Energy Economics 150 (2025) 108822

Contents lists available at ScienceDirect

Energy Economics
journal homepage: [Link]/locate/eneeco

Power Trading within an Energy Community: Applying a fair unequal

sharing rule of photovoltaic energy
∗
Nicolas Boccard , Renan Goetz
Economics Department, University of Girona, Spain

ARTICLE INFO ABSTRACT

Keywords: Energy communities (ECs) with photovoltaic (PV) installations can reduce their collective electricity costs by
Energy community sharing self-generated power among members. This paper presents a comprehensive framework to address key
Peer-to-peer trade challenges such as rooftop sharing, heterogeneous ownership structures, and electricity trading protocols. We
Grid stability
propose the Weighted Constrained Equal Awards (WCEA) mechanism’s proportional and fair allocation rule
Sharing rules
that distributes the benefits of internal energy trade. Unlike existing approaches, the WCEA accommodates
Social choice
Smart City
diverse ownership arrangements and varying consumption patterns. This ‘‘fair-unequal’’ protocol promotes
Photovoltaic both the formation and long-term stability of ECs and is robust to strategic manipulation. We evaluate the
WCEA using high-frequency metering data in a Monte Carlo simulation. Results indicate that a sufficiently
large PV system can reduce median energy costs by up to 65%. Enabling peer-to-peer energy trading yields
additional savings: approximately 2.5% when consumption patterns are similar, and up to 11% when they are
highly heterogeneous. Moreover, shared PV generation reduces reliance on grid electricity, offering ancillary
benefits to the distribution network. However, large-scale PV installations may also stress the grid due to
temporal mismatches between generation and consumption. These findings suggest that policy incentives
should prioritize installations that currently satisfy less than half of the prosumers’ energy demand.

1. Introduction for condominium residents, the PV installation process is more com-

plex due to the collective ownership of rooftop space. Approval must
The European Union has set ambitious targets for the uptake of be granted by the homeowners association, which typically requires
renewable energy resources (RES). A key priority is the expansion of presenting a compelling proposal in a general meeting. Nonetheless,
rooftop photovoltaic (PV) installations beyond single-family homes to- compared to individual prosumers, ECs offer a key advantage which is
wards multi-family units which are prevalent in the densely populated the ability to further reduce electricity costs by coordinating energy
urban regions (aka condominiums).1 To support this transition, Direc- flows internally. Indeed, when one resident is a net producer while
tive (EU, 2018) (REDII) establishes the Energy Community (EC) as an another is a net consumer, energy can be traded locally among them
autonomous legal entity governed by its members. ECs are entitled (aka peer to peer exchange), enabling cost savings by avoiding the
to produce, consume, store and sell renewable energy. Membership is traditional grid interaction–whereby electricity is typically bought at
open to individuals, small and medium-sized enterprises (SMEs), and a premium but sold at a discount. Recall that such an opportunity is
local authorities located in close geographical proximity to the EC’s unavailable to detached home prosumers who are solely connected to
activities. Crucially, the objective of the EC is to offer environmen- the grid and who therefore seek to consume or store their PV output.2
tal, economic, or social benefits to its members rather than seeking In contrast, direct exchanges between condominium residents open new
financial profits (Art. 16 & 22). avenues for cost savings all the while minimizing exchanges with the
When households install rooftop PV panels, they become prosumers grid.
with a view to save on power bills and (quite often) contribute to In this article, we leverage a comprehensive, high-frequency dataset
environmental sustainability (cf. Kocakusak et al., 2024). However,
to evaluate the potential net gains from establishing an EC in a typical

∗ Corresponding author.
E-mail address: [Link]@[Link] (N. Boccard).
1
The condominium is a threefold legal concept, common to many jurisdictions, encompassing individual apartment ownership, joint ownership of land and
shared building infrastructure as well as membership in an owners association. In North America, common property may also include roads and amenities
(cf. Rosen and Walks, 2015).
2
This behavior may lead to a rebound effect, resulting in oversized PV systems (cf. Boccard and Gautier, 2021).

[Link]
Received 30 May 2025; Received in revised form 3 July 2025; Accepted 8 August 2025
Available online 16 August 2025
0140-9883/© 2025 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license ([Link]
N. Boccard and R. Goetz Energy Economics 150 (2025) 108822

condominium. We start by addressing the necessary steps to create an analysis remains theoretical, with no empirical validation. Building on
EC as its formal establishment requires navigating a range of practi- a different game-theoretic reward distribution scheme, Fioriti et al.
cal and institutional challenges. Key issues include allocating limited (2021) determine the optimal size of an EC in the presence of an
rooftop space, selecting an ownership model (individual, communal aggregator and energy service companies.3 Nevertheless, their reliance
or mixed), and deciding whether to purchase or lease the PV system. on simulated irradiance profiles and 15-min load data granularity may
Further stumble-blocks are the design and management of internal limit the real-world applicability of their findings. Gjorgievski et al.
electricity trading—such as tracking power flows, accommodating the (2023) propose a sharing mechanism known as virtual net billing to
entry and exit of members, and deciding whether metering is required allocate cooperative benefits from energy trading. The virtual net
to implement a trading scheme. billing approach guarantees that each member can consume a share
In dealing with these obstacles, we make three contributions to the of the PV output equivalent to their investment share. However, any
literature. First, we develop a blueprint (detailed design) that outlines total excess demand of the community is allocated among the members
strategies for addressing the key practical and institutional challenges in proportion to their individual excess demand relative to the total
involved in the creation and management of energy communities. In excess demand. In this respect, while virtual net billing ensures that
doing so, we support the REDII directive’s call for ‘‘fair, proportionate, each member receives their entitled share of PV output, the allocation
and transparent procedures’’ (Art. 22.d). mechanism is ultimately driven by differences in energy consumption
Second, while existing sharing rules rely on a single criterion—such rather than by ownership structure. Similarly, one can also allocate
as cooperative benefits, individual electricity consumption, or owner- excess supply. Hence, the outcome of the allocation may be perceived
ship shares in the PV installation—we propose a novel mechanism that as unfair, as it is determined solely by each member’s consumption and
ensures a fair but differentiated allocation of the benefits of internal generation relative to the community-wide totals. This approach over-
trade of the surplus and deficit of the PV output. The rationale for looks potential mismatches between a member’s energy needs and their
our proposed sharing rule jointly considers each member’s ownership ownership share in the community’s generation assets. As a result, it
share and actual energy consumption (cf. Section 2.1). Our scalable, may discourage members from investing in solar capacity that matches
transparent, fair and automated protocol is based on the weighted their actual consumption, and instead incentivize underinvestment in
constrained equal awards (WCEA) rule from the theory of social choice. order to benefit from surplus energy generated by others.
As demonstrated in our companion paper (Boccard and Goetz, 2025) The study by Bossu et al. (2024) formulates an optimization prob-
(for short B&G), the WCEA rule satisfies several desirable properties lem with a detailed representation of household appliances. However,
defined in the economic literature. In particular, it is robust against their model does not incorporate real data on load or solar irradiance;
strategic manipulation, ensuring no member can take advantage of the instead, it is calibrated using three representative days throughout the
mechanism at the expense of others. Moreover, the data and technical year. The authors apply cooperative game theory concepts, specifi-
infrastructure required to implement the WCEA are minimal, so that cally the Shapley value and the Nucleolus, to allocate the cooperative
implementing it should not encounter significant barriers. gains of the energy community (EC) among its members. However,
Our third contribution is empirical. To assess our EC formation these methods require a solid mathematical background to fully un-
blueprint, we assemble a distinctive data set of real-world observations derstand its mechanism, which may hinder their transparency and,
to estimate the cost savings of an EC where energy is exchanged consequently, reduce their acceptability among stakeholders. Berg et al.
through the WCEA rule. We compare our setting with two alternative (2024) explore the potential cost reductions for industrial and residen-
scenarios: one without photovoltaic systems (‘‘no PV’’) and the other tial consumers who engage in collaboration. Their analysis shows that
one with PV but without internal trading (‘‘no trade’’), the latter effec- the cost savings from energy exchange are relatively modest and the
tively simulating the case where households reside at the same location Shapley value yields the fairest distribution of cooperative costs and
but in detached houses. Unlike many empirical studies that rely on benefits. In addition to relying on cooperative game theory, whose com-
a single representative load profile, our analysis incorporates a very plexity can make the sharing mechanism less transparent and harder for
wide range of profiles reflecting variation in age, occupation, household all stakeholders to understand, the approach is also constrained by the
composition, and geographic location (solar irradiation). In addition, need for costs and benefits to be known in advance in order to apply
we vary a key parameter which is the ratio of available rooftop space the model effectively.
for PV installation relative to the building’s total energy demand. This Beyond the design of empirical approaches or the choice of the
allows to speak of the relative size of the PV installation and evaluate underlying theoretical frameworks, the literature reveals further limita-
its influence on the benefits of power trading. tions related to the sharing rules applied. As we show in our companion
In Section 2, we review the literature and explain the underlying work B&G, none of the widely studied sharing rules satisfy all five
concept of fairness of the WCEA. In the following section, we present normative criteria derived from social choice and game theory. Their
a step-by-step blueprint for the creation of an EC. Section 4 details the shortcomings stem from relying solely on either ownership structure
mathematical concept of the WCEA as well as the calculations involved or consumption patterns, rather than integrating both. The widely
in its numerical implementation. Section 5 presents the results of the accepted criteria are efficiency (no member receives a higher share than
empirical part of our study. Based on high-frequency real-word data, it their consumption), individual rationality (each member benefits from
determines the cost savings of PV systems of different sizes with and participating in the EC), non-manipulability (strategic behavior aimed
without trade. The last section concludes. to gain a larger share is not rewarded), replacement monotonicity
(when members join or leave, all shares adjust in the same direction),
2. Literature and fairness (the rule can reflect a collectively agreed upon notion of a
fair outcome). These criteria are essential for fostering energy commu-
Scholarly work on energy communities, as summarized in the re- nities, as they increase peer confidence in the fairness and reliability
view by Soto et al. (2021), highlights a broad spectrum of promising of the sharing rule. B&G prove formally that the WCEA mechanism
avenues for future research. This outlook is reinforced by a substantial belongs to the family of sequential sharing rules, and therefore satisfies
and growing body of literature. For a comprehensive analysis, we refer all normative criteria except fairness. Indeed, a sequential rule is not
readers to Schwidtal et al. (2023), who review 135 studies and identify
consumer-focused models as the most frequent.
With regard to energy exchange within energy communities, Tushar 3
An aggregator is an automated device or business entity that consoli-
et al. (2018) conceptualize internal trading as a coalition game and dates the energy resources of multiple ECs to facilitate efficient trading and
apply the concept of the core to evaluate stability. However, their participation in broader energy markets.

2
N. Boccard and R. Goetz Energy Economics 150 (2025) 108822

inherently fair, as this depends on the specific concept one adopts. Consider then a building with available rooftop space to install
Now, establishing a fairness criterion remains key to enhance the broad PV panels. Condominium owners must initially determine whether to
acceptability of a sharing rule, which, in turn, underpins the success- become members of the EC. The resulting 𝑛 members community will
ful formation of new ECs. Section 2.1 thus scrutinizes the fairness then need to establish a method for distributing the investment and
properties of the WCEA. Beyond fairness, the WCEA’s ‘‘replacement maintenance costs, along with the generated energy and the potential
monotonicity’’ also contributes to the long-term viability of a sharing revenues from selling excess energy to other community members or
rule by guaranteeing that the entry or exit of members does not result their distribution network operator (DSO).4 In most countries, the law
in preferential treatment for existing participants. specifies that the flat area of (square footage) determines the unit en-
Our review of the literature has focused on energy sharing and titlements (𝛼𝑖 )𝑖≤𝑛 used to distribute the communal costs of the building
revealed that various of the proposed mechanisms are quite com- and also to weigh the votes over expenses, budget and projects such as
plex, particularly with respect to the optimization procedures, data a major renovation of shared components of the building.5
requirements, and the burden placed on participants’ time and (mental) The REDII directive suggests flat owners should have the option to
training effort. In addition, these studies are often paired with an participate in an EC at least up to their entitlement 𝛼𝑖 and that national
empirical analysis that is not based on actual real-world data, or one legislation should allow them to temporarily withdraw and later rejoin
where the granularity of the meter readings is insufficient to precisely the EC. In the first module of our blueprint, the allocation of PV system
capture all energy transactions among the members of the EC. The ownership is organized among the EC members.
objective of our study is to fill these two deficits in the literature.

2.1. Fairness and sharing rules

#1 EC constitution Being knowledgeable about forthcoming
stages, owner 𝑖 requests a PV share 𝛽𝑖 ∈ [0; 2𝛼𝑖 ]. The final share
Should the EC adopt a fairness notion prioritizing solidarity and
is denoted 𝛾𝑖 ≤ 𝛽𝑖 . The under-subscribers 𝐽 with ∀𝑗 ∈ 𝐽 , 𝛽𝑗 ≤ 𝛼𝑗 ,
simplicity, it could very well choose to allocate the PV output equally
receive their claim i.e., 𝛾𝑗 = 𝛽𝑗 . The surplus of roof space
among its members, regardless of their underlying‘‘shares’’. If how- ∑ +
they free, 𝛿 = 𝑖 − 𝛽𝑖 ) , is )
𝑖∈𝐽 (𝛼( distributed to the remaining
ever, fairness is instead understood in terms of individual merit or
𝛿
contribution, the EC may opt for the Shapley value, which reflects participants as 𝛾𝑖 = 𝛼𝑖 1 + ∑ for 𝑖 ∉ 𝐽 whenever 𝛿 > 0.
𝑘∉𝐽 𝛼𝑘
each member’s marginal contribution, whether the latter is based on Final shares (𝛾𝑖 )𝑖≤𝑛 determine future voting rights within the EC.
entitlements or self-production. Alternatively, one might frame fairness
as the minimization of grievances or the perceived unfairness among
subgroups. If so, the EC might favor a refinement of the core such as
the nucleolus, which seeks to minimize the maximum dissatisfaction In this first module, we allow members to request a PV ownership
any coalition of members might express. Now, and this is our key share that may be equal to, greater than, or less than their entitlement.
objection, none of these three allocation rules simultaneously satisfies To limit strategic behavior, we propose introducing an upper limit on
the five normative criteria derived from social choice and cooperative individual claims.6 Finally, owners who do not want to participate in
game theory, while also maintaining a high degree of transparency and the EC can opt out with a nil claim.
avoiding excessive complexity. Our second blueprint module provides guidance to determine the
Not only does the WCEA checks all 5 properties but it is also aligned size and type of the PV installation. To this end, factors such as the
with the European directive’s proportionality principle. It guarantees total capacity in kWp, the direction and tilt of the panels which affect
that EC members receive a share of the PV output proportional to their daily and seasonal exposure, a maintenance agreement with a
their individual investment in the PV system. Additionally, whenever specialized company for up-keeping and repairs, and the type of cell
a trading opportunity arises among members, its economic value is technology, contrasting older, cost-effective options with newer, pricier
equally split between the seller and the buyer (see blueprint module but more efficient alternatives, are considered.
#3 in the next section). Importantly, the probability of being selected
to participate in these beneficial trades is also proportional to each
member’s investment share in the PV system (entitlement) and the
member’s self-production rate. Under the WCEA, internal trading pro- #2 PV system The EC contracts a technical consultant to develop
motes efficiency, as energy is allocated strictly according to demand, and propose different PV systems. A vote, weighted by rights
preventing any excess or waste. It also promotes fairness by considering 𝛾𝑖 , is taken. The proposal with the fewest votes is eliminated; if
both the members’ entitlements and their self-production rates, that necessary, the consultant breaks ties. Voting rounds continue until
is, the extent to which they are net contributors or net beneficiaries. a single proposal remains. Each member 𝑖 finances the share 𝛾𝑖 of
Members with higher self-production rates have their claims prioritized the investment cost for the chosen system. Additionally, members
and are more likely to receive their full allocation than those with lower contribute to the annual maintenance costs in proportion to 𝛾𝑖 ,
rates. and are entitled to claim a share 𝛾𝑖 of the resulting electricity
The WCEA is also flexible insofar as one may temporarily alter production.
the priorities to reflect a special circumstance (e.g., six-month work
absence or short-term increase in household size). Concluding, the
fairness concept underlying the WCEA is neither based on strict pro-
The total investment cost includes photovoltaic panels, mounting
portionality, nor needs nor no-envy but a flexible and transparent rule
structures, an inverter, smart meters for each flat (assuming they are
which allows the Energy Community to tune it in accordance to the
idiosyncratic fairness concept held by its members.
4
The DSO, a regulated monopoly, maintains the physical connection
3. Setting-up an energy community
between buildings and the grid, reads meters and emits bills. Thanks to
deregulation, households can now contract power from a retailer rather than
We present here a blueprint (detailed design) solving the chal- their DSO.
lenges to the establishment and management of an energy community. 5
cf. Çağdaş et al. (2020) for a comparison of several legal regimes and
Our sensible proposal can reduce transaction costs and save the valu- their subtle differences.
able time that is often spent negotiating agreements in meetings or 6
This constraint significantly reduces the potential for manipulation in the
developing a trading approach in markets. allocation process.

3
N. Boccard and R. Goetz Energy Economics 150 (2025) 108822

not already installed), labor, permits, and all essential cabling within
the building, along with any required connection upgrades to the #4 Traded Quantities Within a time period where the PV system
distribution network. Optionally, the EC can offer users of neighboring generates 𝑧0 kWh, owner 𝑖 receives his share 𝑧𝑖 = 𝛾𝑖 𝑧0 kWh while
buildings the chance to either invest in a portion of itss PV system or consuming 𝑐𝑖 kWh in his own flat and thus has net residual power
integrate their personal PV system or energy storage solution within the demand 𝑟𝑖 = 𝑐𝑖 −𝑧𝑖 . Let 𝑉 denote the set of internal vendors (𝑟𝑗 < 0)
∑
community. The REDII directive explicitly considers this option since and 𝑣∗ = − 𝑗∈𝑉 𝑟𝑗 how much they offer. Likewise, let 𝐴 denote
∑
rooftop space is frequently limited in the dense urban environment; it the set of internal acquirers (𝑟𝑖 > 0) and 𝑎∗ = 𝑖∈𝐵 𝑟𝑖 how much
is therefore essential to facilitate the extension of the EC to neighboring they wish to acquire.
∑
buildings with free roof space.7 ∙ If 𝑣∗ ≤ 𝑎∗ , we solve 𝑣∗ = 𝑖∈𝐵 min(𝑟𝑖 , 𝛾𝑖 𝑧) in 𝑧 to get 𝑥𝑖 , the
The net-metering policy, which was popular with detached house PV output each acquiring member buys internally at price 𝑝. ̂ The
unmet demand 𝑟𝑖 − 𝑥𝑖 is acquired from the DSO at price 𝑝1 .
owners (cf. Boccard and Gautier, 2021) has been replaced in all ju- ∑
∙ If 𝑣∗ > 𝑎∗ , we solve 𝑎∗ = 𝑗∈𝑆 𝛾𝑗 min(−𝑟𝑗 , 𝛾𝑗 𝑧) in 𝑧 to get 𝑥𝑗 , the
risdictions by the net-billing scheme which introduces a grid tariff 𝑝1
PV output each vendor sells at price 𝑝. ̂ The excess supply of the
applicable to imports and a lower feed-in tariff 𝑝2 applicable to exports.8
PV system, −𝑟𝑗 − 𝑥𝑗 is sold to the DSO at price 𝑝2 .
For simplicity, our third module assumes that a single tariff applies
to all EC members, although more complex ones such as real time
pricing could be used without major consequences for the validity of
the proposed blueprint.
In scenarios #1 and #3 (𝑣∗ < 𝑎∗ ), the available PV power is
insufficient to meet the full demand within the building. The available
energy is distributed according to the WCEA rule among claimants
#3 Trade Pricing The EC assesses existing tariffs (𝑝1 , 𝑝2 ) offered by giving due consideration to the initial investments each member
by retailers and selects one via an elimination voting process. made in the PV system. In scenarios #2 and #4 (𝑣∗ > 𝑎∗ ), the PV
Owners may consume or sell their share 𝛾𝑖 of the PV output. power supply exceeds the demand within the building. This leads to a
Transactions between EC members are priced at the mid-market distribution problem, namely determining how much electricity each
( ) member can sell internally at the attractive price 𝑝̂ and how much
rate 𝑝̂ = 12 𝑝1 + 𝑝2 .
has to be sold at the lower feed-in-tariff 𝑝2 . The WCEA rule is again
employed to solve this distribution problem.
Our fifth module addresses financial settlements among EC mem-
bers who are virtual owners of PV panels and engage in virtual power
The fixed price 𝑝̂ is established in advance and can be considered
trades. In every trading period, the WCEA decomposes the individually
fair because it ensures equal allocation of savings: for every kWh
𝑝 −𝑝 metered load into self-consumption, internal trade, export and import.
exchanged, the buyer saves 1 2 2 compared to the grid price 𝑝1 while
𝑝 −𝑝 Over the billing period, totals satisfy
the seller earns 1 2 2 above the feed-in tariff 𝑝2 . Keniston et al. (2024)
demonstrate that the principle of ‘‘splitting the difference’’ is exten- Load (+) = Self-consumption (+) + Internal Trade (±)
sively observed in practice and holds substantial normative appeal + Export (-) + Import (+)
(cf. Schelling, 1960).
The WCEA, to be shortly presented, prioritizes solving energy imbal- Final (net) amounts due can then be calculated by applying the cor-
ances within the EC before relying on the more costly grid. In scenarios responding prices (0, 𝑝,
̂ 𝑝2 , 𝑝1 ) to each component. This operation can
where either all members are buyers (case #1) or all members are be entrusted to the DSO, as smart meters ensure secure and reliable
sellers (case #2), trading cannot occur and all energy is either bought communication. The resolution of outstanding financial claims or obli-
gations within the accounting system of the DSO also allows to comply
or sold from/to the grid at the prices 𝑝1 and 𝑝2 . If there is a slight
with national privacy regulations.
excess of energy (case #3), it is traded internally to satisfy some of
the unsatisfied demand and finally, if there is a slight energy demand
excess (case #4), it is traded internally to satisfy some of the excess
supply waiting to be sold. The WCEA manages all four cases.9 Clearly, #5 Accounting The retailer/DSO serving EC members incorpo-
as the tariff features 𝑝1 > 𝑝2 , it is rational for EC members to cover rates the PV meter into its list of clients and executes the WCEA
their power load from their share of the PV output. Our fourth module algorithm on its computer system. EC members are charged for
thus directly builds from the residual demand. incoming energy at the positive billing rate 𝑝1 , receive negative
billing for outgoing energy at rate 𝑝2 , and charged for ‘‘community
power’’ at rate 𝑝.
̂

7
If going forward with this option, the EC will have to amend its bylaws
and redistribute PV ownership shares in a negotiation involving the new
members and long-time ones. An alternative settlement process can be implemented locally pro-
8
Notice that net-metering is net-billing with 𝑝1 = 𝑝2 . In practice, it required vided that all EC members cooperate:10 the EC sets up a bank account
the DSO to let the meter run backwards when the PV output exceeded as a single entity and signs a contract with the retailer which covers
consumption.
9
the connection cost for all 𝑛 members meters. The bill of the EC is
Because electricity flows through the distribution network (and within
determined by the combined readings of the 𝑛 smart meters and tariff
buildings) according to Kirchhoff’s laws, it cannot be routed linearly from
a specific source to a specific sink in an AC system. Instead, instantaneous
(𝑝1 , 𝑝2 ). A micro-computer reads data from both the smart meters and
imbalances between generation and demand are automatically balanced across the PV system meter, calculating expenses on a minute-by-minute basis.
the grid. The Distribution System Operator (DSO) plays a crucial role in Every month, a program on the micro-computer computes the total,
ensuring that the infrastructure supports such bidirectional flows safely and adds the connection fee, and generates a bill for each member, which
reliably. Accurate billing is enabled by meters installed at both generation and
consumption points. Mechanisms such as the WCEA and other energy-sharing
10
frameworks assign a real-time price per kilowatt-hour (kWh) to each source Cooperation implies for instance that all EC members are in good standing
and sink. with the community expenses and have a solid payment discipline.

4
N. Boccard and R. Goetz Energy Economics 150 (2025) 108822

∑ ∑
is hand-delivered by a trustee. Members electronically pay their bill demand case (1) amounts to solve 0 = buyers min(𝑞𝑖 , 𝑥) + sellers 𝑞𝑗
to the EC account. By design, the sum of all EC members’ bill precisely which is equivalent to
matches the amount due to the retailer. This ‘‘snail’’ protocol eliminates ∑
0 = min(𝑞𝑘 , 𝑥) 𝑥 ∈ R+ (3)
the requirement for an internet connection, thereby removing the
𝑘≤𝑛
necessity for blockchain encryption and a complex trading platform.
Nevertheless, its success depends on a high level of trust among EC since for seller 𝑗, the function of 𝑥, min(𝑞𝑗 , 𝑥) is systematically equal
members. to 𝑞𝑗 < 0 over R+ . Likewise, the excess supply case (2) solves 0 =
∑
𝑘≤𝑛 min(−𝑞𝑘 , −𝑥) for 𝑥 ∈ R− which is equivalent to
4. Social choice theory ∑
0 = max(𝑞𝑘 , 𝑥) 𝑥 ∈ R− (4)
𝑘≤𝑛
The weighted constrained equal awards (WCEA) rationing rule has Eqs. (1) and (2) define a dichotomous structure where the validity
been discussed and explored through various axiomatizations but, as of each equation is mutually exclusive and depends on the sign of 𝑞̄ =
far as we could ascertain, has never been given a practical application ∑ ̄ ̄
𝑘≤𝑛 𝑞𝑘 = 𝐷 − 𝑆. To summarize,
{
the uniform rule allocates to individual
except by Alyami (2024) for the curtailment of RES output when the min(𝑞𝑘 , 𝑥∗ ) if 𝑞̄ > 0
high-voltage transmission network becomes congested. This section 𝑘, the net exchange 𝑥𝑘 = , where 𝑥∗ is the unique
max(𝑞𝑘 , 𝑥∗ ) if 𝑞̄ ≤ 0
formally presents the WCEA, temporarily setting aside the context of {∑
min(𝑞𝑘 , 𝑥) if 𝑞̄ > 0
electricity, households and PV panels. real root over R of function ℎ(𝑥) ≡ ∑𝑘≤𝑛 .
𝑘≤𝑛 max(𝑞𝑘 , 𝑥) if 𝑞̄ ≤ 0

4.1. Constrained equal awards Summarizing, the uniform rule concatenates the CEA twice, once
for an excess demand and once for an excess supply. As a testimony
The constrained equal awards (CEA) was originally proposed by Mai- to its success, Thomson (1994) declared it ‘‘the most important solu-
monides (1175) to settle a bankruptcy among creditors or divide an tion to the problem of fair division in economies with single-peaked
estate among heirs. In a modern application, Goetz et al. (2008) use preferences’’.
the CEA to allocate a scarce water resource to farmers. In the formal
context, 𝑛 individuals each have a claim that, collectively overcomes 4.2. Priority extension
the endowment available for allocation. The CEA mechanism can be
illustrated as follows: claimants gather at a round table while an The CEA is deemed egalitarian because participants are treated
arbitrator divides the endowment into small units, such as one dollar, equally and have the same level of access to the limited resource until
and sequentially distributes a dollar to each claimant. This sequence their claims are fully satisfied. When claimants possess individualized
is repeated until the endowment is exhausted or a claimant’s demand rights 𝛾𝑘 , so-called entitlements or rationing priorities, proportional
is fully satisfied, at which point they exit the process, and the cycle rationing is commonly used to ensure that the awards distribution
resumes. Because the distribution unit is minimal, the seating position preserves the proportionality of these priorities. However, if, there
at the table is irrelevant. A visual representation of this process is exists a cap on individual rewards, the principle of proportionality must
available in Appendix A. be adjusted accordingly. The weighted constrained equal awards (WCEA)
The CEA solves a rationing problem that is structurally identical to proposed by Moulin (2000) precisely achieves the required adjustment
a market imbalance called excess demand because aggregate demand of absolute proportionality.13 To tweak Eq. (3), it is sufficient to multi-
exceeds aggregate supply. In the finance jargon, the demand side is ply the award 𝑥 by the priority 𝛾𝑘 assigned to agent 𝑘. Eq. (4) is likewise
long while the supply side is short. The market analogy suggests looking modified. An extended version of the uniform rule, denoted by WCEA∗ ,
at the inverse situation given by a short demand and long supply. For accounts for different priorities by allocating14
this dual case, the CEA applies as well, simply by exchanging the roles {
min(𝑞𝑘 , 𝛾𝑘 𝑥∗ ) if 𝑞̄ > 0
of market participants. Sellers are now rationed and the endowment ∀𝑘 ≤ 𝑛, 𝑥𝑘 = (5)
max(𝑞𝑘 , 𝛾𝑘 𝑥∗ ) if 𝑞̄ ≤ 0
represents the entire amount of purchases that needs to be distributed
among them. The CEA is therefore able to solve any market imbalance to individual 𝑘 where 𝑥∗ is the unique zero of function
by adequately identifying the short and long sides. In the context of {∑
min(𝑞𝑘 , 𝛾𝑘 𝑥) if 𝑞̄ > 0
market disequilibrium,11 Bénassy (1982) calls it the Uniform Rule.12 ℎ(𝑥) ≡ ∑𝑘≤𝑛 (6)
𝑘≤𝑛 max(𝑞𝑘 , 𝛾𝑘 𝑥) if 𝑞̄ ≤ 0
Mathematically, we consider a group of sellers each supplying the
individual quantity 𝑠𝑗 and a group of buyers each demanding the indi- defined over R. To verify that the WCEA∗ has a unique solution, we
vidual quantity 𝑑𝑖 . Let 𝑆̄ and 𝐷̄ be the aggregate supply and demand. prove the following lemma.
If 𝑆̄ < 𝐷,
̄ the originally formulated CEA solves
∑ ∑ Lemma 1. Function ℎ(.) has a unique real root over R.
𝑠𝑗 = 𝑆̄ = min(𝑑𝑖 , 𝑥) 𝑥 ∈ R+ (1)
sellers buyers ∑ ∑
Proof. Define 𝑓 (𝑥) ≡ 𝑘≤𝑛 min(𝑞𝑘 , 𝛾𝑘 𝑥) and 𝑔(𝑥) ≡ 𝑘≤𝑛 max(𝑞𝑘 , 𝛾𝑘 𝑥).
with final buyer allocation indicated by 𝑥𝑖 = min(𝑑𝑖 , 𝑥∗ ). If 𝐷̄ < 𝑆,
̄ the Observe that both functions are stepwise linear, thus continuous and
mirror CEA solves weakly increasing, ordered (𝑔 ≥ 𝑓 ) and satisfies 𝑓 (0) = −𝑆̄ < 0, 𝑔(0) =
∑ ∑ 𝐷̄ > 0 and 𝑓 (+∞) = 𝑞̄ = 𝑔(−∞). Applying the intermediate value
𝑑𝑖 = 𝐷̄ = min(𝑠𝑗 , 𝑥) 𝑥 ∈ R+ (2)
theorem to 𝑓 and 𝑔, we note that one and only one of them has a
buyers sellers
To unify the two cases, we denote the number of agents by 𝑛 and the unique zero over R, the sole zero of ℎ(.). More precisely, if 𝑞̄ > 0, the
individual net market demand by 𝑞𝑘 which either 𝑑𝑖 or −𝑠𝑗 ; the excess weakly increasing 𝑓 (.) must cross the horizontal axis between 0 and
+∞ while the weakly increasing 𝑔(.) starts from a positive value at −∞
and therefore remains above the horizontal axis forever. The case 𝑞̄ < 0
11
Historically relevant cases are massive unemployment (workers long and is entirely symmetrical. ■
employers short) and idleness of productive capacities during economic crises
due to inelastic price adjustments (firms long and consumers short).
12 13
He also proposes the proportional rule, though it is susceptible to ma- cf. Flores-Szwagrzak (2015) and Zhao and Ohseto (2022).
14
nipulation since individuals may exaggerate their claims if their veracity is The starred notation is only used in this section to indicate the
unverifiable. mathematical implementation of the WCEA.

5
N. Boccard and R. Goetz Energy Economics 150 (2025) 108822

As we show in Proposition 1 of B&G, the WCEA∗ belongs to the class excess PV output from neighbors at the attractive mid-market rate 𝑝, ̂
of sequential allotment rule defined by Barberà et al. (1997). As such, rather than paying the higher retail rate 𝑝1 . Our numerical simulation
it complies with the properties of efficiency, individual rationality, for the original dataset reveals that in the long run, an average 5% of
non-manipulability and replacement-monotonicity.15 The WCEA∗ also the building load is traded among members of the EC, thus cutting their
𝑝 −𝑝
satisfies the fairness criterion, since agents’ priorities can be defined in combined energy bill by 1𝑝 2 × 5% ≈ 2.5%. Therefore, while members
1
any manner representing their agreed collective view of fairness. generally save 2.5%, the ‘‘wise’’ investor who contributes minimally
to the shared PV system manages to reduce cost by at least 5% with
4.3. Numerical solution negligible investment. This constitutes a typical case of ‘‘freeloading’’.
To eliminate this socially wasteful incentive, we adopt the proportion-
The WCEA∗ numerical implementation resembles with the algo- ality principle which guarantees that future internal trade will reflect
rithm used by Sönmez (1994) to characterize the uniform rationing the initial investment effort to establish the EC. In other words, the
rule: it arranges claims in ascending order, serves the claims that do distribution of PV output should align with each owner’s investment,
not exceed the participant’s entitlement and removes those claimants represented by the share 𝛾𝑖 of the PV system. Therefore, we utilize the
from the pool and starts anew. In applying the WCEA∗ , we shall use the WCEA∗ along with the PV rights vector 𝛾⃗ as the primary mechanism for
fact that both 𝑓 (.) and 𝑔(.) are piecewise linear and thus exhaustively handling the allocation of any surplus power among the EC members.
𝑞
summarized by their kinks. We compute ∀𝑖 ≤ 𝑛, 𝑥𝑖 = 𝛾𝑖 and order As revealed by the previous steps of the analysis, and reinforced
𝑖
(increasingly) this series of numbers, relabeling 𝑥̂ 𝑖 that in position #i. by its underlying mathematical formulation, the WCEA∗ algorithm can
We now evaluate handle high frequency fluctuations in the prices 𝑝1 and 𝑝2 as well as
( ) in the priority vector 𝛾⃗ . It can therefore be considered well-suited for
∑ ∑
∀𝑚 ≤ 𝑛, 𝑓 (𝑥̂ 𝑚 ) = 𝑞̂𝑘 + 𝑥̂ 𝑚 1 − 𝛾̂𝑘 (7) dynamic power trading scenarios.
𝑘≤𝑚 𝑘≤𝑚

and search for 𝑚∗ such that 𝑓 (𝑥̂ 𝑚 ) < 0 < 𝑓 (𝑥̂ 𝑚+1 ). If found, the solution 5. Net benefit of an energy community
is
∑ In this empirical section, we compute the cost savings resulting
− 𝑘≤𝑚∗ 𝑞̂𝑘
𝑥∗ = ∑ >0 (8) from the installation of a rooftop PV system and the potential to trade
1 − 𝑘≤𝑚∗ 𝛾̂𝑘 energy within the EC. We calibrate the PV system as well as the house-
otherwise 𝑓 (𝑥̂ 𝑛 ) < 0 and we go on to evaluate holds consumption across various geographies and load profiles. Our
( ) approach relies on a Monte-Carlo experiment, the component of which
∑ ∑ are illustrated in Fig. 1 and are explained further down. We obtain a
∀𝑚 ≤ 𝑛, 𝑔(𝑥̂ 𝑚 ) = 𝑞̂𝑘 + 𝑥̂ 𝑚 1 − 𝛾̂𝑘 (9)
𝑘≥𝑚 𝑘≥𝑚 range of cost savings across various configurations. Furthermore, we
evaluate the influence of the PV system size upon the stability on the
to identify 𝑚∗ such that 𝑔(𝑥̂ 𝑚 ) < 0 < 𝑔(𝑥̂ 𝑚+1 ) with solution
∑ grid managed by the DSO.
− 𝑘>𝑚∗ 𝑞̂𝑘
𝑥† = ∑ <0 (10)
1 − 𝑘>𝑚∗ 𝛾̂𝑘 5.1. Economic & technical calibration
In practice, our implementation needs about 4𝑛 operations to solve
We consider representative values for solar irradiance, household
system (7)–(10). To obtain the solution for 𝑛 = 10 individuals, the
demand, PV capacity, roof-space and power prices. Observe firstly that
Wolfram kernel requires a tenth of a millisecond.
Hamburg (Northern Germany) receives about 𝑔 = 1000 full hours
of solar irradiance per year compared to 1600 for Girona (Northern
4.4. Energy community
Spain) and 3000 for Faro (Southern Portugal). We shall speak of
the weak, medium and strong irradiance regimes when referring to a
We now return to the energy community setting and demonstrate
building located across these latitudes. A second key consideration is
how the WCEA∗ facilitates internal trading opportunities while block-
the household electricity consumption which averages 𝑑 = 4 MWh in
ing abusive behavior. Assume the EC is composed of 𝑛 members, each Europe.16 This figure is close to the Spanish one, but is twice as much
∑
owning a positive share 𝛾𝑖 of a PV system (i.e., 𝑖 𝛾𝑖 = 1). Within a in Scandinavia and only one half in the least affluent EU countries.17
given time period, the PV asset output is 𝑧0 ≥ 0 while each owner The second dimension arises from the realization that an energy
consumes 𝑐𝑖 > 0; we may thus form the net residual demand vector community is composed of a variety of users whose power consumption
𝑟⃗ = 𝑐⃗ − 𝑧0 𝛾⃗ ∈ R𝑛 where 𝑐⃗ = (𝑐𝑖 )𝑖≤𝑛 is the load vector and 𝛾⃗ = (𝛾𝑖 )𝑖≤𝑛 is differs. A two-adults household might require only half of the national
the PV entitlement vector. If 𝑟⃗ is positive, everybody must buy power average, whereas a larger family could need double that amount. In
from the retailer at price 𝑝1 whereas if 𝑟⃗ is negative, everybody must addition, numerous urban buildings feature stores and businesses at
sell power to the retailer at price 𝑝2 . Otherwise, there is both a positive ground level, resulting in an even higher energy usage. Considering
∑ ∑
supply 𝑆̄ = − sellers 𝑟𝑗 and a positive demand 𝐷̄ = buyers 𝑟𝑖 inside these factors, we allow the EC members consumption to vary between
the building. To fully leverage this opportunity to trade power at the 2 MWh and 16 MWh annually. Table 1 presents the range of variation
attractive mid-market rate, the long side must be rationed. The uniform for our first two parameters.
rule seems to be an excellent option because it effectively prevents To cover the anual energy consumption using a PV system, a typical
manipulation since participants have no incentives to report an energy household in Girona (medium irradiance) needs to install a productive
consumption that does not match their actual power load. capacity 𝑘 = 𝑑𝑔 = 4000 1600
= 2.5 kWp. Once we take into account
Nevertheless, the CEA that forms the basis of the uniform rule mo-
tivates members to invest minimally into the PV system. Indeed, some
may opt for the smallest eligible ownership stake to take advantage of 16
According to eurostat, the population in 2022 was 446 million people.
the favorable mid-market rate for electricity consumption. The number of households (HH) stood at 199 million, and the total household
A person who invests the minimum admissible amount will almost energy consumption amounted to 708 TWh. This results in an average annual
systematically remain on the demand side of the local market, acquiring consumption of 3.6 MWh per household, with 2.24 people per household on
average.
17
For the United States, EIA (2023) reports a higher average household
15
Individual rationality is ensured only when there are personal upper limits consumption of 10 MWh per year, with for instance California at 6.5 MWh
on rewards, but no minimum guaranteed rewards for individuals. and Florida at 14 MWh.

6
N. Boccard and R. Goetz Energy Economics 150 (2025) 108822

Fig. 1. Energy community framework.

Table 1 𝑐0 = 𝑝1 𝑑 = 800 e. To this, one must add fixed charges (aka connection
Two key energy community parameters. cost) commonly ranging between 200 e and 600 e as detailed in
User demand Single Couple Family Business Appendix B. In the polar case where the household installs a fully
Solar irradiance 2 MWh 4 MWh 8 MWh 16 MWh
sized PV system and signs a virtual battery agreement with the retailer
Weak (1000 h) The actual load and PV generation (i.e., 𝑝2 = 𝑝1 ),18 the annual energy cost of power would be zero. In
Medium (1600 h) will be determined by parameters
this study, we select 𝑝2 = 12 𝑝1 = 10 ¢/kWh which reflects the low
Strong (3000 h) during the Monte Carlo experiment
market value of the large surpluses of PV generated electricity across
the country. Recall that when 𝑝1 > 𝑝2 , a household accumulates a
Table 2 financial debt with the retailer as it is often selling daytime excess PV
Third energy community parameter. output at a discount to subsequently ‘‘buy it back’’ in the evening at a
Roof space Narrow Regular Wide premium. Ultimately, the reduction in the energy bill will be influenced
Energy coverage 25% 50% 100% by the solar irradiance pattern over the year and its ongoing interaction
with the household’s annual load profile.
The last component needed to develop a fully fledged numerical
analysis is the data library of high-frequency load and PV output profiles
the geographical and social/family variability, the capacity will range
created by Boccard (2025). The basic component is a vector containing
between 1 kWp and 16 kWp. To monetize this technical specification,
half a million minute-by-minute readings collected over an entire year,
we adopt the installation cost stated by Fraunhofer (2024) p11 as 𝜔 = 2
either at an actual single family home or a flat or a research facility
e/W. The full investment is therefore 𝐾 = 𝜔𝑘 = 5000 e. The cost in
from a university or a DSO. We conduct a Monte Carlo experiment by
Hamburg (low irradiance) would be 60% higher, while it would be 47%
repeatedly selecting, at random, 𝑛 load datasets to represent households
less in Faro (strong irradiance). Each family type would face a cost in
and one PV dataset. Armed with this data bundle, we compute for
line with these variations.
every of the half million minutes of the year, how much power each
The third key dimension relates to the necessary roof area. Although
household needs, how much PV generation is available, how much
a 2 × 1 meters PV panel can generate up to 333 W, this 6 m2 /kWp
power it might be selling or buying, and finally based on the WCEA
guideline should be adjusted to about 10 m2 /kWp to account for the
scheme, how much power is shared among the residents of the building
lost space caused by the existence of chimneys, air conditioning units,
to minimize costly purchases from the retailer. Yearly totals are than
and shading from nearby structures—a minor issue for standalone
calculated.
houses but a very acute one in the high-rise urban environment. Ul-
timately, a family residing in Girona would need to free a significant
5.2. Condominium building & power trading
25 m2 of roof space. Given that urban buildings typically have mul-
tiple floors with numerous owners, it is unlikely that there would be
We consider a hypothetical condominium consisting of six standard
adequate square footage for all of them. We therefore assume that the
and four larger apartments. Total annual energy consumption is set at
available roof space 𝑆 for installing a PV system will cover a range
70 MWh i.e., an average 7 MWh per flat or an equivalent continuous
of 25% to 100% of the annual power consumption 𝐷 of the building
𝑆 /𝐷 load of 8 kW = 70 MWh
. To simplify the interpretations of the results,
i.e., 14 ≤ 10 𝑔
≤ 1 as synthesized in Table 2. Note that the upper 8760 h
we assume owners ask for their guaranteed share so that PV ownership
limit would typically require a connection to an adjacent roof and also shares are 𝛾⃗ = 𝛼⃗ . Key to our construction is the heterogeneity of the
constitutes a virtual maximum since the EU directive mentions in its households’ consumption and ownership profile. Large flats tend to be
article 2.16.c that the primary purpose of an EC is not financial profits. occupied by larger families who tend to consume more. There is thus a
This limitation would be breached if the PV system were so large that
the EC would become a systematic power seller.
Regarding retail power prices, we select the fixed price 𝑝1 = 20 18
The retailer ‘‘stores’’ noon and summer excess PV generation for night and
¢/kWh as an intermediate value within the retail market range. The winter use. This operation is virtual insofar as the metered excess energy is
annual household electricity cost prior to the PV installation is thus instantaneously consumed by someone else over the network.

7
N. Boccard and R. Goetz Energy Economics 150 (2025) 108822

Table 3
Building basic characteristics.
Flat #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 Building
Area (m2 ) 50 60 70 80 90 100 100 100 200 150 1000
PV share (%) 5 6 7 8 9 10 10 10 20 15 100
Load (W) 250 333 500 667 500 667 750 1000 1500 1833 8000
Load share (%) 3.1 4.2 6.3 8.3 6.3 8.3 9.4 12.5 18.8 22.9 100
𝜇 (%) 60 44 12 −4 44 20 7 −20 7 −35 0

positive correlation between the two dimensions but important varia- 5.3. Condominium power trade and savings
tions are commonly observed. We therefore specifically choose 10 flat
areas and 10 annual consumption that are imperfectly correlated. These The Monte Carlo experiment is repeated, this time incorporating
are shown in Table 3 with their respective shares (in the building); the prices, to compute the savings achieved by EC members over the year
share of PV
bottom row displays an indicator 𝜇 = share − 1 of the relative while allowing the PV system size to vary from 25% to 100% coverage.
of load
advantage each owner holds with respect to the building average.19 We also provide a focus on the amount of internal trade and exchanges
We now use the data library to compute the likeliness of trade with the network.
among EC members. To do so, we assign every minute of the year into For each member of the EC, the WCEA decomposes for each minute
a sunshine class according to the likelihood of trade. This partitioning the load 𝑐 into self consumption, internal trade, export and import,
will reflect the complex relationship between solar irradiance and the a vector (𝑠, 𝑡, 𝑒, 𝑖). Over a year, these sums to 𝐶, 𝑆, 𝑇 , 𝐸, 𝐼 and still
load profiles of all EC members. Our 5 sunshine classes are:20 satisfy 𝐶 = 𝑆 + 𝑇 + 𝐸 + 𝐼. Applying the price vector (0, 𝑝, ̂ 𝑝2 , 𝑝1 ) to
the RHS yields the energy cost of the consumer bill. In the absence
of PV, 𝐶 = 𝐼 since 𝑆 and 𝐸 are uniformly nil. In the presence of
nil no sunshine (night)
PV but without internal trade, we have 𝑇 = 0 so that the export
component (𝐸) increases in absolute value and simultaneously the
low all members of the EC have a positive residual load (no trade)
import component (𝐼) decreases. Comparing the energy costs of these
two scenarios allows us to quantify the cost savings resulting from the
weak the negative residual load helps to cover some of the positive
PV installation. Similarly, enabling internal trade in the annual energy
residual load (trade)
balance makes it possible to assess the additional savings attributable
to the combined effect of PV generation and internal exchange.
strong the positive residual load helps to absorb some of the negative
residual load (trade)
5.3.1. Energy bill savings
Table 4 presents our first findings using a medium solar irradiance
massive all members have a negative residual load (no trade)
and a PV system designed to meet the entire energy demand of the
building (i.e., some PV panels are likely installed on the rooftop of an
In the two trading scenarios (weak & strong ), a portion of the
adjacent building). The first row shows the savings in the absence of
building’s residual power load is traded internally, while the remainder
internal trade, treating each of the 10 flats as if it were a detached
is either imported from or exported to the DSO, depending on whether
house. On average, the savings for the whole building from this PV
the case is classified as weak or strong. In the low case, all residual power installation are 65%. This is a substantial amount, even though it does
is imported and in the massive case, all residual power is exported. The not reach the theoretical 100% maximum. The second row illustrates
temporal distribution of the sunshine classes (except the night one) the cost savings achieved by utilizing internal power trading among
is illustrated graphically. We use a medium solar irradiance profile owners. Total savings increase for everyone, with an additional average
(e.g., Girona) and size the PV system to cover 50% of the annual of 2.5 percentage points saved as a result of trade (68.0% vs. 65.5%),
energy consumption of the building. Fig. 2 shows the probabilities aligning with prior research findings (cf. Long et al., 2018; Fina et al.,
of the presence of the classes low, weak, strong and massive during 2018 or Zhou et al., 2020). A possible explanation for the limited
daytime for every day of the year. For example, conditions that allow additional savings is discussed in Section 5.3.3.
trade are present about 30% of the daytime in July and only about 15% One can appreciate how the diverse households benefit from the
in January (the sum of the ocher and green line). EC’s PV system. For instance, HH#1 consumes less energy than its
Fig. 3 provides a transversal view by displaying the probabilities corresponding share of the PV output, enabling it to sell enough surplus
of the classes low, weak, strong and massive during the day. At noon, to cover its entire energy cost. On the other hand, HH#10’s high energy
favorable conditions sum up to approximately 60% while in the early consumption only allows it to reduces its energy bill by approximately
morning or late afternoon, they are approximately 40%. 50%. Given the previously computed default energy cost of 800 e, an
Figs. 2 and 3 illustrate the probabilities that EC members trade average family will be able to cut its bill by at least 500 e. In addition,
energy but they do not show the amount of power traded. For this the exchange of energy with neighbors provides only additional savings
purpose, we determine the lesser amount between internal supply and of 20 e.
internal demand. Fig. 4 illustrates the amount of power traded during Next, Table 5 examines two significant variations in solar irradiance
each day of the year while Fig. 5 shows the amount of power traded at and the extent to which the PV system meets the building’s energy con-
each hour during a day. The shape of these curves are well known and sumption. The weak northern regime derives from existing British and
recall us that no matter how much we aim to trade power within our German PV installations with a capacity factor (CF) lower than 12%,
EC, climate remains in the driver seat insofar as there is more trade in while the strong southern regime comes from American, Brazilian, and
the summertime and around noon. Australian systems with a CF over 17%. The medium sunshine regime
is drawn from installations located between these latitudes.
The extent of internal trade increases with the coverage of the PV
19
By definition this statistic is nil for the building. system, except under conditions of intense solar irradiance, where a
20
The irradiance regime of the actual geographical location of the building 50% coverage may be considered ideal. Increasing the coverage of the
will impact the size of these 5 classes. Notice that the sunshine classes mirror PV system beyond this threshold creates systematic excess production.
cases 1 to 4 of the fourth component of the blueprint. This surplus cannot be used internally and has to be sold to the DSO

8
N. Boccard and R. Goetz Energy Economics 150 (2025) 108822

Fig. 2. Probabilities of the solar irradiance classes during the year.

PV system 50% coverage & medium solar irradiance.
Probabilities do not add to 100% due to absence of the night class.

Fig. 3. Probabilities of the solar irradiance classes during the day.

PV system 50% coverage & medium solar irradiance.
Probabilities do not add to 100% due to absence of the night class.

Table 4
Energy cost savings from a maximally sized PV system.
Household #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 Mean
PV-no trade (%) 96.8 88.7 71.9 63.3 88.4 76.2 69.8 54.8 69.8 46.4 65.5
PV+trade (%) 100.5 91.6 74.2 65.3 91.7 78.6 72.0 57.2 72.0 48.7 68.0

at lower rates. As noted in the last row of Table 5), a small sized PV Regarding internal trade, we first observe that a limited coverage
system is most efficient, registering savings of 23%, which are close factor of 25% leads to trade among members to represent a share
to the theoretical maximum of 25% achievable with a virtual battery. 2.7
59.4
≈ 4% of all flows. When the PV system is augmented and covers
At 50% coverage, efficiency declines slightly with savings at 39% -
well below the theoretical maximum of 50%. The efficiency is lowest one half of the building load, then the energy bill can be cut by up
3.6
at 100% coverage, since about a third of the solar energy is not used to 36.8% and internal trade rises to 65.5 ≈ 5% of all flows; power
within EC but injected into the grid with limited economic return. exchanges with the grid are also reduced by 1 − 65.5
70.1
≈ 7%.

9
N. Boccard and R. Goetz Energy Economics 150 (2025) 108822

Fig. 4. Building internal trade distribution across the year. (PV system 50% coverage & medium solar irradiance).

Fig. 5. Building internal trade distribution across the day. (PV system 50% coverage & medium solar irradiance).

Table 5
Varying solar irradiance regime and PV coverage.
Solar irradiance regime Weak (north) Medium (central) Strong (south)
PV coverage 25% 50% 100% 25% 50% 100% 25% 50% 100%
Building load (MWh) 70.1 70.1 70.1 70.1 70.1 70.1 70.1 70.1 70.1
Flows (MWh) 60.1 67.1 92.5 59.4 65.5 91.0 55.4 58.3 83.6
Trade (MWh) 2.6 3.5 3.5 2.7 3.6 3.7 3.2 4.8 4.3
Savings PV no-trade (%) 20.5 36.5 65.1 20.8 36.8 65.5 21.7 38.6 67.7
Savings PV+trade (%) 22.4 38.9 67.4 22.7 39.3 67.9 24.0 42.1 70.7

To test the reliability of our estimates, we adjust our model and major divergence that should be prevented, perhaps by implementing a
let the power load for each condominium to be directly proportional regulatory cap.21 Even under a medium sunshine regime, raising the PV
to square footage (without any incentives or surcharges). We find our coverage from 25% to 50% and subsequently to 100% yields a similar
numerical results to be altered by merely a decimal place. phenomenon: flows within the distribution network will increase on
average over the year. Contrariwise, a limited PV system meeting 25%
5.3.2. Impact upon the DSO of the building’s demand reduces energy flows from 70 MWh (no PV)
The energy transition presents substantial challenges for DSOs to 59 MWh which is a 1 − 59.4 = 14% decrease. This is a significant
70.1
whose traditional regulatory framework did not account for urban
local generation. The rise of ECs, with rooftop PV installations, raises achievement that enhances the DSO’s ability to manage the expansion
concerns among DSOs about the additional costs required to ensure of Distributed Energy Resources (DER) without having to raise tariffs.
network reliability. Beyond this, the DSO’s most critical responsibility is to ensure grid
We observe that under a low solar irradiance, an EC is likely to ne- stability and prevent blackouts. It is therefore warranted to investigate
cessitate the installation of additional panels which shall then increase
the flows in and out of the building. Should the annual energy load of
the building exceed 70 MWh, the DSO could potentially claim a breach 21
Unless other segments of the distribution network can absorb these
of its contractual obligation to serve the building. The 92.5 MWh persistent generation surpluses, the situation leads to an inefficient allocation
volume (total coverage in the northern region, see Table 5) indicates a of resources from a macroeconomic perspective.

10
N. Boccard and R. Goetz Energy Economics 150 (2025) 108822

Fig. 6. Top 1% residual demand (medium location, 50% coverage).

Fig. 7. Bottom 1% residual demand (medium location, 50% coverage).

how ECs might impact the DSO’s ability to fulfill this role. To that and therefore calls for careful prospective studies before implementing
end, we analyze the building’s maximum bidirectional power flows in widespread rooftop PV deployment across urban areas.
order to assess the necessary network dimensioning required to serve Note finally that the DSO/retailer is concerned with the granularity
it reliably. For this analysis, we simulate a medium solar irradiance of meter readings. When data is aggregated to an hourly level rather
and a PV system sized to cover 50% of the building load. We display than recorded per minute, numerous physical flows between readings
in Fig. 6 the top 1% demand at each minute of the day. The peaks cancel each other out, leading to a financial loss (given that 𝑝2 < 𝑝1 ).
observed on Fig. 6 typically occur during times of minimal sunlight Consequently, the DSO is likely to advocate for the use of the most
– early morning and late afternoon – with the highest peak reaching detailed granularity available.24
21 kW.22 The demand pattern resembles the historical one - prior to the
installation of the PV system. Consequently, the DSO can reliably serve 5.3.3. Sensitivity analysis
the building without requiring additional infrastructure reinforcement. We design a series of additional Monte Carlo experiments to explore
In Fig. 7, we display the bottom 1% of net demand which is closely the effects of greater heterogeneity among participants. We redefine
related to the maximum outflow from the building. Remarkably, this the initial set of 10 flats to include 3 small, 3 medium, and 4 large
value is lower but nearly equal to the maximum inflow.23 It is consumers. Although their annual energy consumption levels remain
important to highlight that expanding the PV system to cover the entire unchanged, the consumption profiles are now drawn from more diverse
energy needs of the building (100% coverage) causes the maximum set of household data, capturing distinct usage patterns across different
outflow to increase from 18 kW to 29 kW, significantly exceeding the household types. The second modification increases the size of the
maximum inflow of 21 kW observed in Fig. 6. Consequently, oversizing community, first to 20 and then to 30 flats while maintaining the
PV systems could worsen the problems faced by DSOs as more DERs same proportions of small, medium, and large consumers. In each case,
are integrated into their grid. The scale of this impact is significant we randomly draw 20 or 30 complete yearly load profiles, thereby
introducing greater variability and more realism. In a fourth variant,
we replace the 4 large flats with 4 non-residential profiles – such as
22
For reference, note that the continual equivalent load is 8 kW.
23
The analysis assumes that reverse power flows into the distribution
24
network, as long as they remain below the maximum allowable inflow According to our Monte Carlo experiment, the monetary loss remains
(corresponding to the building’s maximum PV generation capacity), do not relatively small, approximately 1% or 140e for the entire building over the
compromise grid stability. course of a full year.

11
N. Boccard and R. Goetz Energy Economics 150 (2025) 108822

Table 6 At moderate scale, small PV systems can be successfully integrated

Varying the energy community heterogeneity. into the existing grid as their requirement for rooftop space is limited
PV coverage Quarter Half Full and they offer two advantages over to larger systems. Firstly, they
Flows 79.0% 68.1% 69.4% deliver a greater return on investment (ROI) since savings present
10 flats
Trade 1.5% 3.7% 9.7%
decreasing returns to scale. Second, widespread deployment of small
Flows 77.9% 65.8% 64.4% scale PV systems will reduce the operating costs of the grid, whereas
20 flats
Trade 1.7% 4.3% 11.2%
the integration of large scale PV forces costly upgrades within buildings
Flows 78.1% 67.0% 64.4% and across the entire distribution network. Based on these findings, we
30 flats
Trade 1.8% 4.4% 11.3%
recommend that energy policies adopt a highly selective approach and
Flows 77.3% 64.6% 60.9%
Non-residential prioritizing the installation of new small PV systems or the progressive
Trade 1.2% 3.2% 8.0%
extension of existing small systems. Furthermore, the presence of large
dimension PV within the grid should be avoided or at least, not actively
incentivized.
schools or hospitals – representing offices or service-sector buildings
with different load profiles. CRediT authorship contribution statement
The results of these experiments are summarized in Table 6, where
each entry represents a share of the building’s total annual electricity
Nicolas Boccard: Writing – review & editing, Writing – original
consumption. We find that a large PV system consistently reduces
draft, Software, Resources, Investigation, Formal analysis, Data cura-
exchanges with the external grid, especially in scenarios where non-
tion, Conceptualization. Renan Goetz: Writing – review & editing,
residential locations are included. Moreover, increasing the community
Supervision, Project administration, Methodology, Conceptualization.
size enhances internal energy sharing: up to 11% of the building’s
total consumption can be met through peer-to-peer exchanges. In other
words, roughly one-out-of-nine households is effectively ‘‘living off- Declaration of competing interest
grid’’, with its energy needs fully covered by its own PV generation
and surplus energy from neighbors. The authors declare that they have no known competing finan-
cial interests or personal relationships that could have appeared to
6. Conclusion influence the work reported in this paper.

In this article, we propose a blueprint to address the key challenges Acknowledgments

that must be overcome to establish and manage an energy commu-
nity. We code an algorithm implementing the Weighted Constrained Nicolas Boccard reports financial support from Ministerio de Cien-
Equal Awards (WCEA) mechanism, enabling virtual peer-to-peer energy cia, Innovación y Universidades, Spain (PID2022-138003NB-I00), Gen-
trading among community members. Unlike sharing rules commonly eralitat de Catalunya, Department of Research and Universities, Grant
analyzed in the literature, the proposed mechanism satisfies five norma- 2023, CLIMA 00090 and SGR 1360. Renan Goetz reports financial
tive criteria drawn from social choice and game theory. In particular, support from Generalitat de Catalunya, Department of Research and
it upholds fairness and replacement monotonicity, ensuring that mem- Universities, Grant 2023, CLIMA 00090 and SGR 1360.
bers are treated in a way that reflects their specific contributions
and circumstances, while maintaining stability as participants join or
Appendix A. CEA and WCEA visualization
leave the community. These properties not only support the practical
implementation of energy communities but also help mitigate potential
internal conflicts. The CEA is given an hydraulic visualization by Kaminski (2000).
Based on a comprehensive and high-frequency real-world data set Imagine 3 farmers who share a reservoir in the middle of a drought
of household loads and solar irradiance, our numerical simulations and must find a way to allocate the little water it contains. On the left
indicate that internal trade among residents offers minimal savings panel of Fig. A.8, Farmer 𝑖 owns an empty vessel closed by a lid at
for the EC in contrast to the substantial savings achieved through height 𝑑𝑖 , representing his claim to water usage. Vessels are joined at
the installation of PV panels. As an example, a photovoltaic system the bottom by a communicating tube linking to the reservoir; as water
covering a quarter of the building load (under medium solar irradiance) flows down the tube, it goes into all 3 vessels and rise. According to
cuts the energy bill by at least 20.8% while the internal trade of power Pascal’s law, the water level ascends uniformly throughout. Once the
allows cutting the energy bill by an additional 2.3%. In scenarios where first vessel is filled up, water keeps rising in the remaining ones until
the roof area is sufficiently large and the PV system is configured the second vessel is filled up and so on.
to meet the entire energy demand of the building, the reduction of The CEA mirror situation of excess supply is illustrated on the right
the electricity bill can reach 68%. However, this would result in a panel of Fig. A.8 and goes as follows: heavy rain is expected and
minor contribution to internal trade, covering just 2.5% of total energy will surely overflow the reservoir, meaning that each farmer will have
flows. Importantly, such a large PV system is likely to increase power to accept some flooding of his field. The hydraulic solution requires
exchanges with the network due to the mismatch between load and aligning the vessels at the top rather than the bottom. The task is to
consumption profiles across the day. This may entail an unexpected empty them until the bottom pool (safety reservoir) is filled. If of small
additional cost for members of the EC because they have to contract size, the water level is the same everywhere i.e., everyone saves the
a greater connection power (kW). If widespread over the city, such a same amount of flooding (darkest blue). If the reservoir is bigger, the
development will also increase the cost for the distribution system op- first field dries (dark blue); the process goes on as the reservoir size
erator because it requires a grid reinforcement. Moreover, our findings increases.
suggest that savings from internal trade can only be increased if the The hydraulic visualization is easily extended to the WCEA by
member’s daily consumption profiles are sufficiently heterogeneous. making the vessel width of farmer 𝑘 equal to priority 𝛾𝑘 and its height
𝑞
With families, businesses and public entities operating appliances and equal to 𝛾𝑘 , this to keep volume at 𝑞𝑘 . As water flows into the vessels,
𝑘
machines at different moments of the day and night, gains from trade more is awarded to someone with a large entitlement, compared to
(energy bill savings) rise from 2.5% to 11%, a notable increase. someone else with a small entitlement.

12
N. Boccard and R. Goetz Energy Economics 150 (2025) 108822

Fig. A.8. Hydraulic visualizations.

Appendix B. Connection cost calibration CPUC, 2024. Demand Flexibility Rulemaking. Technical Report R.22-07-005,
California Public Utilities Commission, URL: [Link]/-/media/cpuc-
website/divisions/energy-division/documents/demand-response/demand-flexibility-
The non-energy cost of power paid to the retailer (or DSO) include oir/ab205_factsheet_050824.pdf.
grid connection cost as well as taxes, duties and levies; the sum is EIA, 2023. Residential Energy Consumption Survey (RECS) Dashboard. Technical
highly variable across jurisdictions and to complicate matters even Report, Energy Information Administration, URL: [Link]/3UlvTUP.
more, duties are sometimes proportional to load. Since there is no EU, 2018. Directive on the promotion of the use of energy from renewable sources.
Off. J. Eur. Union URL: [Link]/eli/dir/2018/2001/2023-11-20.
unified nor standardized reporting on this subject, we shall estimate
Eurostat, 2024. Electricity Prices for Household Consumers. Technical Report, Eurostat,
a range by analyzing original documents and infer the breakdown of [Link]
total cost into energy and other components, hereafter the connection Fina, B., Fleischhacker, A., Auer, H., Lettner, G., 2018. Economic assessment and
fee. business models of rooftop photovoltaic systems in multiapartment buildings: Case
According to federal (BDEW, 2024), the 2023 german residential en- studies for austria and germany. J. Renew. Energy 2018, 9759680. [Link]
org/10.1155/2018/9759680.
ergy price was 24 ¢/kWh, complemented by 11 ¢/kWh for connection
Fioriti, D., Frangioni, A., Poli, D., 2021. Optimal sizing of energy communities with fair
and another 11 ¢/kWh for duties and RES subsidies. Hence, the share 𝜆 revenue sharing and exit clauses: Value, role and business model of aggregators
11+11
of non-energy cost is 24+11+11 ≈ 48%. For France, the (Eurostat, 2024) and users. Appl. Energy 299, 117328. [Link]
full cost figure of 28 ¢/kWh for 2023 decomposes into respectively 18, 117328.
5 and 5 ¢/kWh (cf. MNE, 2024), leading to 𝜆 = 5+5 28
≈ 36%. In the US, Flores-Szwagrzak, K., 2015. Priority classes and weighted constrained equal awards
rules for the claims problem. J. Econom. Theory 160, 36–55. [Link]
the California and Texas residential full rates are 30 and 14 ¢/kWh, 1016/[Link].2015.08.008.
including 𝜆 = 17% and 24% for connection fees (cf. CPUC, 2024). The Fraunhofer, 2024. Levelized Cost of Electricity - Renewable Energy Technologies.
𝜆
connection fee is then approximated by 1−𝜆 800 e which leads to the Technical Report, Fraunhofer Institute for Solar Energy Systems ISE, Christoph Kost,
range [200, 600] e. URL: [Link]/content/dam/ise/en/documents/publications/studies/
EN2024_ISE_Study_Levelized_Cost_of_Electricity_Renewable_Energy_Technologies.
pdf.
References Gjorgievski, V.Z., Velkovski, B., Minuto, F.D., Cundeva, S., Markovska, N., 2023. Energy
sharing in european renewable energy communities: Impact of regulated charges.
Alyami, S., 2024. Fairness and usability analysis in renewable power curtailment: A Energy 281, 128333. [Link]
microgrid network study using bankruptcy rules. Renew. Energy 227, 120424. Goetz, R.U., Martinez, Y., Rodrigo, J., 2008. Water allocation by social choice rules: The
[Link] case of sequential rules. Ecol. Econom. 65, 304–314. [Link]
Barberà, S., Jackson, M.O., Neme, A., 1997. Strategy-proof allotment rules. Games ecolecon.2007.06.018.
Econom. Behav. 18, 1–21. [Link] Kaminski, M., 2000. Hydraulic rationing. Math. Social Sci. 40, 131–155. [Link]
BDEW, 2024. Strompreisanalyse. Technical Report, Bundesverband der Energie und org/10.1016/S0165-4896(99)00045-1.
Wasserwirtschaft, URL: [Link] Keniston, D., Larsen, B.J., Li, S., Prescott, J., Silveira, B.S., Yu, C., 2024. Splitting the
Bénassy, J.-P., 1982. The Economics of Market Disequilibrium. Academic Press, difference in incomplete-information bargaining: theory and widespread evidence
URL: [Link] from the field. Internat. Econom. Rev. [Link]
disequilibrium. Kocakusak, D., Greenberg, M.R., Andrews, C.J., 2024. Key influences on residential
Berg, K., Foslie, S.S., Farahmand, H., 2024. Industrial energy communities: Energy photovoltaic solar panel adoption in the united states. Energy Sources B: Econ.
storage investment, grid impact and cost distribution. Appl. Energy 373, 123908. Plan. Policy 19, 2418126. [Link]
[Link] Long, C., Wu, J., Zhou, Y., Jenkins, N., 2018. Peer-to-peer energy sharing through a
Boccard, N., 2025. The Power Chunk. Technical Report, SSRN, URL: [Link]/ two-stage aggregated battery control in a community microgrid. Appl. Energy 226,
abstract=5174031. 261–276. [Link]
Boccard, N., Gautier, A., 2021. Solar rebound: The unintended consequences of Maimonides, M., Book of Judgements, Vol. 4, Moznaim Publications, Chabad-
subsidies. Energy Econ. 100, [Link] Lubavitch. URL [Link]/library/article_cdo/aid/1161176/jewish/Malveh-
Boccard, N., Goetz, R., 2025. The Sequential Sharing Rule As an Alternative to Estab- [Link].
lished Sharing Rules for Energy Communities: A Comparative Analysis. Technical MNE, 2024. Evolution du Prix de l’Électricité. Technical Report, M’ediateur national
Report 5176737, SSRN, URL: [Link]/abstract=5176737. de l’énergie, URL: [Link]/calculette/process/electricity-
Bossu, A., Durillon, B., Davigny, A., Barry, H., Robyns, B., Belaïd, F., Saudemont, C., price.
2024. Coalitional game-based gain generation and distribution for collective self- Moulin, H., 2000. Priority rules and other asymmetric rationing methods. Econometrica
consumption in an energy community. Math. Comput. Simulation 225, 129–147. 68, 643–684. [Link]
[Link] Rosen, G., Walks, A., 2015. Castles in toronto’s sky: condo-ism as urban transformation.
Çağdaş, V., Paasch, J.M., Paulsson, J., Ploeger, H., Kara, A., 2020. Co-ownership shares J. Urban Aff. 37, 289–310. [Link]
in condominium – a comparative analysis for selected civil law jurisdictions. Land Schelling, T., 1960. Strategy of Conflict. Harvard University Press, URL: [Link].
Use Policy 95, 104668. [Link] com/books?id=7RkL4Z8Yg5AC.

13
N. Boccard and R. Goetz Energy Economics 150 (2025) 108822

Schwidtal, J., Piccini, P., Troncia, M., Chitchyan, R., Montakhabi, M., Francis, C., Thomson, W., 1994. Consistent solutions to the problem of fair division when
Gorbatcheva, A., Capper, T., Mustafa, M., Andoni, M., Robu, V., Bahloul, M., preferences are single-peaked. J. Econom. Theory 63, 219–245. [Link]
Scott, I., Mbavarira, T., Espa–a, J., Kiesling, L., 2023. Emerging business models 10.1006/jeth.1994.1041.
in local energy markets: A systematic review of peer-to-peer, community self- Tushar, W., Saha, T.K., Yuen, C., Liddell, P., Bean, R., Poor, H.V., 2018. Peer-to-peer
consumption, and transactive energy models. Renew. Sustain. Energy Rev. 179, energy trading with sustainable user participation: A game theoretic approach. IEEE
113273. [Link] Access 6, 62932–62943. [Link]
Sönmez, T., 1994. Consistency, monotonicity, and the uniform rule. Econom. Lett. 46, Zhao, Z., Ohseto, S., 2022. Strategy-proof and fair reallocation with single-peaked
229–235. [Link] preferences. Soc. Choice Welf. 58, 791–800. [Link]
Soto, E.A., Bosman, L.B., Wollega, E., Leon-Salas, W.D., 2021. Peer-to-peer energy 021-01374-3.
trading: A review of the literature. Appl. Energy 283, 116268. [Link] Zhou, S., Zou, F., Wu, Z., Gu, W., Hong, Q., Booth, C., 2020. A smart community energy
10.1016/[Link].2020.116268. management scheme considering user dominated demand side response and p2p
trading. Int. J. Electr. Power Energy Syst. 114, 105378. [Link]
[Link].2019.105378.

14