*2.1. Research Design*

To answer the research question, a new analytic framework was first designed that decomposes both energy communities and UBEM tools to information entities relatable to each other, namely: EC progression factors and capabilities of UBEM tools. Then, a two-tiered systematic literature was conducted into the research of EC and UBEM, respectively, to collect these information entities. EC use-cases and lifecycle phases were identified to answer research sub-questions. Finally, the results were matched to see the potential interactions between UBEM and EC as the EC-specific affordances of UBEM (Figure 3).

**Figure 3.** Overview of research process.

To meet the research objective of providing a manual for EC practitioners, the EC analytic framework must consist of features corresponding to disaggregation requirements of the main research question, namely: different use-cases, a breakdown of typical life-cycles, and a collection of progression factors. In the framework, the features for the analysis of UBEM are the progression factors, while use-cases and lifecycle phases are structural metadata assigned to the factors. The working definition for progression factor in this study is any condition that is indicative for the progression of energy communities through their lifecycle. Progression factors were extracted from the reviewed articles and labelled by which lifecycle phase and which use-case they are relevant for. This labelling was essential to answer research questions 2 and 3. Apart from essential structural metadata, supplementary labelling schemes describing the importance of each factor and the discipline with highest authority in them were added. The supplementary metadata were chosen to support the practical objective of providing a manual for EC planners, and were selected due to availability of information, based on a preliminary review of the literature. All categories in essential and supplementary metadata were defined from the literature, and not top-down—meaning alternative categorizations are valid. With four distinct categorizations, the factors were analysed on their relationships to each other by

inspecting pairwise correlations among categories and by agglomerative hierarchical clustering using the UPGMA algorithm [76] (Supplementary Materials). The reason for inspecting this relationship is to validate whether the categories are redundant, and to see how factors can be bundled together for communication to EC practitioners.

The UBEM analytic framework was an amalgamation of the frameworks of four most prevalent UBEM tool reviews [22,61,62,77]. The features used to describe and classify UBEM tools in these four articles were taken as UBEM capabilities for coupling against EC progression factors. However, not all UBEM approaches were considered relevant for the research question. Generally, UBEM can be classified into two distinct approaches: top-down and bottom-up [57–59], complemented by hybrid approaches which combine the two [59,60,78]. Top-down modelling was excluded from this study, as they are incapable of modelling complex scenarios in energy transitions due to their reliance on aggregated historical statistical data and they are not able to consider different energy saving measures in different spatiotemporal situations, or occupancy types. [78,79].

Categorisation of UBEM tools evolved naturally as they became more mature and acquired more functions. Analysis of frameworks from four [22,61,62,77] of the collected review articles were used to define a comprehensive analytic framework.

For the analytic framework of this study, all features of bottom-up and hybrid modelling tools that were present in more than one article were automatically retained; the rest went under a systematic preselection process (see Table 1 for list of selected and filtered features). Features were excluded if: (1) they were duplicates, or included in the other features (e.g., exo- or endogenous demand modelling, Impact of user behaviour, Time horizon); (2) does not correspond to the working UBEM definition (e.g., building stock location, building characterization); (3) there is not sufficient information in the reviews and the original articles of the tools (e.g., non-residential type of building, input type).


**Table 1.** List of selected and filtered urban building energy modelling (UBEM) tool features.

1 duplicates, or included in the other features, 2 not correspond to the working UBEM definition, 3 insufficient information in the reviews and the original articles of the tools.

However, not all analysed tools were present in all four reviews. The original articles of the included tools were used to fill in missing features to avoid information gaps. All, but 12 feature-tool couplings were filled this way. As in the case of EC progression factors, a typology of tools based on these features was produced via agglomerative hierarchical clustering, using the UPGMA algorithm [76], and a pairwise correlation matrix was produced. Again, this exercise was used to see whether there is a useful categorisation of tools, and to test whether the UBEM capabilities (the final list of features) are all necessary.

Finally, the relationship between EC and UBEM was justified through the concept of affordances, using EC progression factors as key performance indicators for the UBEM capabilities. Constructing an affordance is not defining a single feature but is discussing the way a certain user appropriate features for goal-oriented actions [80]. At its core, an affordance is the dynamic between features and actions, that is: the feature, a set of actions made possible by that feature, and the way the feature facilitates those actions [81]. This also means that an affordance is situational, as facilitation would occur for certain users, in certain contexts, which must be specified to justify an affordance [82]. Finally, the actions afforded must relate to the goal of the user [73,74].

Therefore, these six components, features, actions, facilitation, user, context and goal must be present to define an affordance (Figure 4).

**Figure 4.** Components defining technological affordances.

In this study, the EC progression factors, and their metadata define the context and goals, while UBEM capabilities refer to features. Taking the EC planner as the user, an affordance can be specified by describing facilitation, and actions that are linked to user goals. Thus, an affordance exists if (1) in the context defined by EC use-case and lifecycle, (2) for an EC planner or project manager as a user, there is a (3) set of UBEM capabilities that (4) facilitate (5) a set of actions to (6) reach the goal of the user defined by meeting one or more progression factors. If the six components are justifiably present, the affordance exists. As progression factors were already classified by use-case and lifecycle phase, affordances could also be examined against both. This provides the necessary disaggregation for the research sub-questions, while a collection of justified affordances is the answer to the main research question.

## *2.2. Data Extraction*

To extract information for analysis, two secondary data sources are investigated: review articles and case studies of energy communities. Thus, the feature extraction phase consists of an overview of reviews with the scope of review articles and meta-analyses and a systematic review with a scope of case studies. This distinction is chosen as recent reviews will not cover recent case studies, and because case studies are expected to yield more information on project lifecycle, while reviews are expected to give a better overview of EC use-cases. Both data sources are expected to return progression factors. The review of UBEM tools will rely on the secondary data source of UBEM proof of concept studies. The second phase of the research is also a systematic review, with the scope of UBEM tools.

The Scopus database was chosen for publication selection for its larger share of unique citations in both social science and engineering citations [83]. For the overview of reviews, the search term ("energy community" OR "community energy" OR "energy cooperative" OR "citizen energy") AND "review" was used for review. The time period for search was set to 2018-2020 at first, with annual extensions planned if insufficient information would have been generated—however, database was saturated with the first batch, see paragraphs below. For the review of case studies, ("energy community" OR

"community energy") AND "case study" was used for the same years. For the review of UBEM tools, "urban" AND "energy" AND "model" was used for years between 2015 and 2020.

In all cases, the articles went through a preselection process (Figure 5). Titles were checked for overall domain relevance, while the abstracts were read for a relevance of the narrower topic. For example, an article on nutrition [84] was filtered out in the first step, while another [85] was filtered out in the second step, as it was a study on energy, however not on community energy as per definition. In addition to relevance check, articles had to be reviews, case studies and original articles introducing new UBEM tools, respectively. Finally, the following exclusion criteria were defined: the geographical scope is limited to the EU/US and cases where EC is the only viable alternative for energy distribution due to various constraints such as remote communities. These criteria were applied to meet the practical objectives of the study.

**Figure 5.** Selection process of references disaggregated to data sources.

The final articles, after confirmation in the text, that they contribute to study results, were selected in a bottom-up manner, by defining a threshold for data saturation. The data saturation threshold is a way of determining after how much articles does the research become redundant [86]. In studies, where the task of the observations is to provide new labels or classes, there is a characteristic saturation curve, plotted as the number of observations against the quantity of new labels accumulated with each observation. This curve is steep for the first "n" observations and flattens out afterwards. A flat curve means that repeated observations will likely not yield new labels, the database is saturated. Data saturation curves were used for the overview of reviews and the review of UBEM tools (Figure 6). The observations were the articles read, and the labels were the progression factors and the UBEM tools respectively. The threshold conditions for saturation were set as the difference quotient for observations On − On−3; On − On−5; and On − On−10. The three ranges were chosen to decrease sensitivity to small-scale disturbances, and to set a minimum number of articles to be read.

The actionable set of articles included 25 EC reviews, 18 EC use-cases and 12 UBEM articles. Saturation for UBEM tools was reached at 12 articles and 43 tools. A total of 34 out of 2115 search hits were assessed before saturation, out of which 12 was preselected. Due to lack of information or failure to meet UBEM/USEM definitions, 21 tools were excluded. The 22 remaining tools were further analysed. For the progression factors, saturation was reached at 20 articles, with 49 progression factors. Including preselection, this meant that 108 of the 126 articles were processed, out of which 41 was preselected by domain relevance and 33 by topical relevance. A total of 8 unique use-cases were identified, supplemented by hybrid use-cases as one category. However, 5 use-cases were considered

for use-case-specific progression factors, due to lack of data, or conflict with the definition (see Section 3.1). Three use-cases, community choice aggregation, microgrids and green neighbourhoods were not considered for unique progression factor extraction.

**Figure 6.** Log of data saturation: (**a**) Number of accumulated progression factors plotted against number of energy community (EC) reviews on the left; (**b**) number of accumulated UBEM tools plotted against number of UBEM reviews on the right.

#### **3. Results**

The results are presented following the logic of affordances: Section 3.1 introduces the different identified use-cases for ECs (corresponding to affordance context), Section 3.2 describes phases of a generalized EC project lifecycle (affordance context), Section 3.3 collects EC progression factors (affordance goal) and examines according to the selected structural metadata, Section 3.4 constructs affordances from UBEM features (affordance capabilities), and disaggregates them as per the research questions.

#### *3.1. Energy Community Use-Cases*

The following use-cases were identified during the study: renewable energy production, peer-to-peer energy market, demand-response providing community, bulk investment in energy conservation measures, community choice aggregation, collective grid ownership and community energy storage. Additionally, the green neighbourhood is a special case not strictly an EC, and three hybrid use-cases were also identified.

Three use-cases, community choice aggregation, microgrids and green neighbourhoods were not considered for unique progression factor extraction. Community choice aggregation (CGO) allows cities or other local government units to aggregate customers within their jurisdictions and to procure energy for them, either through contracts or through ownership of generation [40]. Although the subject of the article was a comparison between community choice model and renewable energy community, the former is not in fact an energy community, but a pooling of consumers under a single trusted intermediary to bargain on their behalf. Collective grid ownership is a valid EC, however it never appeared on its own in the literature but integrated to one of the other use-cases [45,49]. Finally, green neighbourhoods are unique models targeting complex sustainability goals on the neighbourhood scale [32,55,56]. These ambitious projects are often government-funded flagship projects or experimental niches, or unique market niches of bottom-up initiatives. They operate partially or fully on a combined waste-water-energy nexus, seeking to leverage all three circulations to close loop, essentially leaving behind no waste, wastewater and taking in no energy from external sources [55,87]. Green neighbourhoods can be models for energy communities on the long-term,

but the multiplicative impacts of such projects scale at the cost of scaling their complexity, operational and investment costs. Given that green neighbourhoods fit the definition of multi-purpose energy communities, any universal EC progression factor will also be valid for them as well.

From the core investigated use-cases, the energy conservation investment community (ECC) pools resources to bulk invest in interventions reducing their energy consumption, such as purchasing energy-efficient appliances, retrofitting the building envelope, replacing windows or multiple of the above in deep retrofit projects (Figure 7). The reduction of operational costs makes energy conservation communities relevant for ESCO financing, and the contribution to decarbonisation by demand reduction are measurable contributions to government sustainability policy [33].

**Figure 7.** Energy conservation investment community use-case. Dashed line denotes the community.

Peer-to-peer energy markets (P2PM) rely on community microgrids to sell and buy electricity produced locally in distributed plants, and externally if the microgrid is connected (Figure 8). Regarding market structure, the model can either be a full market, a decentralized market or a hybrid solution in between. The main difference between full and decentralized markets is the lack/presence of a community representative, who acts as intermediary for both internal market governance and as a medium between the decentralized community and the energy market. In hybrid models, members may join individually, or through the representative [53].

**Figure 8.** Peer-to-peer energy market use-case. Dashed line denotes the community.

Renewable energy communities (REC) are the most well-researched use-cases. Historically, decentralized renewable energy communities kicked of the social innovation niche, buy pooling resources to invest in energy production (Figure 9). In a decentralized model, there is a single or

several power plants for the entire community, which can both be utilized to meet the demands of the community itself, but also to sell excess energy if they have access to wholesale markets. In a distributed model, the community functions more akin to energy conservation communities, to bulk invest in renewable production on household level [42]. Distributed production, alongside with microgrid ownership, are constituent use-cases for peer-to-peer energy markets.

**Figure 9.** Renewable energy community use-case. Dashed line denotes the community.

Demand-response energy communities (DRC) are one of the more novel, experimental use-cases that stem from the mainstreaming of renewable production and distributed energy production. Since both trends have threats to maintaining grid balance, a community offering to harmonize load curves en masse is a viable service for grid operators (Figure 10). This involves consumers in the management of stable grids, therefore such interventions are labelled demand-side responses. The key strategies for managing load curves is through changing consumption habit, with load shaving meaning changing consumption amounts at certain times of a days, while load shifting is offsetting supply to other uses.

**Figure 10.** Demand-response community use-case. Dashed line denotes the community.

Community energy storage (CSE) is a modular, scalable, virtual energy storage built up from a grid of distributed storage units owned by community members (Figure 11). On their own, virtual storage communities can offer flexibility services like demand-response communities, while in combination with local production, they may serve as buffers to locally produced surplus. Community storage, if place-based is its own category in terms of scale (tens to hundreds of kWh) and is not a substitution for large storage with capacity levels in the MWh-GWh scale. Most common technologies include lithium-ion batteries, lead-acid batteries, flow batteries and more recently hydrogen for electricity, and water tanks and phase-change materials for thermic energy.

**Figure 11.** Community energy storage use-case. Dashed line denotes the community.

Hybrid use-cases are trending due to the associative nature of the use-cases and the potential of stacking services [88]. We have seen before that renewable energy communities can evolve into peer-to-peer markets, but it is also possible for prosumers to build a diverse grid stabilisation portfolio, by having both demand and supply side options (Figure 12). Renewable production can mitigate or eliminate undersupply, while community energy storage can act as a buffer in case of oversupply. Energy conservation measures are supplementary to all other use-cases, expanding external markets of any local production use-case by decreasing internal demand [46]. It is also an option for renewable energy communities to distribute some production and assets, and keep others at community level, while also acquiring energy storage to provide full stack energy services internally [14,89,90].

**Figure 12.** Hybrid renewable-demand response community use-case. Dashed line denotes the community.
