![](bg2.png)
discuss the background about Smart Grids and their
reliability. In section 3 we discuss about adaptation to
the Mosaik framework (Schütte et al., 2011) for dy-
namic changes to topologies for SG failure modelling.
In section 4 we showcase the modified co-simulation
platform with an experimental evaluation to compare
different patterns of failure emergence. Section 5 con-
cludes the paper.
2 BACKGROUND
A Smart Grid (SG) can be seen as a modern
power grid enabling two-way power flow and at
the same time bi-directional communication between
power suppliers and consumers (Fang et al., 2011).
Controllers, sensors, computer systems, automation
equipment are integrated to provide efficient trans-
mission of electricity, fast restoration in case of fail-
ures, and overall reduced costs for utilities (Goel
et al., 2015). SGs achieve lower power costs for
consumers, reduced peak demand, increased integra-
tion of large-scale renewable energy systems—among
other benefits. Real-time monitoring and recovery of
power generation and distribution is another key char-
acteristic, as the actual state of the grid is monitored
and reported to the network, adapting the power out-
put to the real needs. SGs are also important to in-
crease the usage of renewable sources (e.g., solar en-
ergy), as excess energy generated can be sold, more-
over attempting to reach balance in demand response
programs (Siano, 2014).
Decentralization of the SG led to the introduction
of microgrids. A microgrid is an independent and
small network of electricity users (consumers / pro-
sumers) that can carry out operations independently
from the centralized grid and even isolate itself from
the rest of the power network in case of failure of
the grid, with the so-called islanding mode (Hebner,
2017). The decentralization attempts to avoid sin-
gle points of failure and "domino effects" of failures,
leading potentially to large blackouts.
Internet of Things (IoT) devices also play an im-
portant role in the context of SG, as they open the way
to so-called smart energy scenarios. For example, a
household using a solar-power system (with batter-
ies and sensors) can decide on the best moments for
recharging a Electric Vehicle (EV) (Hebner, 2017).
The many sensors, devices, automation equipment
and different layers pose many challenges both in
terms of security and reliability concerns, with the
smart grids expecting to provide self-monitoring and
self-healing capabilities.
Reliability of a software system is defined as the
probability that the system will function as required
without failures and errors for a certain period of time
(O’Connor, 2012). From this definition, reliability
can be seen as associated to the concept of quality of
service. In traditional grids, reliability, and security of
cyber elements were not considered as critical for the
overall stability of power grids because of the less rel-
evant dependencies between cyber and physical parts.
This situation changed drastically with Smart Grids:
the incorporation of modern IT aspects moved SG
more towards cyber-physical systems, which brings
tighter constraints related to security and reliability.
In modern grids, the cyber part is essential for the
proper functioning of the whole power grid, as it pro-
cesses sensors’ data, monitors the grid, handles se-
curity, and makes power distribution decisions (Lei
et al., 2018). The physical part is thus strongly depen-
dent on the availability of the cyber layer. Power-grid
stability needs to take into account also possible cyber
failures. However, SGs can have a relative advantage
over traditional grids when examining fault-tolerance
and fault-recovery: Supervisory Control And Data
Acquisition (SCADA) systems employed in SGs can
communicate with a multitude of sensors in real-time.
In case of failures, SCADA systems can locate the
area subject to the failure and start self-healing and
notification activities. Such large availability of data
can support a multitude of anomaly detection algo-
rithms and platforms (Rossi and Chren, 2020; Lip
ˇ
cák
et al., 2019).
As every CPS, also SG can be modelled formally
as the interaction of reactive systems constrained by
temporal constraints. Defining formally the compo-
nents can allow to find conditions under which the
constraints are violated. Some parts can be even
solved by means of analytical equations. However,
the complexity of the interactions and the compu-
tational complexity required by the many solvers,
forces in many cases to rely on the usage of simu-
lations. However, due to the many system states, sim-
ulations can be only use to disprove the correctness
constraints defined by invariants in the falsification
process, that is simulations cannot cover all the possi-
ble run cases (Alur, 2015).
Due to the complexity for analytical solutions,
many models were proposed over time for failure
propagation in power grids, each one covering differ-
ent aspects (Guo et al., 2017; Cai et al., 2016; Xiao
and Yeh, 2011): topological models (based on net-
work analysis), stochastic simulation models (proba-
bilistic simulations), statistical models, dynamic sim-
ulation, and interdependent models (studying cou-
pling of interdependent networks cyber and physical).
Since traditional power grids were designed and
COMPLEXIS 2021 - 6th International Conference on Complexity, Future Information Systems and Risk
64