IEEE Systems, Man and Cybernetics Magazine - July 2021 - 16

challenge is that the generation plus storage must match
the demand. If the server admits more packets than
those available on the supply side, some of the clients
will not be served, which decreases the quality of the
service provided. Another important aspect is the possibility
of emergency messages. If a client's requests are
rejected many times, the client may send an emergency
message to convey that its packets need to be treated
with high priority. Emergency messages might also be
sent if some appliances or devices have to be used
urgently. Different approaches could be utilized as part
of the admission control process, for example, queuing
theory, machine learning, and probabilistic forecasting
of weather and energy.
2) Classification: The accepted energy packets may be separated
into priority classes. For computational purposes,
it is preferable to have a limited number of priorities.
However, the preferred number of classes and the techniques
to perform the classification are open questions.
Energy routers can preprocess a request and classify it at
the household level, or the classification can be related to
the appliance type. The classification task can then be
built with algorithms using rule-based approaches, hierarchical
classifications, or statistical methods. Advances in
machine learning and artificial intelligence promise good
performance for dynamically classifying the requests.
3) Availability: Once the accepted packets are classified, the
server needs to check the availability of the energy packets
for the upcoming time slot(s); i.e., how many packets
can be served considering the following constraints: 1) the
uncontrollable baseload, 2) the active flexible loads that
cannot be turned off after they are started (e.g., using dishwashers),
3) the short-term generation forecasts, and 4)
the possibility of using storage. This will define the system
capacity for that specific slot. The capacity can be divided
into smaller slices based on priority classes; i.e., some
packets can be reserved for different packets based on
their priority. We call this energy network slicing (see
Figure 5). Machine learning and predictive algorithms for
inventory management can be expected to provide good
supporting tools for achieving this task.
4) Allocation: With the packets classified and availability
defined, the server needs to select which packets will be
allocated for delivery in the upcoming slot(s). In this
case, the server will fill the vacant positions of each slice
with packets from its associated class. Other prioritization
rules may be developed; for example, a closer deadline
is first served, or an earlier arrival is first served.
Alternative options, such as random rules, can also be
devised and used. Research into resource allocation in
software-defined networks, including different optimization
approaches and queuing theory, may provide a
good background for building and assessing different
approaches to the availability problem.
To illustrate the concept, let us consider a toy example
consisting of a case where the server has two priority
16 IEEE SYSTEMS, MAN, & CYBERNETICS MAGAZINE July 2021
levels to classify the packets. Additional flexibility is
provided by the energy buffer. The energy packets need
to be delivered over the next time slots considering a
time horizon of, for instance, 30 slots. Once the packets
are classified, they can be allocated and delivered using
energy network slices, which are simply the number of
packets in each priority class that are allocated in a
given slot. It is important to remember that the proposed
implementation may be vulnerable to cyberattacks [36].
A new element of the SDEN-mapping the role of the
firewall in the Internet-needs to be designed to guarantee
the safe operation of a virtual microgrid. A systematic
investigation of such a protective agent is, however,
beyond the scope of this article and should be the focus
of future research.
Figure 5 presents the energy network slicing technique
considering the following setting: 1) two slices, namely priorities
1 (green) and 2 (blue), considering the available
energy from the available sources; and 2) additional packets
available from the buffer/cache (yellow). This shows
that the proposed management is capable of directly generating
a specific demand curve using packet multiplexing.
Note that this profile can be dynamically modified to
reflect the actual state of the system and guarantee the
balance between supply and demand based on the operational
needs of the devices and the wishes of the end users.
Moreover, this approach does not assume any price signals
or cost-related optimization techniques, and it only considers
direct usage requests.
Toward a Bright Future With
Free Electricity Based on Sharing?
The proposed solution aims to achieve a future fossil-fuelfree
energy system governed as a commons, i.e., a system
with abundant " clean " electric energy that is shared and not
held privately [19]. In our proposed solution, such a commons-based
renewable electrification paradigm can be
achieved by prioritizing electricity sharing within the
microgrid over purchasing from the main grid. Nevertheless,
the solution also needs to comply with the current electricity
market structure to develop, expand, and eventually
become the dominant means by which electricity is supplied.
This is the main argument in the work of Nardelli et
al., 2019 [20], where the authors discuss the concept of the
energy Internet. In fact, the proposed SDENs and virtual
microgrids are not too far from existing commercial solutions
(see, e.g., Klein et al. [11]), and so the idea is generally
compatible with European markets. Scaling up electricity
sharing is also feasible with the use of exclusive group
(EXG) bidding schemes [14], [37], where different load profile
curves are bid in the market. In the market clearance, a
load curve is selected and provided to the bidder, who, in
turn, needs to " generate " such a demand profile (otherwise,
it needs to cover the costs related to up- or downregulation).
This approach is an interesting way of managing the flexibility
of loads, but it has an important drawback: if the EXG
IEEE Systems, Man and Cybernetics Magazine - July 2021

Table of Contents for the Digital Edition of IEEE Systems, Man and Cybernetics Magazine - July 2021
