Simulation & Emulation platform for smart grid technologies

Abstract

Energy management has become a very popular topic in the past few years, especially with the protection of the environment. The use and management of renewable energy sources (RES) has therefore become a necessity. In this context, innovative EMSs (Energy Management Systems) are required to bring a new way to monitor energy flows in micro grids. The interaction between physical and digital elements is relying on the use of Information and Communication Technologies (ICTs). A side effect is the appearance of Multi-Agent System (MAS) whose aim is to manage efficiently energy flows to reduce consumer bills, to minimize the use of fossil fuel in favor of RES. In this paper, we introduce "GYSOMATE", a simulation platform to test different management strategies, with the implementation of predictive tools and MAS to improve the EMS.

Full text

Simulation & Emulation platform for smart grid
technologies
Yassine Gangat
Energy, Electronic, and Processes
Laboratory (LE²P) & Computer
Sciences and Mathematics Lab. (LIM)
University of La Réunion
Saint Denis, Réunion, France
yassine.gangat@univ-reunion.fr
Nicolas Coquillas
Energy, Electronic, and Processes
Laboratory (LE²P)
University of La Réunion
Saint Denis, Réunion, France
nicolas.coquillas@univ-reunion.fr
Dominique Grondin
Energy, Electronic, and Processes
Laboratory (LE²P)
University of La Réunion
Saint Denis, Réunion, France
dominique.grondin@univ-reunion.fr
Michel Benne
Energy, Electronic, and Processes
Laboratory (LE²P)
University of La Réunion
Saint Denis, Réunion, France
michel.benne@univ-reunion.fr
Taher Issoufaly
Energy, Electronic, and Processes
Laboratory (LE²P)
University of La Réunion
Saint Denis, Réunion, France
taher.issoufaly@univ-reunion.fr
Jean-Pierre Chabriat
Energy, Electronic, and Processes
Laboratory (LE²P)
University of La Réunion
Saint Denis, Réunion, France
jean-pierre.chabriat@univ-reunion.fr
Abstract— Energy management has become a very popular
topic in the past few years, especially with the protection of the
environment. The use and management of renewable energy
sources (RES) has therefore become a necessity. In this context,
innovative EMSs (Energy Management Systems) are required
to bring a new way to monitor energy flows in micro grids. The
interaction between physical and digital elements is relying on
the use of Information and Communication Technologies
(ICTs). A side effect is the appearance of Multi-Agent System
(MAS) whose aim is to manage efficiently energy flows to
reduce consumer bills, to minimize the use of fossil fuel in
favor of RES. In this paper, we introduce “GYSOMATE”, a
simulation platform to test different management strategies,
with the implementation of predictive tools and MAS to
improve the EMS.
Keywords—smart grids, HIL, multi-agent system,
I. INTRODUCTION (HEADING 1)
Our modern society is based on energy. Whenever there
is a power outage, most of our activities are paralyzed. On
the one hand, the energy demands are increasing
exponentially: growth in population, enhancement of
building services and comfort levels, in conjunction with the
rise of time spent inside buildings, have raised their energy
consumption to the levels of transport and industry. [1]
On the other hand, continuation of fossil fuel uses will
bring us to confront difficulties such as consumption of non-
renewable energy source, environmental alteration, etc.
These issues show an unsustainable situation. [2]
Renewable energy sources (RES) such as solar or wind
energy is the answer to this challenge. But these sources are
intermittent and vary through the day. [3]
The Energy Management Systems (EMS) is a response to
the energy challenges and the concern for the protection of
the environment of industrialized and developing countries.
It brings together techniques to reduce energy consumption
for the sake of both financial economy and reducing the
ecological footprint.
Reunion Island, which is located in a tropical area,
provides a fertile ground for research in this domain.
Due to its small surface area (2500 km²), nuclear power
generation cannot be implemented on the island: first, the
unitary size is oversized in relation these needs; then, the
maintenance of such a structure is impossible.
In addition, despite its small area, Reunion Island
presents the same energy problems as in many countries of
the world, including:
The presence of many energy producers, from
individual level to a collective one.
A large number of consumers distributed in a non-
homogeneous way.
A lot of different microclimates. [4]
Different energy source types.
These parameters make the island an ideal place to
generate research and innovation in the field of the rational
use of energy.
A smart grid in Reunion Island should be built around
these two following axes:
an EMS that takes into account the topology of our
electricity distribution networks
a weather prediction tool that should be able to
consider microclimates
Moreover, the intermittency, stochastic variability and
low predictability of RES, including the solar resource, as
well as the distributed nature of these units, require the
development of intelligent, decentralized control strategies
for all the elements such as conversion sources, storage
systems and consumers.
Distributed energy resources (DER) expansion represents
a change in the paradigm of power grids, making micro
generation and storage units relevant, both in terms of
sustainable development and demand side energy (DSE). In
this context, the security of small and isolated to large-scale
interconnected grids depends on an efficient EMS, resulting
in a constrained optimization problem including two stages:
unit commitment (UC) and economic dispatch (ED). Since
the objective (e.g. balance equation) and constraints
(technical limits, operational and security constraints) are
expressed as linear functions, involving discrete and
The 2018 International Conference on Control, Electronics, Renewable Energy and Communications (ICCEREC)
978-1-7281-1993-9/18/$31.00 ©2018 IEEE 122

Fig. 1. Agent’s hierarchy
continuous variables, UC and ED can be formulated as a
Mixed-Integer Linear Programming (MILP) problem.
Commercial solvers as GUROBI, CPLEX or EXPRESS are
able to solve MILP problems and considering thousands of
variables and constraints.
Distributed control, monitoring and energy management
have proved to achieve operational efficiency of
interconnected DER, and many publications present multi-
agent systems (MAS) as one of the best technology for
introducing distributed intelligence in large scale systems.
But developing smart grids in Reunion Island will require
research & development that suits the island's characteristic.
It cannot be achieved with theories and simulation practices
alone. Several smart grid simulations have already been
compared, such as: MatPower, PSAT, InterPSS and
GridLAB-D in [5], OpenDSS, GridLAB-D and APREM in
[6]. But these simulations lack circumstances that physical
experiment includes.
Developing a test-bed such as in [7] will be useful for
testing control, communication but it won't suffice for a
whole smart grid.
This study presents the first step in the development of an
intelligent EMS taking into account the energy balance for an
urban micro grid with battery storage, and using MAS to
improve the communications scheme between the distributed
facilities, and the commercial solver GUROBI to minimize
cost and facility constraints. A testing platform called
GYSOMATE (french acronym for Supervision, Dynamic
Management and Optimization of Urban Micro grids for
Island Electricity Self-sufficiency) is proposed.
GYSOMATE will not only simulate but also emulate a fully
customizable Smart grid. Thanks to this platform we will be
able to test different EMS and prediction tools.
II. DESIGN OF GYSOMATE’STESTING PLATFORM
The purpose of our testing platform is to provide a
sandbox in order to test several approaches of EMS and
prediction. Our goal is not to provide the best EMS or
prediction method, but rather to allow comparison between
different configurations. This platform has been built from
these 5 following modules:
a Multi-Agent Systems (MAS)
a Real-Time Simulator (OP4510 from Opal-RT)
a nanogrid build from Raspberry-Pi & Arduino
mostly
a Human Machine Interface (HMI).
a NoSQL Database (InfluxDB)
The multi-agent system is not only a module, but it is
also the binder between all modules as we will see further.
A. The Multi-agent System
Multi Agents Systems have already shown its strength in
addressing issues related to complex systems involving
ecological, economic and social dynamics [8]. In addition to
this, Multi Agents offer formalizations that take into account
both collective management and person-centered modeling
[9]. The coupling between the individual and the collective,
realized through the interactions, as well as the semantic
richness of the spatialized environments make MAS a
particularly adapted approach for our model of energy
management [10] [11] [12].
The potency of our MAS architecture is that it's not only
in charge of the EMS as in most of MAS such as
Powermatcher and Volttron. It also serves as a prediction
tool and a binder between GYSOMATE's components.
1) MAS Architecture
We have chosen an architecture that allows MAS to be
the barycenter of our platform. At the beginning of the
project, we chose the SKUAD framework (Software Kit for
Ubiquitous Agent Development) developed by the Collective
Adaptive Systems team of the LIM (Laboratory of Computer
Science and Mathematics).
One of the reasons we have chosen this framework is that
it has already all the leeway scope we need for GYSOMATE
such as it includes Raspberry Pi & Arduino API. Moreover,
we can work together to develop new needed APIs.
But after analyzing the various articles dealing with the
subject, we realized that JADE (Java Agent Development
Framework) was one of the most used multi-agent platforms.
We decided to create parallel MAS with JADE, which was
not difficult because both JADE and SKUAD are written in
Java.
An agent hierarchy as shown in the Fig.1 has been
implemented to structure agents to fit our needs while being
close to physical hardware distribution and network
topology. There are 4 types of agents:
Grid (in green): A Grid agent can "manage" another
Grid agent or Supervisor agent.
Supervisor (in yellow): A Supervisor agent always
depends on a Grid agent. It "manages" Aggregator
agents of different types.
Aggregator (in blue): An Aggregator agent is
intended to aggregate Device Agents of the same
type. These aggregators are optional.
The 2018 International Conference on Control, Electronics, Renewable Energy and Communications (ICCEREC)
978-1-7281-1993-9/18/$31.00 ©2018 IEEE 123

Device (in orange): A Device Agent is an agent that
is directly linked to a physical device that it controls
(AC, light bulbs, switch, etc.). These devices could
either be controlled or not.
The basic principle is each agent send back data to his
"superior" and acts regarding its superior's orders. However,
in some cases (i.e.: when the superior does not answer for
example, etc.), the agents can communicate with other
agents. The naming of the agents will be as follows:
Town: Agent in charge of the city.
Neighborhood: Agent in charge of a neighborhood.
Building: Agent in charge of a building or at least the
end-user (housing in a building). It can also be a
battery or a PV farm.
Load: all possible loads of the building.
Storage: all types of energy storage (hydrogen,
battery, etc.).
Production: all types of energy production.
Our agents will first start with an 'initialization step' by
checking its internal components (social links, physical links
and battery status) before the agent starts its specific job:
behaviors for either proactive activities (i.e.: an agent asks
the current load of a battery) and reactive activities (i.e.: an
agent will send his status whenever he is asked).
Each agent possesses some local multi-objective
optimization method, controls the voltage and belongs to a
global-objective and participates in voltage regulation of a
complete system. All agents are coordinating together in a
cooperative way.
2) MAS as an EMS
After setting up our MAS for its basic functions, we
focus on EMS. We have emulated (through Raspberry-PI)
and simulated (through OP4510) a physical environment (i.e.
a study case). In order to prove that our MAS will act as an
EMS, we have formulated a very simple EMS that will be
described in the chapter III.
The MAS has been built in such a way that we can easily
replace the EMS by anything we want to code, as it has been
separated from the kernel of the MAS.
3) MAS as a prediction tool
Our MAS has also been connected to our DB and is
currently using Datamining's techniques such as
classification and clustering. For each hour, we build a
sliding horizon of classes. For each class, we have associated
a set of behaviors adapted to it. This prediction method is
used for both energy production and consumption.
In the same way as for the EMS, the prediction part of
the agent has been clearly identified and could be changed by
anything we want to implement.
4) MAS as a binder
In order to bind all the GYSOMATE modules together,
we decided to “agentify” them. Through integration in the
multi-agent's space, the OP4510 has now become a
ubiquitous component and is now reachable by any agents.
We have used the communication protocol library
implemented in SKUAD, called Ubiquity. Indeed, it uses the
concept of “space” (set of grouped elements interconnected)
with the implementation of UDP protocol with an application
layer that allows the management of sending and receiving
packets along with the security of data sent on the channel.
This mode of communication will enable the interaction
between each element regardless of their type or their
specification (database, JADE agent, simulator…). This
generic aspect is a key factor for our proposed model.
Agents are also able now to access the DB and send logs
to the HMI. Agents have also been implemented on
Raspberry-pi and are now able to connect to Arduino in
order to access real world's sensors and actuators.
Multi-agent system can operate in both real-time and
simulated environment, giving us the best of both.
B. Real-time simulator
1) Concept
Today, many tools are emerging concerning the analysis
of power systems. However, smart grids use case becomes
more and more complex to meet the constraints in terms of
energy efficiency. We find out that one tool cannot handle all
the processes to achieve performance results. Therefore, the
combination of several tools is required. In response to these
growing challenges, real-time simulation platforms and
suites are very suitable for integrating models, software and
hardware components. This proposal is a concept approach
that allows the analysis of large-scale scenarios. [13]
Many works in this domain first introduced a new
method of rapid prototyping called HIL (Hardware-In-the-
loop Simulation), a simulation that implements the modeling
of the studied environment and the physical equipment in
closed loop, to bring more solutions for this challenge. The
method was initially applied in the automotive industry [14]
and has been used successfully to estimate the behavior of
continuous analog processes under various control laws [15],
[16]. The implementation of a HIL system is efficient in
terms of computing power and speed of execution to
simulate in real time the physical behavior of the systems
with accuracy. Besides, development costs are significantly
reduced by the introduction of a HIL system.
By definition, Hardware-In-the-Loop (HIL) characterizes
a numerical simulation of the functioning of an electronic
control unit (ECU) for which the hardware environment is
simulated: the inputs and outputs of the tested system
(hardware) are connected to a computer which reproduces
the functioning of the environment
As explained in [17], HIL simulation is being
increasingly used as an important design and development
step in the manufacturing process of many industries [18].
HIL simulation provides a means for the operation of
physical hardware, such as power components and control
hardware, while interfaced to a computer simulation of the
system in which the physical hardware is intended to
function.
Opal-RT is leading testing equipment use for the
hardware-in the loop and software-in-loop testing for various
electro mechanical, electrical and power electronics system.
Opal-RT is used for validating results of system simulations
which is to be tested in real time.
2) Architecture
The 2018 International Conference on Control, Electronics, Renewable Energy and Communications (ICCEREC)
978-1-7281-1993-9/18/$31.00 ©2018 IEEE 124

Fig. 2. Nanogrid
In fact, the real time simulation is realized with a real-
time platform which is named RT-LAB, which is perfectly
integrated with MathWork Matlab/Simulink.
RT-LAB converts Simulink models to implement them in
a real-time environment, and then run them on one or more
processors
There are two ways of communication between the host
processors and the execution processors, the first via physical
inputs/outputs, the second through an Ethernet connection.
3) Interface
The developed EMS must be able to act on the simulator.
It is therefore necessary to build a communication tunnel
which passes, on the one hand, the controls of the energy
management system to the models simulated by RT-LAB,
and on the other hand, which transfers the state variables of
the simulator to the energy management system. To achieve
this goal, we designed the architecture, to establish a socket
for the model (service), then, to allow the agents, to pass the
commands and the measures via this socket with a structured
format of the data.
Each service (model) simulated in the RT-LAB has a
physical counterpart in the energy management system; the
correspondence is made by the alias of the actuator agent
dedicated to the service.
The configuration of this system is done in a Json file
with different parameters such as codes of instructions; data
send or receive, agents aliases, etc.
Nanogrid
It is very complex to connect production, storage and
consumption units for information feedback in real time and
even more for piloting, particularly because of both juridical
and technical aspect. Thus, we have decided to develop
within the laboratory a connected and controllable nanogrid.
This nanogrid would be built from:
Emulators for RES,
Emulators of devices consuming energy,
Energy storage devices,
Sensors and Actuators,
Few electrical appliances connected and controllable
within the laboratory (air conditioning, water heater).
The architecture deployed will be similar to that studied
in the different cases of studies. Emulators, sensors, and
actuators will be deployed using Arduino and Raspberry Pi
microcontrollers. These devices are inexpensive and open,
providing a greater degree of freedom in development.
The 2018 International Conference on Control, Electronics, Renewable Energy and Communications (ICCEREC)
978-1-7281-1993-9/18/$31.00 ©2018 IEEE 125

TABLE I. OUR PARTNER'S AND OUR LAB'S DATA
Type
Number
of sites
Settings
Time
step
Period
Photovoltaic
Production
27
Power
(Converter
output)
10 min
Up to 6
years
Sun power
14
Irradiance,
Temperature
10 min
1 min
(LE²P)
Up to 10
years
(LE²P)
Tertiary
consumption
3
Energy
(General and
AC)
10 min
More than
1 year
Fig. 3. Study case
The nanogrid will operate on a 12V voltage level for the
emulated part. The storage device will be dimensioned to be
integrated at this voltage level. The link with other electrical
equipment will be via the energy management system and
more particularly the multi-agent platform. This real
nanogrid can be connected to other virtual nanogrid from the
OP4510 real-time simulation.
The configuration of an elementary brick of this nanogrid
is given in the Fig.2. The connection of microcontrollers via
Wi-Fi has been considered because it allows a management
of Arduino wirelessly by the Raspberry Pi in a transparent
way.
The introduction of this nanogrid brings us several
advantages. It allows us in particular to visualize live-effects
of the EMS and to interact with the system in a real way. In
addition, given the low cost of the equipment and the fact
that it is only running on 12V DC, the eventual damage will
be limited in the case of a bad EMS.
C. Human Machine Interface
A multilevel HMI is under development. The target
audience for this HMI is the following people:
The end-user: this is a person who has access to a
supervising agent and all agents underneath.
The "neighborhood" responsible user: this is a person
who has access to the first level of the grid agent, who
is under the supervision of the supervisory agents. He
will have access to only partial and/or anonymized
information.
The responsible user of the "city": this is a person
who has access to the second level of agents’ grid,
which is under the responsibility of other agents’
grids.
For each user, the HMI has several views:
Located view: a view that shows agent localization,
available actions and logs.
Graphic view: a view with mostly graphics and alert
trigger
Dashboard: a simple view that could also be rendered
as a WebApp, with minimal information
Configuration: this where we can configure rules for
the ESM or prediction.
The final HMI is still under development but the
prototype we used for our testing is operational.
The HMI is initialized by an XML file that contains the
scenario configuration (id, nature, information of each agent
and elements) and is fed by data and logs from agents. It can
also send orders or requests to some agents.
D. NoSQL Database
In our project, we have several types of data, from
different sources:
1. Energy-related Data (production, consumption and
storage) from our partners,
2. Energy related Data and logs from our simulations
and emulations,
3. Energy related Data and logs from our Laboratory
(sensors, prediction, etc.)
Energy production, consumption and storage data are
time series with values. The logs are also time series
containing text. Among the Time Series Data Bases,
InfluxDB was chosen particularly because several services
are already ready: graphics, export, calculation of averages,
API web/rest, management of alert, etc. InfluxDB comes
with a stack that includes all the tools needed to set up the
DB. The Table 1 shows a summary of our partner's and our
Lab's data.
III. PRELIMINARY RESULT
To put forward a first version of our system, we decided
to focus on an elementary brick composed of one element of
storage, production and consumption. This environment is
presented with the Fig.3. There is a house wired with the
utility grid which represents the consumption part equipped
with a photovoltaic plant that aims to produce energy and a
battery for the storage. Indeed, it is a very simple example,
but it will assure us that everything is working together.
The architecture in our system to reproduce the elemental
brick will use the OPAL-RT simulator which creates all
physical elements by interconnecting them. For each
physical element, we will associate them with an agent
implemented by our multi-agent part. Therefore, we’ll have a
battery agent plugged to the battery, a charge agent linked
with the consumption element and a PV agent for the
photovoltaic plant. A supervisor agent will be created to
manage these previous listed agents.
The scenario is described as follows: the charge part is
simulated with a consumption profile save into a file read by
the simulator to generate consumption values according to
time. For the generation of photovoltaic production value, we
used real data collected in our database, recorded in a file
and read according to time by the simulator. Periodically, the
battery agent will ask for the current level of the battery. The
charge agent will do the same process by collecting the
The 2018 International Conference on Control, Electronics, Renewable Energy and Communications (ICCEREC)
978-1-7281-1993-9/18/$31.00 ©2018 IEEE 126

Fig. 4. Results
generated consumption. Finally, the PV agent will get the
production values by interacting with the photovoltaic
physical element in the simulator. All of this information will
be sent to the supervisor. After analyzing them, the latter will
monitor the energy source by interacting with the
photovoltaic production and the battery. Indeed, he will
choose to use the energy of the photovoltaic production or
the energy stored in the battery according to the consumption
curve simulated.
The preliminary results are described with the Fig.4.
There are four curves. The first one represents the
photovoltaic production according to time. The second one
refers to the state of charge of the battery. The third curve is
the position of the switch button and the last indicates the
consumption of energy according to the time.
The supervisor agent can take three actions with a switch
button: use the production energy (position 1), use the energy
stored in the battery (position 3) or take to the energy from
the electrical network (position 2). At the beginning, the
energy stored in the battery is used.
We can see that when the photovoltaic production
increases and is superior to the consumption, the agent will
switch the position to 1 and take this source of energy to
meet the consumption as shown with green boxes.
The second case is highlighted with blue boxes. There is
a peak of consumption and the production of photovoltaic
panels is low. The only solution is to take energy from the
battery. This is the reason why the supervisor agent switches
the position to 1.
Finally, the last situation is represented by purple boxes.
In this scenario, the production of photovoltaic is not enough
to meet the need of the consumption. Besides, the energy
stored in battery is low. The action taken by the supervisor
agent will consist of plugging to the electrical network to
ensure the necessary needs for consumption (switch position
to 2).
These results clearly show us that all modules of our
platform were well connected through a simplistic model of
elemental brick. This has also shown the impact of the multi-
agent system for better energy management. These tests
ensure the robustness of the basic elements of our system
which will allow the development of other more complex
models in the future.
IV. CONCLUSION & PERSPECTIVES
In recent years, EMS (Energy Management System)
represented centralized systems with pre-defined and non-
dynamic algorithms. The introduction of multi-agent
systems, which is very popular, has led to the restructuring of
this type of system in order to monitor, control and optimize
the generation and transmission of energy sources. New
architectures tend towards a decentralized approach with the
competence of the agents to act according to the environment
in which they are located.
Plethora of energy management systems are presented in
the literature. In this paper, we focused on the development
of a testing platform called “GYSOMATE”. The aim is to
implement a generic EMS with the use of multi-agent system
allowing the interconnection of different entities. The
preliminary results emphasize the success of the
interconnection of entities through sending and receiving
packets with the same protocol of communication using a
simplistic model of elemental brick.
Our next step will be to reflect the complexity of real-
world smart grid including unpredictable renewable energy
sources by using more complex scenarios as illustrated in the
Fig 5.
In the future, our generic system will improve with the
use of predictive algorithms to help agents for their decision
making. The latter will propose different methods such as
classification, determinist approach, game theory or
stochastic process. The aim is to simulate the behavior of our
agents using a set of algorithms and analyze the results to
choose the better one according to the context. Besides,
implementation of optimization algorithms is being studied
whose first experiment was conducted with the GUROBI
solver. Finally, the introduction of negotiation process
between agents for the energy management is also planned.
In this case, there is no supervisor agent that sends orders to
inferior agents. All agents will be at the same level and will
interact with each other to find themselves in the most
accurate solution to optimize sources of energy.
The 2018 International Conference on Control, Electronics, Renewable Energy and Communications (ICCEREC)
978-1-7281-1993-9/18/$31.00 ©2018 IEEE 127

Fig. 5. Perspectives
To conclude, “GYSOMATE” platform aims to become a
generic tool allowing us to easily plug different entities and
analyze the results of the different simulations to have a
better overview of agents’ behavior and actions taken while
finding the right compromise according to the
context/constraints.
ACKNOWLEDGMENT
Special acknowledgments go to the European Regional
Development Funds (ERDF) and the Regional council of
Reunion for GYSOMATE project funding, the project’s
partners Corex Solar, TEEO. We also address special thanks
to Dr. Denis Payet, from the Lab. of Computer Science and
Mathematics (LIM) of University of La Reunion for the
SKUAD framework.
REFERENCES
[1] Pérez-Lombard, L., Ortiz, J., & Pout, C. (2008). A review on
buildings energy consumption information. Energy and buildings,
40(3), 394-398.
[2] Atilgan, B., & Azapagic, A. (2015). Life cycle environmental impacts
of electricity from fossil fuels in Turkey. Journal of Cleaner
Production, 106, 555-564.
[3] Carrasco, J. M., Franquelo, L. G., Bialasiewicz, J. T., Galván, E.,
PortilloGuisado, R. C., Prats, M. M.,... & Moreno-Alfonso, N. (2006).
Power-electronic systems for the grid integration of renewable energy
sources: A survey. IEEE Transactions on industrial electronics, 53(4),
1002-1016.
[4] David, M., Andriamasomanana, F. H. R., & Liandrat, O. (2014).
Spatial and temporal variability of PV output in an insular grid: Case
of Reunion Island. Energy Procedia, 57, 1275-1282.
[5] Pöchacker, M., Sobe, A., & Elmenreich, W. (2013). Simulating the
smart grid. arXiv preprint arXiv:1304.2503.
[6] Moffet, M. A., Sirois, F., & Beauvais, D. (2011). Review of open
source code power grid simulation tools for long-term parametric
simulations. CanmetENERGY, Tech. Rep., 137.
[7] Leskarac, D., Bennett, C., Moghimi, M., Stegen, S., & Lu, J. (2015,
November). Testing facility for research and development of Smart-
MicroGrid technologies. In Power and Energy Engineering
Conference (APPEEC), 2015 IEEE PES Asia-Pacific (pp. 1-5). IEEE.
[8] Bousquet, F., Barreteau, O., Le Page, C., Mullon, C., & Weber, J.
(1999). An environmental modelling approach: the use of multi-agent
simulations. Advances in environmental and ecological modelling,
113, 122.
[9] Ferber, J., & Weiss, G. (1999). Multi-agent systems: an introduction
to distributed artificial intelligence (Vol. 1). Reading: Addison-
Wesley.
[10] Gangat, Y., Duizabo, A., Courdier, R., Payet, D., Bessafi, M., &
Chabriat, J. P. (2016, October). Simulation of microgrids using a
multi-agent system and measurements of solar ressources on the
island of Reunion (original title : Simulation de microgrids à l'aide
d'un système multi-agents et de mesures de gisements solaires sur l'ile
de La Réunion.) In JFSMA'16 (pp. 247-248). Cepadues.
[11] Gangat, Y., Courdier, R., & Payet, D. (2009, October). Energy
management of a territory-a multi-agent simulation approach (original
title : Aménagement énergétique d'un territoire-une approche par
simulation multi-agents). In Journées Francophones Systèmes Multi-
Agents (JFSMA’09) (pp. 237-240). Cepadues.
[12] Labeodan, T., Aduda, K., Boxem, G., & Zeiler, W. (2015). On the
application of multi-agent systems in buildings for improved building
operations, performance and smart grid interaction–A survey.
Renewable and Sustainable Energy Reviews, 50, 1405-1414.
[13] Opal-RT, http://www.opal-rt.com, accessed September 2018
[14] Wei Deng, Wei Pei, Ziqi Shen, Zhenxing Zhao, "IEC 61850 based
testbed for micro-grid operation control and protection", Electric
Utility Deregulation and Restructuring and Power Technologies
(DRPT) 2015 5th International Conference on, pp. 2154-2159, 2015.
[15] Patricio Cortes, David O. Boillat, Johann W. Kolar, "Phase-oriented
control of a modular 3-Phase 3-level 4-leg inverter AC power source
supplying floating or grounded loads", Energy Conversion Congress
and Exposition (ECCE) 2013 IEEE, pp. 1357-1363, 2013.
[16] Patricio Cortes, David O. Boillat, Hans Ertl, Johann W. Kolar,
"Comparative evaluation of multi-loop control schemes for a high-
bandwidth AC power source with a two-stage LC output filter",
Renewable Energy Research and Applications (ICRERA) 2012
International Conference on, pp. 1-10, 2012.
[17] Ren, W., Sloderbeck, M., Steurer, M., Dinavahi, V., Noda, T.,
Filizadeh, S., ... & Belanger, J. (2011). Interfacing issues in real-time
digital simulators. IEEE Transactions on Power Delivery, 26(2),
1221-1230.
[18] Dufour, C., Abourida, S., & Belanger, J. (2005, November).
Hardware-in-the-loop simulation of power drives with RT-LAB. In
Power Electronics and Drives Systems, 2005. PEDS 2005.
International Conference on (Vol. 2, pp. 1646-1651). IEEE.
The 2018 International Conference on Control, Electronics, Renewable Energy and Communications (ICCEREC)
978-1-7281-1993-9/18/$31.00 ©2018 IEEE 128