)).
On the other hand, the Message Club feature wasn’t felt to be as useful as participants “never felt [they] needed to [message]”(P11). P1 and P7 attributed this lack of usefulness to the perceived importance of the message received: “Well, I saw the odd message pop up, but then you just thought, ‘Oh, why? Oh.’ And some of them [.] – when you read them, they weren’t like the world’s going to end kind of ” (P7). For P17 the dislike of the messaging tool related to the lack of response by the other participants indicating the importance of two-way communication. “ [.] that was the only message I sent and so for me that was like I didn’t use it again when I thought I hadn’t got a reply I thought well they can’t be bothered so I won’t.” (P17). P15 also mentioned their saturation on the modes of communication and not needing a new one “about cooking or washing”(P15).
During the first group deployment, all four households mentioned separately they would have liked an initial workshop so as to get a ‘shared understanding’ (P1). However, even when we organized a brief workshop for the subsequent two groups (online, to simplify participants’ attendance), it still did not seem to fulfill the expectations and desires of some for collective discussion “When we were doing the Zoom, I thought we were going to discuss that more, and I was up for really – I thought we were going to have to do an agenda. Do you know what I mean? Or [take out] a calendar and say, ‘OK, does anyone want this?’”(P8).
5.5 Simulation, Imagination and Empathy
Related to the desire among many participants to know more about who was making bookings and why, participants frequently imagined how they might respond towards under particular circumstances – from their neighbors hosting a dinner party, to neighbors living with a disability. In some cases, these imagined scenarios prompted empathy and a desire to be accommodating, but in other cases, they were envisioned as occasions where energy communities could break down.
The simulation aspects in our study, seemed to have helped some of our participants reflect about what it would mean to live with only renewable energy and how their practices would need to change, sometimes radically. P4 and P8 ran-through a scenario in which renewables such as solar were the only source of energy available. “And then I was like – it was slowly dawning, well, how am I going to eat my dinner? [.] So I would even have considered, there’s different ways, perhaps lunch would be my main meal or… Do you know what I mean?” (P8). P4 who had mentioned how they found this experiment ‘restrictive’ equally explained how they considered such kind of restriction necessary in order to change society. “I think that, you know, if we were in a situation where, as I said earlier, where this wind and solar were the only things available, then yeah, I would accept that that is restrictive and that would mean even, I suppose we would as a society, society then have to change. And I’d be accepting of those things and engage with them in order to make it work.” (P4).
Practices of imagination not only pertained to ‘extreme’ scenarios, but also to the changes required in daily life. For instance a common reflection was considering the effect of trying to coordinate during the winter months when there is less solar. Daily activities such as an immediate need of electricity or then simply a dinner party were also presented as occasions were energy communities could break down “I was only thinking, if you had a dinner party, it’s bad enough trying to find a date for when everybody can come together for a dinner party. Then working out and then having to communicate between people maybe too - well, you wouldn’t necessarily know how much energy there was.” (P2).
Besides making participants reflect on alternative personal scenarios, the deployment also seemed to generate more empathy towards others, within and beyond the club, either through direct knowledge of others’ circumstances or through hypotheses. “Families of four must be really going through it. And it’s reminded me – living here on my own – what a low-energy user I am compared with people who’ve got multiple people under the same roof.” (P10).