Conference PaperPDF Available

Efficient Power Sharing at the Edge by Building a Tangible Micro-Grid -the Texas Case

Conference Paper

Efficient Power Sharing at the Edge by Building a Tangible Micro-Grid -the Texas Case

Abstract and Figures

Information and Communication Technology (ICT) is now touching various aspects of our lives. The electricity grid with the help of ICT is transformed into Smart Grid (SG) which is highly efficient and responsive. It promotes two-way energy and information flow between energy distributors and consumers. Many consumers are becoming prosumers by also producing energy. The trend is to form small communities of consumers and prosumers leading to Micro-grids (MG) to manage energy locally. MGs are parts of SG that decentralize the energy flow by allocating the produced energy within the community. Energy allocation amongst them needs to solve issues viz., (i) how to balance supply/demand within micro-grids; (ii) how allocating energy to a user affects his/her community. To address these issues we propose six Energy Allocation Strategies (EASs) for MGs-ranging from simple to optimal. We maximize the usage of the energy generated by prosumers within MG. We form household-groups sharing similar characteristics to apply EASs by analyzing thoroughly energy and socioeconomic data of households. We propose four metrics to evaluate EASs. We test our EASs on the data from 443 households over a year. By prioritizing specific households, we increase the number of fully served households up to 81% compared to random sharing.
Content may be subject to copyright.
Efficient Power Sharing at the Edge by Building a
Tangible Micro-Grid – the Texas Case
Nikos Kouvelas1, R. Venkatesha Prasad1, Akshay U Nambi2
1Embedded and Networked Systems, TU Delft, 2Microsoft Research India
Abstract—Information and Communication Technology (ICT)
is now touching various aspects of our lives. The electricity
grid with the help of ICT is transformed into Smart Grid
(SG) which is highly efficient and responsive. It promotes two-
way energy and information flow between energy distributors
and consumers. Many consumers are becoming prosumers by
also producing energy. The trend is to form small communities
of consumers and prosumers leading to Micro-grids (MG) to
manage energy locally. MGs are parts of SG that decentralize
the energy flow by allocating the produced energy within the
community. Energy allocation amongst them needs to solve
issues viz., (i) how to balance supply/demand within micro-grids;
(ii) how allocating energy to a user affects his/her community. To
address these issues we propose six Energy Allocation Strategies
(EASs) for MGs – ranging from simple to optimal. We maximize
the usage of the energy generated by prosumers within MG. We
form household-groups sharing similar characteristics to apply
EASs by analyzing thoroughly energy and socioeconomic data
of households. We propose four metrics to evaluate EASs. We
test our EASs on the data from 443 households over a year. By
prioritizing specific households, we increase the number of fully
served households up to 81% compared to random sharing.
I. INTRODUCTION
Traditionally, the energy distribution network (grid) is
centralized. Substations are primarily used to interface cen-
tralized generators to a large number of end-users. Further,
the electricity grid, utilizing ICT, is now transformed into a
highly efficient and responsive grid, also known as Smart Grid
(SG). Apart from drawing energy from the power line some
consumers generate energy using renewable sources and are
called prosumers. To manage the requirements of prosumers
and consumers efficiently, SG employs intelligent monitoring,
control, and bidirectional communication. This enhanced the
efficiency, reliability and sustainability of the electricity grid.
SGs deploy large numbers of smart meters. These Internet-
enabled devices collect fine-grained data regarding energy
usage and offer real-time information to enhance efficiency
in energy generation and distribution and bring consumption-
awareness. Prosumers generate power using solar (mostly),
wind, hydro, etc., which can be allocated to other customers
in the vicinity. This makes SGs dynamic and less dependent on
the substation. However, renewable sources of energy are in-
termittent and require forecasting. Thus, the presence of power
distribution lines of substations as stable electricity suppliers is
imperative. Micro Grids (MGs) are small communities of con-
sumers and prosumers that have evolved to support distributed
control from SGs. MGs allocate energy between consumers
and prosumers while complying with policies prioritizing
Fig. 1: Models of MG with Central Controller (CC); (left) CC used
only for communication; (right) CC also has storage.
certain users. The energy redistribution at a local level is also
economically beneficial (see Fig. 1). Buying energy from the
substation is more expensive compared to getting it from the
neighbourhood while selling back to the substation is less
lucrative compared to selling directly to neighbours [1]. To
share energy at a neighborhood level, storage point coalitions
of utility companies and municipalities are used. They keep
the generated excess energy and supply it according to the ser-
vice priorities and policies of their respective MGs. However,
allocating energy among prosumers and consumers is non-
trivial because of several constraints: (i) individual consumers
present varying energy requirements over time, and hence
allocation mechanisms need to be adaptive; (ii) prioritizing
certain households causes bias in the community, therefore it
is essential to develop rigorous Energy Allocation Strategies
(EASs); (iii) the predictability of the generated energy is
limited; and (iv) socioeconomic characteristics (often private)
affect consumption and generation of energy (e.g., size of
households, income, and age of residents) [2], [3]. We propose
EASs aiming to achieve fairness, defined for particular groups
of consumers or over entire MGs. Specifically, encompassing
the above issues we answer the general question: How to
optimize the allocation of produced energy excess between
the members of a community under various constraints?
To this end, (a) we propose three optimal EASs to maximize
energy sharing and minimize the energy borrowed from the
substation based on game theoretic and information theoretic
formulations [4]; (b) we propose three simple EASs for
MGs without centralized energy storage; (c) we demonstrate
the efficacy of our proposed algorithms on a real-world
dataset collected over a year from 443 households located in
Texas [5]. Though we use some commonly used methodology
978-1-7281-5089-5/20/$31.00 ©2020 IEEE
well-known in communications, the metrics and treatment
are different and novel. This work targets the problem of
sharing the energy locally and the intricacies involved
therein rather than the grid related issues.
II. RE LATE D WOR KS
Morstyn et al. propose a virtual power plant created through
P2P transactions among prosumers in order to incentivize
them to coordinate and trade their excess energy [6]. Sim-
ilarly, in [7] individual households control the energy they
generate through renewables by an energy sharing coordinator.
Prosumers of a micro-grid store their excess energy in a
common storage unit for later usage in [8], and a function that
accounts for the historical consumption data of the households
is designed to re-allocate the stored energy in households and
to schedule their consumption. The problem of online energy
management in networked MGs is considered in [9]–[11].
Shi et al. propose a stochastic model of the power flow in MGs
for real-time energy management based on Lyapunov opti-
mization [9]. Online energy management of MGs by applying
the Alternating Direction Method of Multipliers (ADMM) on
the historical data of the generated energy is proposed in [10],
[11]. Liu et al. consider a centralized operator per MG that
constructs and controls an energy exchange network between
prosumers and the power grid, while Ma et al. consider
privately owned MGs exchanging energy with adjacent MGs
based on power flow constraints using the power line. Game-
theoretic approaches are considered in [12], [13]. Motivated
by the cooperative game theory, Du et al. form coalitions
of MGs, which coordinate sharing of surplus in electrical
and thermal energy in order to minimize their operational
costs [12]. The economic benefits for households applying
a game-theoretic peer-to-peer energy trading scheme are an-
alyzed in [13], where the aforementioned coalitions among
different prosumers are proven to be stable. The majority of
works above use community-simulators and numerical case
studies to apply energy allocation strategies [8]–[10], [12],
[13], however, we incorporate many methods well-known in
ICT domain with new metrics on a real-case data set to
propose new energy sharing strategies.
III. SYS TE M MOD EL
Fig. 1 depicts an abstract model of an MG neighborhood-
community. From an energy perspective, MGs are sets of
households with different energy needs, equipped with a
number of electrical appliances. In addition, among the
households, some are prosumers generating energy through
renewable sources. Note that if the households cannot cover
their own needs by generating energy, the deficit is drawn
from the power distribution line of the substation. In an MG
community of cconsumers and pprosumers, let the group
of consumers be C={C1, C2, ..., Cc}and, similarly, P=
{P1, P2, ..., Pp}representing prosumers. Both Cand Pare
connected to the power line of the substation, which is also
mandatory for energy transactions between them, as Cand
Pdo not possess the infrastructure required to share energy
directly. To this end, the role of applying EASs between
households is the responsibility of a central controller (CC),
owned by the MG-operator (utility companies). In Fig. 1, the
CC is connected to all the households, to route information
about the energy needs of consumers and the amounts of
energy generated by the prosumers. The decisions of CC
about any energy transition are forwarded to the involved
prosumers and consumers. However, apart from the MG
models in which the CC is solely a communication point,
there are also models in which it connects to the power line
of the substation, to store and forward the excess energy from
prosumers to (members of) Cusing the EAS-algorithms (cf.,
right part of Fig. 1) [4]. Since prosumers have their own
energy needs, they cannot allocate all their generated energy to
consumers. Once the total produced excess energy is stored
in CCs, the CCs are informed by the consumers regarding
their energy requirements, Ea={Ea,1, Ea,2, ..., Ea,c}, and
then, the dictated allocation strategy (EAS) is applied. As
a result, every consumer i[1, c]receives an amount of
energy represented by Eg={Eg,1, Eg,2, ..., Eg,c}, to cover
his/her needs partially, Eg,i < Ea,i, or totally, Eg,i =Ea,i,
depending on his/her priority of service within the MG. In this
work, MG communities with users having their own battery
storage are not considered. Using batteries in houses incurs
capital and maintenance costs. Furthermore, battery round-
trip efficiency has to be taken into account, i.e., power losses
during charging-discharging. We assume that the aforemen-
tioned costs and losses are undertaken by the utility (company,
business operator) that controls CC. In addition, in our study
case, we assume a small neighborhood where we consider
neither the losses when CC stores/distributes energy nor the
physical limitations of the distribution grid (seen in larger
residential areas).
IV. MET HO DO LO GY
A. Characterization
We use fine-grained data regarding consumption of ap-
pliances and generation by renewable energy sources. Us-
ing the consumption/generation data, we compute the de-
ficiency/excess of energy for every household. To achieve
convergence, we smooth the daily (and hourly) differences in
energy by averaging the measurements over weekly intervals.
To associate every household with the others in its community,
we use clustering to distribute households into different groups
(clusters). In this paper, we use the Expectation-Maximization
(EM) algorithm to define the exact number of clusters that can
best accommodate the households regarding their attributes
(e.g., consumption, generation), and distribute every house-
hold uniquely to one cluster (c). To acquire energy consump-
tion/generation perspective of households over longer periods
of time (e.g., yearly), the metrics of temporal membership
and adaptability are used. Cluster membership refers to the
presence of a household in one of the clusters that are defined
for an energy attribute and cluster adaptability refers to the
transition between different clusters of the same attribute in
consecutive time intervals (clustering periods) [4], [14]. The
terms temporal membership and temporal adaptability assess
the probability that a household is a member of a cluster or
performs a cluster transition. For the analysis, we considered
anonymized data.
B. Energy Allocation Strategies (EAS)
To show the evaluation of strategies, we mention simple
strategies but delve more into the optimal strategies and
provide in-depth discussion. All EASs are found in [4].
Simple allocation strategies create prosumer-consumer
pairs, and the energy flows from the prosumer to the consumer
of each pair, using the power line. CC is used only for routing.
Random strategy: Every prosumer sends information about
his/her available energy to the CC and the CC chooses
randomly a consumer to allocate the energy. If the consumer
is covered fully, the remaining energy is allocated randomly
to another.
Greedy strategy: The CC lists consumers in a priority se-
quence and they are served as the sequence dictates. Energy is
transferred by every prosumer to its corresponding consumer-
pair by First-In-First-Served. In the greedy approach, the order
of service is the same for every time interval. This order rela-
tion results in consumers being served in the same sequence
at every time interval, leading to dissatisfied consumers in the
community. To ensure fair energy allocation, we propose the
λlevel of service.λis a percentage limit of service, imposed
on every household. When this limit is reached the following
household will be served, and consequently more households
will be served with the same amount of energy.
Round-robin strategy: This mechanism ensures that served
households in an interval are moved to the end of the service
sequence. This sequence is initially created by the priority
policy at T= 1. At T= 2, the algorithm moves the previously
served households to the end of the service sequence (and
redefines it). This mechanism continues until a predefined
limit of time intervals, called Time-Limit (T L), is reached. T L
reveals the number of service rounds until reinitialization; it
resets the service sequence at Tmod T L = 0. Consequently,
T L defines the depth of service diversity.
In optimal allocation strategies, CC, besides routing,
stores energy too; and computes the amount to be distributed
to every consumer. Optimal EASs define Relations of Weight
when serving the consumers. Weights are assigned to the
members of C. The exact amount of energy to be received
by a consumer is found using his/her weight as follows,
p
i=1
Ee,i =x
c
j=1
wj, where at first, the total amount of
energy that is saved by the prosumers during a time interval
is gathered at CC. Then, by using the weights wgiven
to every consumer of C, the single unit of energy, x, is
computed, and every consumer, j, receives an amount of
energy corresponding to xwj[4]. Within the community,
weight-ratios between consumers dictate differences in the
amounts of energy that they are entitled to. As the ratio
between the assigned weights of two consumers increases,
(a) Game Theoretic (b) Water Filling
Fig. 2: Optimal Energy Allocation Strategies
the difference in the amount of energy allocated to each of
them also increases.
Weighted strategy: The total excess energy, on every Tdu-
ration, is gathered by the central controller (CC). The CC
splits consumers Cinto Nsubgroups, C=N
n=1Cn. To
each subgroup, it assigns a weight, wn, same for all the
consumers of a subgroup (n). The highest weights are as-
signed to the subgroups of prioritized consumers. The priority
policies used by this EAS are based on size and energy
(deficiency) attributes, for increased accuracy of prioritization.
The excess energy from pprosumers is distributed according
to
p
i=1
Ee,i =x
N
n=1
(wnCn).
Algorithm 1: Game Theoretic (GT)
consumer and prosumers are indexed by kand i
At the beginning:
1: CC assigns weights wkk[1, c]according to a CPP
At each time interval
Initialization phase:
2: CC collects excess energy from prosumers, p
i=1 Ee,i
3: Consumers Csend their deficiencies Eato the CC
4: CC defines the heights of service Husing H=Ea/w
Energy Allocation phase:
5: while c
k=1 Hk>0do
6: CC chooses non-zero minimum height of service, min(H)nz
7: if (min(H)nz c
k=1 wk)p
i=1 Ee,i then
8: Ea,k Ea,k min(H)nzwk,k[1, c]
9: p
i=1 Ee,i p
i=1 Ee,i (min(H)nz c
k=1 wk)
10: Consumer with min(H)nz is fully served
11: wmin(H)nz = 0
12: HHmin(H)nz
13: else
14: min(H)nz p
i=1 Ee,i
c
k=1 wk
15: Ea,k Ea,k min(H)nzwk,k[1, c]
16: Break
17: end if
18: end while
Game Theoretic strategy (GT): In GT all the consumers seek
energy according to their weights from the CC simultaneously,
as shown in Fig. 2a. They withdraw only when they are fully
served. The concept behind this algorithm relies on Game
Theory, and specifically on the existence of an equilibrium
based on the choices of non-cooperative consumers-players
on energy allocation, where everyone is bound to a certain
decision. After assigning a different weight, w, to each
Algorithm 2: Water-Filling (WF)
Prosumers are indexed by i
j,lrepresent the indices of the most and least prioritized
consumer being served simultaneously
At the beginning:
1: CC assigns weights wkk[1, c]according to a CPP
At each time interval
Initialization phase:
2: CC collects the excess energy from prosumer, p
i=1 Ee,i
3: Csend info on their deficiencies Eato the CC
4: CC defines initial heights of service by H=Ea/wand forms
them in ascending order, Hini
5: j= 1, l = 1
6: HHini
Energy Allocation phase:
7: while j6cdo
8: Perform GT algorithm for energy allocation phase on the
following:
group of (l+ 1 j)consumers with weights assigned in
step 1 with p
i=1 Ee,i and additional heights h
if l+ 1 6c, hk={Hl+1 Hk,if Hl+1 <2Hini,k
2Hini,k Hk,otherwise
else hk= 2Hini,k Hk
for k: [j, l]
After GT algorithm:
9: Total excess decreased, p
i=1 Ee,i updated
10: Individual deficiencies of (l+ 1 j)households decreased
or covered, Ea,k updated k[j, l]
Updating the Heights of service:
11: aj
12: for k: [a, l]do
13: HkHk+hk,
14: if Hk= 2Hini,k then
15: jj+ 1
16: end if
17: end for
18: if (Hl=Hl+1 or Hl= 2Hini,l ) and l+ 1 6cthen
19: ll+ 1
20: end if
21: end while
consumer according to the imposed priority policy, the CC,
holding information about the deficiency of all the consumers,
defines the ratios of deficiency and weight, termed Levels
of Service,H, with H=Ea/w. The amounts of energy
that are individually received fit into the specific energy and
socioeconomic characterization of each consumer.
Water-Filling strategy (WF): At the beginning, different
weights are given to each consumer by the CC depending
on the priority policy that is followed. Then, being informed
regarding the deficiency of each consumer, the CC defines
their H. However, in this EAS, the CC arranges the Hof the
consumers in ascending order, which becomes their order of
service. The difference between this algorithm and the GT is
that some consumers can ask for energy before others. Many
consumers often have to wait until the prioritized households
are fully covered, as can be seen in Fig. 2b. Let us assume
that the transferred energy is added on top of the Hof
every consumer, as additional service-level,h=Eg/w. As
the CC starts sharing energy with the first consumer in the
order of service, its level, h1increases until h1=H2H1.
Then, assuming there is enough excess energy stored, the CC
starts transferring to the second consumer in the order too;
until h2=H3H2=h1(H2H1)h1=H3H1.
This procedure continues until the need of every consumer is
covered or the energy is depleted. A consumer jis withdrawn
from service only when fully covered (hj=Hj). For two
consumers, jand l, with Hl> Hj, it is also possible that
HlHj>Hj, and thus the consumer jis fully covered
before lstarts requesting for energy. A number of consumers
can be served simultaneously at any time instance, as long
as they have equal sums of Hand h(cf. Fig. 2b).
V. EXPERIMENTAL EVALUATI ON
To test our EASs, we employed the readily available and
standard Pecan Street dataset, which is located in Texas
Austin and composed of 443 households. Among them, 180
households generate energy using solar panels. We used one
year of consumption and generation data (in kW) from the
smart meters of all the households, and we computed the
deficiency and excess of energy for every household. The
smart meters offered fine-grained data for accurate analysis.
We only selected those households having data for more than
300 days. At first, we analyzed the metrics that focus on
households being served. These metrics refer to the consumers
of an MG community. Thus, for a consumer k, we answer with
1 (true) or 0 (false) the following questions; (a) Is kserved
fully?, (b) Is knot served at all?, (c) Is it the first time that
kis served in timespan T?
To quantify the potential of a strategy in covering com-
pletely the needs of (a group of) consumers cwithin a
community, we define the Served Ratio (SR) metric for Tas,
SR =
c
k=1
Cserved,k/c. To evaluate the efficiency of prosumers
in serving (a group of) consumers during T, we define the
Prosumers Beneficial Ratio (PBR), PBR =
c
k=1
CnotServed,k/p.
Low values of PBR imply efficient prosumer usage. For the
EASs that use priority sequences for consumer service, we use
Uniqueness Ratio (UR), which quantifies the service diversity
of a sharing strategy for (a group of) consumers for any
set of consecutive time intervals, denoted as TbTa, with
Ta, Tb[1, Tmax], UR =
Tb
T=Ta
c
k=1
CT
unique,k/c.
To quantify satisfaction regarding the service offered to a
consumer during a timespan T, we use the ratio of the amount
of energy given to a household (or a group) and its total
energy sought. We term this ratio Energy Ratio (ER) and, for a
consumer k, during T, the ER is defined as ERk=Eg,k/Ea,k .
When ER = 0, no energy is received. However, to evaluate
fairness in service we have to consider the priority that
every household possesses within its group. Under a priority
policy, the coverage of deficiency of every household impacts
differently the community. Prioritized households are more
important in terms of service and should receive higher
(a) Membership (b) Adaptability
Fig. 3: Temporal energy behaviors
amounts of energy than the rest. For a consumer k, applying a
weight that mirrors his/her significance in the community turns
ER into its weighted form, ERw,k =wkERk. To evaluate it,
we use the log2relation to define the Social Welfare (SW) for
any consumer k, SWk=wklog2(1 + ERk). However, SWk
cannot be characterized as high or low and thus fairness in
serving consumers according to their significance cannot be
evaluated by SW. It needs to be compared with the maximum
possible value of SWk. Obviously, when a consumer is fully
served ERk= 1, and then SWk,max =wk. Thus the metric to
characterize every consumer regarding the fairness in energy
allocation is the Social Welfare Ratio (SWR), defined as
SWRk=SWk/wk. In order to expand the individual-SW
to group-SW, or further, to SW for a whole community of c
consumers, we have, SWRc=
c
k=1
log2(1 + ERk).
VI. IM PL EM EN TATION RESULTS
We evaluate the temporal energy behavior of households
using membership and adaptability. In Fig. 3a, the x-axis
shows the clusters in terms of consumption; c1represents low
consumption and c5high. Further, the position of the clusters
on x-axis represents cluster centroids. The yearly membership
ratio for a household being in a particular cluster is θm.
In Fig. 3a, about 400 households consumed low amounts
of energy, out of which, 115 households were in c1for
more than 75% of the year (white). This result implies that
115 households can be prioritized by policies that focus on
low deficient consumers. In Fig. 3b, the x-axis presents the
beneficial cluster transitions in consumption. For a household,
the ratio of particular cluster transitions (x-axis) over all the
performed transitions is θt. Direct transitions between two
non-consecutive clusters (e.g., c3to c1) are rare, because they
demand higher energy regulation potential from the house-
holds. As shown in Fig. 3b, most of the households regulate
their consumption between c1, c2, and c3; this explains the
higher numbers of households in these clusters (Fig. 3a).
In Fig. 4a and Fig. 4b, we present SR for different target
groups of consumers, created based on energy deficiency and
size. These groups are served for three consecutive months,
using round-robin and greedy EASs. As seen in Fig. 4b,
round-robin EAS serves households from different groups–
not only from the prioritized ones. Moreover, under the round-
robin strategy, because of the repositioning of highly deficient
(a) Greedy (b) Round-robin
Fig. 4: Average SR according to different priority policies
(a) PBR (b) UR
Fig. 5: Prosumer usage – Service diversity (week 38-49)
consumers at the end of the service sequence, high deficiency
and large size priority policies serve more households. The
opposite happens for the policies prioritizing small and less
deficient consumers. In Fig. 5a and Fig. 5b, we evaluate how
efficiently the prosumers are used (PBR) and how diverse
is the consumer service. Note that the lowest values present
the most efficient behaviors as the PBR metric is related to
the consumers not served weekly by the prosumers. Among
the EASs that serve consumers in sequential order, the WF
sharing approach utilizes the prosumers more efficiently than
the other approaches, keeping at the same time a satisfactory
UR (0.5, Fig. 5b). Because of no priority in serving, the
random approach has much lower PBR (Fig. 5a) and high
diversity; serving almost 85% of the consumers (Fig. 5b).
Service-fairness in a community is described by the SWR
metric. In Fig. 6a, the advantage of optimal algorithms against
the simple approaches on energy sharing is clear –they provide
higher fairness in service for every particular priority policy.
Further, generally by choosing policies that prioritize the less
deficient consumers we manage to serve more households
than by promoting the highly deficient ones, because the
prioritized households are easily served. On the contrary,
high deficiency policy aims to serve those in high needs
requiring large amounts of excess energy. The performance
of the random policy stays between other policies, as it gives
priority to none. Specifically, for the WF and GT EASs, under
the same policy, weights, deficiency, and stored energy, WF
EAS manages higher SWR. Focusing only on these two EASs,
in Fig. 6b and Fig. 6c, their impact on different groups of
households (which have been assigned with the same priority
(a) Overall
(b) Game Theoretic (c) Water Filling
Fig. 6: Community social welfare ratio (yearly)
weights) is observed. WF prioritizes the targeted household
groups stricter –maximizing SW for the members of these
groups. On the other hand, in GT strategy the social welfare
results for different groups of households are closer because
all households receive energy simultaneously. Note here that
the big-sized or the highly deficient groups of households
present deficiencies that are not covered easily, thus the impact
of their weights in SWR is lower than the impact of other
groups when they are prioritized. The WF approach presents
overall higher SWR results per priority policy, as confirmed
by Fig. 6a. Further, the GT EAS is more stable than WF,
because in WF we observe more outliers.
VII. CONCLUSION
With the growing adoption of renewable energy sources,
consumers and prosumers are able to redistribute energy
efficiently. ICT infrastructures provide communication needed
between consumers and prosumers to share the available
energy locally, avoiding energy transportation losses. In ad-
dition, prosumers have higher economic benefits by selling
excess energy locally compared to selling it to the central
stations. In this paper, we proposed and evaluated six EASs
that could be easily computed at the edge of SGs, which
control the allocation of excess energy in an MG community.
We considered many novel approaches such as using both
fine-grained energy-data and social attributes to exploit the
temporal energy dynamics of communities. Our approach is
novel in the way we characterize an MG community. We
clustered households into multiple groups thereby making it
easy to analyze the complex behaviour of the community.
We show that there is no “one-size-fits-all” strategy when
prioritizing households and distributing the excess energy in
an MG since the energy needs of households in a community
keep varying while energy harvested also varies. We analyzed
one year of data from 443 houses to test our algorithms and
their impact. The most optimal allocation strategy was WF,
having the highest social welfare ratio, higher by a factor
of 2.5 compared to greedy approaches. This work provides
many knobs to control energy allocation under various sce-
narios with different focuses. This work is one of the highly
comprehensive studies of energy sharing in MGs of small
consumers/prosumers. Expecting every household to be a
prosumer in future it will be interesting to evaluate the scaling
potential of our EASs in a system of distributed MGs.
ACKNOWLEDGEMENT
This research was carried out within the SCOTT project
(scott-project.eu) funded from the Electronic Component Sys-
tems for European Leadership Joint Undertaking under grant
agreement No 737422. This joint undertaking is supported
from European Union’s Horizon 2020 research and innova-
tion program and Austria, Spain, Finland, Ireland, Sweden,
Germany, Poland, Portugal, Netherlands, Belgium, Norway.
REF ER EN CE S
[1] E. Mengelkamp et al., “Designing microgrid energy markets: A case
study: The Brooklyn Microgrid,” Applied Energy, 2018.
[2] I. S. Bayram and T. S. Ustun, “A survey on behind the meter energy
management systems in smart grid,” Renewable and Sustainable Energy
Reviews, 2017.
[3] Y. Wang et al., “Deep learning-based socio-demographic information
identification from smart meter data,” IEEE Transactions on Smart Grid,
May 2019.
[4] N. Kouvelas, “Energy allocation strategies for Micro-grids,MSc thesis,
Delft University of Technology, 2017.
[5] “Pecan Street,” www.pecanstreet.org, [Online; accessed 25-Aug-2019].
[6] T. Morstyn et al., “Using peer-to-peer energy-trading platforms to
incentivize prosumers to form federated power plants,Nature Energy,
2018.
[7] C. Long et al., “Peer-to-peer energy sharing through a two-stage
aggregated battery control in a community microgrid,” Applied Energy,
2018.
[8] T. AlSkaif et al., “Reputation-based joint scheduling of households
appliances and storage in a microgrid with a shared battery,Energy
and Buildings, 2017.
[9] W. Shi et al., “Real-Time Energy Management in Microgrids,” IEEE
Transactions on Smart Grid, Jan 2017.
[10] T. Liu et al., “Energy management of cooperative microgrids: A
distributed optimization approach,” International Journal of Electrical
Power and Energy Systems, 2018.
[11] W. Ma et al., “Distributed energy management for networked microgrids
using Online ADMM with Regret,” IEEE Transactions on Smart Grid,
March 2018.
[12] Y. Du et al., “A cooperative game approach for coordinating multi-
microgrid operation within distribution systems,” Applied Energy, 2018.
[13] W. Tushar et al., “A motivational game-theoretic approach for peer-to-
peer energy trading in the smart grid,” Applied Energy, 2019.
[14] A. U. Nambi et al., “Temporal self regulation of energy demand,” IEEE
Transactions on Industrial Informatics, 2016.
Article
The electricity grid, using Information and Communication Technology, is transformed into Smart Grid (SG), which is highly efficient and responsive, promoting two-way energy and information flow between energy-distributors and consumers. Many consumers are becoming prosumers by also harvesting energy. The trend is to form small communities of consumers/prosumers, leading to Micro-grids (MG) to manage energy locally. MGs are parts of SG that decentralize the energy flow, allocating the excess of harvested energy within the community. Energy allocation amongst them must solve certain issues viz., 1) balancing supply/demand within MGs; 2) how allocating energy to a user affects his/her community; and 3) what are the economic benefits for users. To address these issues, we propose six Energy Allocation Strategies (EASs) for MGs – ranging from simple to optimal and their combinations. We maximize the usage of harvested energy within the MG. We form household-groups sharing similar characteristics to apply EASs by analyzing energy and socioeconomic data thoroughly. We propose four evaluation metrics and evaluate our EASs on data acquired from 443 households over a year. By prioritizing specific households, we increase the number of fully served households to 81% compared to random sharing. By combining EASs, we boost the social welfare parameter by 49%.
Article
Full-text available
Peer-to-peer trading in energy networks is expected to be exclusively conducted by the prosumers of the network with negligible influence from the grid. This raises the critical question: how can enough prosumers be encouraged to participate in peer-to-peer trading so as to make its operation sustainable and beneficial to the overall electricity network? To this end, this paper proposes how a motivational psychology framework can be used effectively to design peer-to-peer energy trading to increase user participation. To do so, first, the state-of-the-art of peer-to-peer energy trading literature is discussed by following a systematic classification, and gaps in existing studies are identified. Second, a motivation psychology framework is introduced, which consists of a number of motivational models that a prosumer needs to satisfy before being convinced to participate in energy trading. Third, a game-theoretic peer-to-peer energy trading scheme is developed, its relevant properties are studied, and it is shown that the coalition among different prosumers is a stable coalition. Fourth, through numerical case studies, it is shown that the proposed model can reduce carbon emissions by 18.38% and 9.82% in a single day in Summer and Winter respectively compared to a feed-in-tariff scheme. The proposed scheme is also shown to reduce the cost of energy up to 118 ¢ and 87 ¢ per day in Summer and Winter respectively. Finally, how the outcomes of the scheme satisfy all the motivational psychology models is discussed, which subsequently shows its potential to attract users to participate in energy trading.
Article
Full-text available
Peer-to-peer (P2P) energy sharing allows the surplus energy from distributed energy resources (DERs) to trade between prosumers in a community Microgrid. P2P energy sharing is being becoming more attractive than the conventional peer-to-grid (P2G) trading. However, intensive sensing and communication infrastructures are required either for information flows in a local market or for building a central control system. Moreover, the existing pricing mechanisms for P2P energy sharing could not ensure all the P2P participants gain economic benefits. This work proposed a two-stage aggregated control to realize P2P energy sharing in community Microgrids, where only the measurement at the point of common coupling (PCC) and one-way communication are required. This method allows individual prosumers to control their DERs via a third party entity, so called energy sharing coordinator (ESC). In the first stage, a constrained non-linear programming (CNLP) optimization with a rolling horizon was used to minimize the energy costs of the community. In the second stage, a rule based control was carried out updating the control set-points according to the real-time measurement. The benefits of P2P energy sharing were assessed from the community’s as well as individual customers’ perspective. The proposed method was applied to residential community Microgrids with photovoltaic (PV) battery systems. It was revealed that P2P energy sharing is able to reduce the energy cost of the community by 30% compared to the conventional P2G energy trading. The modified supply demand ratio based pricing mechanism ensures every individual customer be better off, and can be used as a benchmark for any P2P energy sharing model. For consumers, the electricity bill is reduced by ∼12.4%, and for prosumers, the annual income is increased by ∼£57 per premises.
Article
Full-text available
Smart meters provide large amounts of data and the value of this data is getting increased attention because a better understanding of the characteristics of consumers helps utilities and retailers implement more effective demand response programs and more personalized services. This paper investigates how such characteristics can be inferred from fine-grained smart meter data. A deep convolutional neural network (CNN) first automatically extracts features from massive load profiles. A support vector machine (SVM) then identifies the characteristics of the consumers. Comprehensive comparisons with state-of-the-art and advanced machine learning techniques are conducted. Case studies on an Irish dataset demonstrate the effectiveness of the proposed deep CNN-based method, which achieves higher accuracy in identifying the socio-demographic information about the consumers.
Article
Full-text available
Power networks are undergoing a fundamental transition, with traditionally passive consumers becoming ‘prosumers’ — proactive consumers with distributed energy resources, actively managing their consumption, production and storage of energy. A key question that remains unresolved is: how can we incentivize coordination between vast numbers of distributed energy resources, each with different owners and characteristics? Virtual power plants and peer-to-peer (P2P) energy trading offer different sources of value to prosumers and the power network, and have been proposed as different potential structures for future prosumer electricity markets. In this Perspective, we argue they can be combined to capture the benefits of both. We thus propose the concept of the federated power plant, a virtual power plant formed through P2P transactions between self-organizing prosumers. This addresses social, institutional and economic issues faced by top-down strategies for coordinating virtual power plants, while unlocking additional value for P2P energy trading.
Article
Full-text available
Due to the decreasing revenues from the surplus renewable energy injected into the grid, mechanisms promoting self-consumption of this energy are becoming increasingly important. Demand response (DR) and local storage are among the widely used mechanisms for reaching higher self-consumption levels. Deploying a shared storage unit in a residential microgrid is an alternative scenario that allows households to store their surplus renewable energy for a later use. However, this creates some challenges in managing the battery and the available energy resource in a fair way. In this paper, a reputation-based centralized energy management system (EMS) is proposed to deal with these issues by considering households’ reputations in the reallocation of available energy in the shared storage unit. This framework is used in an optimization problem, in which the EMS jointly schedules households’ appliances power consumption and the energy that each household can receive from the storage unit. The scheduling problem is formulated as a Mixed Integer Linear Programming (MILP) with the objective of minimizing the amount and price of energy absorbed from the main grid. The MILP problem is coded in GAMS and solved using CPLEX. Numerical analysis is conducted using real data of renewable energy production and appliances’ demand profiles for different classes of households and different annual periods in Spain. Simulation results of the different scenarios show that by using the proposed framework higher cost savings can be achieved, in comparison with the classical scheduling scenario. The saving can reach up to 68% when different classes of households exist in the microgrid. The results also show that the fairness in energy allocation is guaranteed by the reputation-based policy, and that the total power absorbed from the main grid by the whole microgrid is significantly decreased.
Article
Full-text available
We propose a distributed algorithm for online energy management in networked microgrids with a high penetration of distributed energy resources (DERs). A high penetration of DERs introduces high uncertainty of power generation to the microgrids. In general, the state-of-the-art forecasting for non-dispatchable DERs such as solar energy is not sufficiently accurate, which results in inaccurate energy scheduling. To address the high uncertainty issue in the networked microgrids, we propose an online energy management based on the online alternating direction method of multipliers algorithm with the past power generation information from the DERs. The online algorithm provides less conservative schedule than the robust optimization based approach. The effectiveness of the proposed algorithm is verified by various numerical examples.
Article
The cooperation of multiple networked microgrids (MGs) can alleviate the mismatch problem between distributed generation and demand and reduce the overall cost of the power system. Energy management with direct energy exchange among MGs is a promising approach for improving energy efficiency. However, existing methods on microgrid cooperation usually overlook the underlying distribution network with operating constraints (e.g., voltage tolerance and power flow constraints). Hence the results may not be applicable to actual systems. This paper studies the energy management problem of multiple MGs that are interconnected by both the direct current (DC) energy exchange network and the alternating current (AC) traditional distribution networks. In our problem, each MG is equipped with renewable energy generators as well as distributed storage devices. In order to handle the non-convex power flow constraints, we exploit the recent results of the exact optimal power flow (OPF) relaxation method which can equivalently transform the original non-convex problem into a second-order cone programming problem and efficiently determine the optimal solution successfully. The objective of our problem is to minimize the overall energy cost in a distribution network consisting of multiple MGs, with the practical operating constraints (e.g., power balance and the battery’s operational constraints) explicitly incorporated. Considering the privacy and scalability, we propose a distributed algorithm with convergence assurance based on the alternating direction method of multipliers (ADMM). We also implement our method based on the model predictive control (MPC) approach in order to handle the forecasting errors of the renewable energy generation. Simulations are made for different MG exchange topologies on three radial distribution network testbeds. Numerical results demonstrate that certain topologies are more favorable than others, and the cooperation strategy for the energy exchange is significantly affected by the MGs’ locations in the distribution network.
Article
Generation from distributed renewable energy sources is constantly increasing. Due to its volatility, the integration of this non-controllable generation poses severe challenges to the current energy system. Thus, ensuring a reliable balance of energy generation and consumption becomes increasingly demanding. In our approach to tackle these challenges, we suggest that consumers and prosumers can trade self-produced energy in a peer-to-peer fashion on microgrid energy markets. Thus, consumers and prosumers can keep profits from energy trading within their community. This provides incentives for investments in renewable generation plants and for locally balancing supply and demand. Hence, both financial as well as socio-economic incentives for the integration and expansion of locally produced renewable energy are provided. The efficient operation of these microgrid energy markets requires innovative information systems for integrating the market participants in a user-friendly and comprehensive way. To this end, we present the concept of a blockchain-based microgrid energy market without the need for central intermediaries. We derive seven market components as a framework for building efficient microgrid energy markets. Then, we evaluate the Brooklyn Microgrid project as a case study of such a market according to the required components. We show that the Brooklyn Microgrid fully satisfies three and partially fulfills an additional three of the seven components. Furthermore, the case study demonstrates that blockchains are an eligible technology to operate decentralized microgrid energy markets. However, current regulation does not allow to run local peer-to-peer energy markets in most countries and, hence, the seventh component cannot be satisfied yet.
Article
Over the last few years, the fast-growing energy needs across the world have intensified a central challenge: how to reduce the generation and operation costs in power systems and, in parallel, to minimize the hydrocarbon emissions. Moreover, one-quarter of world's population still lacks access to electricity, as the cost of building conventional power grids is not affordable by third world countries. On the other hand, behind-the-meter (BTM) energy systems offer cost-effective solutions to aforementioned challenges, as they enable end-users to satisfy their energy needs with distributed energy generation and storage technologies. To that end, this paper presents a detailed survey of BTM energy management systems. The paper starts with the classification of the electrical loads with respect to their physical properties, priority ranking, and sizes. Next, the literature on BTM energy management systems is systematically classified into three main categories: technology layer, economic layer, and social layer. The technology layer spans the studies related to power systems including distributed generation and storage technologies, whereas the economic layer shows how economic incentives along with optimization and scheduling techniques are employed to shape the energy consumption. The social layer, on the other hand, presents the recent studies on how to employ social sciences to reduce the energy consumption without requiring any technological upgrades. This paper also provides an overview of the enabling technologies and standards for communication, sensing, and monitoring purposes. In the final part, a case study is provided to illustrate an implementation of the system.