1164 IEEE TRANSACTIONS ON SMART GRID, VOL. 7, NO.

2, MARCH 2016

Railway Energy Management System:

Centralized–Decentralized Automation
Architecture
Sara Khayyam, Student Member, IEEE, Ferdinanda Ponci, Senior Member, IEEE, Javier Goikoetxea,
Valerio Recagno, Valeria Bagliano, and Antonello Monti, Senior Member, IEEE

Abstract—The modern railway system is a massive grid con- examples of distributed systems trying to achieve optimal
nected complex system with distributed active loads (trains), energy scheduling, while Costanzo et al. [6] proposed a sys-
sources (particularly distributed renewable sources), and storage tem architecture for smart buildings and the focus of [7] is
(wayside or on-board storage systems). Its energy management
therefore requires the concepts and techniques used for man- optimal energy management of SGs considering unpredictable
aging energy in the smart grid (SG). Accordingly, the new load demands and DERs.
railway energy management system (REM-S) is developed to In the context of SG, energy management is already a key
integrate on-board, wayside, and coordination services. REM-S is issue for railway systems as a massive distributed system. The
driven by the idea that regeneration, loads, storage, and volatile variety of operational scenarios within the system adds com-
distributed energy resources should be coordinated dynami-
cally to achieve optimal energy usage. This paper presents plexity to the solutions suitable for all users, i.e., operators
the proposed REM-S architecture, which is based on a hybrid and infrastructure managers (IM) alike. In this sense, several
centralized–decentralized concept and developed according to European projects have been carried out since 2005 in railway
SG architecture model framework. systems, such as MODURBAN (http://www.modurban.org/)
Index Terms—Automation architecture, energy management, and OSIRIS (http://www.osirisrail.eu/) focusing on urban
railway system, smart grid (SG). railway systems or Railenergy (http://www.railenergy.org/)
with more technological emphasis on energy efficiency
I. I NTRODUCTION
improvement. Holistic, coherent information systems which
NERGY management is currently a big challenge for pol-
E icy makers, utilities, industry managers, engineers, and
even residential consumers. Reduction of emissions, 2020 and
integrate major railway subsystems is the focus of projects
like InteGRail (http://www.integrail.eu/).
However, the findings of previous projects lack an inte-
2050 EU energy targets [1], [2] are aimed at mainly through grated approach and they cannot tackle the energy man-
increase of distributed energy resources (DERs) penetration, agement for the entire rail network. Hence, the MERLIN
net reduction of consumption, management of energy flows project (http://www.merlin-rail.eu/) was defined to investigate
for maximum usage of the available renewable energy, and and demonstrate the viability of an integrated EMS and to
creation and exploitation of flexibility. In this context, the achieve a more sustainable and optimized energy usage in the
European railways have committed to reduce their own emis- European electric mainline railway systems. This implies that
sions by 30% by 2020. Considering that in the European energy consumers, producers, and storages are not isolated
railways, the share of electricity from distributed resources elements, but players of the global energy game. A smart and
is dramatically increasing [3], updating energy management coordinated contribution of each of them brings more savings
methods consistently is of great importance. and provides more flexibility for the system to manage the
Distributed energy management systems (EMSs) are energy flow more efficiently.
frequently used in smart grid (SG) solutions. Home The example of energy flow in the European electric main-
EMSs [4], [5], smart city quarters, and smart cities are line railway system shown in Fig. 1 indicates that the amount
of energy regenerated by the train is comparable with the
Manuscript received June 30, 2014; revised November 20,
2014 and February 25, 2015; accepted March 21, 2015. Date of energy consumption of “NO TRACTION LOADS” (signal-
publication May 5, 2015; date of current version February 17, ing, switches, etc.) or “OTHER LOADS” [stations, workshops,
2016. This work was supported by the European Union through electric vehicle (EV) charge stations, etc.]. Regenerating
FP7 Project MERLIN (http://www.merlin-rail.eu/) under Grant 314125.
Paper no. TSG-00661-2014. energy on trains, together with wayside storage and on-board
S. Khayyam, F. Ponci, and A. Monti are with the E.ON Energy train storage, brings notable flexibility to railway system and
Research Center, RWTH Aachen University, Aachen 52074, Germany (e-mail: supports it in being an important actor in the electricity market
[email protected]).
J. Goikoetxea is with CAF Company, Beasain 20200, Spain. by selling surplus energy to the public grid or by reduc-
V. Recagno and V. Bagliano are with D’Appolonia Company, Genoa 16145, ing energy purchases following market price variations in the
Italy. electricity market.
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. In order to achieve this goal, the railway energy manage-
Digital Object Identifier 10.1109/TSG.2015.2421644 ment system (REM-S) developed in MERLIN and presented
1949-3053 c 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: b-on: Instituto Politecnico de Lisboa. Downloaded on July 25,2024 at 17:59:02 UTC from IEEE Xplore. Restrictions apply.
KHAYYAM et al.: REM-S: CENTRALIZED–DECENTRALIZED AUTOMATION ARCHITECTURE 1165

Fig. 2. SGAM framework [9].

Fig. 1. Energy flow of the European country case.

communication requirements for the ICT infrastructure and
new business objectives must be defined. To this aim, the
here, integrates on-board, wayside, and coordination services SGAM framework [9] was applied as common language.
by developing a system that monitors the energy consump- The SGAM originally developed in support of the SG
tion of different railway subsystems and their components, standardization process, can be used as aid in designing SG
and then suggests a “smart” solution for coordinating optimal architectures in a structured manner according to its five inter-
energy usage in the different parts of the system. The contri- operable layers (business, function, information, communica-
bution of this paper consists in the REM-S architecture, and tion, and component), zones and domains [9]. Fig. 2 illustrates
in the demonstration that SG concept and the SG architecture the SGAM layers, zones, and domains.
model (SGAM) tools and framework are applicable to railway In this paper, after clarifying the REM-S objectives and con-
systems. cept, relevant use cases are defined [10], [11] and analyzed.
REM-S implements two automation architecture stand- This phase yields the definition of the architecture for the
points: 1) centralized; and 2) decentralized. According to function layer and business layer. Then the development
railway system specifications there is possibility to partition of component layer, information layer, and communication
the system to local areas and also there is possibility to layer is shown together with the standardization gap analysis
define local or global targets for system optimization. These for harmonization of SG and railway communication stan-
two main factors along with some other factors such as dards. Eventually, some key performance indicators (KPIs)
level of system complexity, size of the system, information for REM-S architecture verification are presented and
and communication technology (ICT) structure and depend- discussed.
ability of different system layers influenced the choice of
a hybrid centralized–decentralized concept for REM-S. In the II. REM-S AUTOMATION C ONCEPT
hybrid centralized–decentralized REM-S architecture, while The objectives of REM-S are defined as follows.
the global EMS is executed in control center considering the 1) Optimize energy consumption in operating the railway
whole railway network for the following day, the local EMS system while ensuring the fulfillment of the applicable
will be done in local subnetworks during every 15 min. performance requirements.
The existing SG standards, communication protocols, and 2) Optimize power demand of operating the railway sys-
ICT technologies are suitable for designing centralized– tem ensuring the fulfillment of the applicable perfor-
decentralized automation architecture. Hence, the design of mance requirements. The reduction of power demand,
the architecture presented here is based on the representa- and especially of power peaks, can free the electricity
tion of the railway distribution system as an SG. Besides, it network capacity which has a direct effect on invest-
accommodates different time horizons. With reference to the ment strategies for network development; also, it may
latter feature, in [8], the model for data management in con- reduce the global energy bill and limit or avoid financial
trol center of SG is studied at three time window modes. This penalties.
architecture must be able to deal with different stakeholders, 3) Optimize costs relevant to energy consumption needed
applications, and networks, and it must be interoperable with to operate the railway system while ensuring the fulfill-
the public grid and energy markets. The development of this ment of the applicable performance requirements. The
architecture requires harmonization of the standards, proto- energy costs can be lowered through a more rational
cols, and data models of different domains. Furthermore, the purchasing strategy, for example buying energy at a

Authorized licensed use limited to: b-on: Instituto Politecnico de Lisboa. Downloaded on July 25,2024 at 17:59:02 UTC from IEEE Xplore. Restrictions apply.
1166 IEEE TRANSACTIONS ON SMART GRID, VOL. 7, NO. 2, MARCH 2016

low price (during off-peak hours), storing it and using it

when the price is higher (peak hours). By this approach
(that is the target of cost optimization in REM-S); a cost
reduction can be achieved even without reducing the
total energy consumption.
The consumption optimization is driven by the idea that
energy regenerated or spared by some actors in the railway
system can be distributed to other actors thus leading to a net
decrease in the energy demand to the public grid.
Because of the distributed nature of the railway system,
the system size, complexity, and uncertainties, as well as the
dynamic and moving nature of the loads, the energy man-
agement must be based on dividing the system into different
parts (subnetworks), and must be implemented in a hybrid
centralized–decentralized way. Fig. 3. REM-S automation architecture concept.
In the proposed automation architecture, each subnetwork
is in contact with the control center and through it with the
electricity market via one intelligent interface substation. The (electricity market), it makes sense to adopt a similar
subnetworks receive the global optimization plan from the con- time structure for the optimization, yielding three opera-
trol center and implement it locally in their own area, locally tional modes.
accommodating unanticipated mismatches. Each subnetwork 1) Day Ahead Optimization (DAO): Calculates the opti-
is in contact with neighboring subnetworks, in order to coor- mum behavior of the network, including power profiles,
dinate and resolve its border area issues about energy and energy and power purchase, power sale, and so on, for
power optimization. Multiagent systems (MAS) technology is the next day time horizon (24 h).
employed for developing automation and control system in 2) Minutes Ahead Optimization (MAO): Locally predicts
subnetworks [12]. Each subnetwork consists of the following and optimizes the following 15 min of subnetwork
active entities. status. Following the DAO profile, MAO covers the
1) An intelligent substation (ISST) in communication with interaction with all subnetwork agents, considering
all energy related components within the subnetwork, power restrictions in the subnetwork as well as the
to send energy consumption/generation commands or surpluses and needs of the adjacent subnetworks and
suggestions to them. Each energy related component according to them suggests actions to subnetwork agents
equipped with an intelligent entity, called agent, and such as SSTs, RSSTs, DERs, ESSs, or DOEMs of the
has the ability to communicate and the intelligence to trains passing through the subnetwork in the next 15 min
make decisions whether to follow the commands or sug- time step.
gestions coming from ISST. ISST is the subnetwork 3) Real-Time Operation (RTO): Fulfills the calculated
manager and acts as main agent in its own area. 15 min MAO profiles for the subnetwork, taking into
2) Several reversible substations (RSST) and nonreversible account the real-time status and behavior of all the
substations (SST) connected to public grid as fixed components of the subnetwork.
agents in negotiation with the main agent of the Fig. 3 depicts the hybrid centralized–decentralized concept
subnetwork. with the three operational modes related to time and location.
3) Wayside energy storage systems (ESSs) considered as In this architecture, by DAO a global EMS runs for the whole
fixed agents. railway network (limited to one IM domain) yielding energy,
4) DERs: Renewable sources which belong to the rail- power, and cost optimization with a top-bottom approach
way system and are located in subnetwork area are based on train timetables and power profiles, DERs genera-
considered as fixed agents as well. tion and ECs power demand. In minutes ahead time slices,
5) Dynamic On-Board Energy Managers (DOEMs) a local EMS is executing in the area of each subnetwork with
Installed on the Trains: They are responsible for energy the target of following the global EMS plan and optimizing
management inside the train and are in contact with power profiles locally. The local EMS is done by coordinating
subnetworks ISST to follow its recommendations. resources to address fast, unanticipated occurrences, such as
This actor is a moving agent which passes through some regenerated energy from the passing train, surplus energy
subnetworks and is in contact with main agent of each stored in ESSs or requests for more energy for a train which is
subnetwork. delayed. This level of optimization is the link between central-
6) External Consumers (ECs): Workshops, stations, or ized EMS and the RTO in all decentralized subnetwork agents.
any other loads such as EV charging stations Solving the optimization problem for all short-term flexibilities
inside the railway system which are fixed agents in in the massive railway system is unfeasible, while by apply-
developed MAS. ing the decentralized automation architecture with subnetwork
Given that the generic load “railway system” interacts agents provided for local EMS, the short-term (minutes ahead)
with the larger power system (public grid) and its market optimization is achievable.

Authorized licensed use limited to: b-on: Instituto Politecnico de Lisboa. Downloaded on July 25,2024 at 17:59:02 UTC from IEEE Xplore. Restrictions apply.
KHAYYAM et al.: REM-S: CENTRALIZED–DECENTRALIZED AUTOMATION ARCHITECTURE 1167

Fig. 4. MAO UML (a) use case diagram and (b) sequence diagram.

DAO and MAO require power flow calculation. Power for each of them. The HLUCs are general actions or compli-
flow runs in order to calculate the power demand at each ance functionality which are characterized as generic, i.e., as
point of common coupling (PCC). Regarding to massive size describing a general concept and not a specific outcome.
of the railway network, the moving nature of most of the The HLUCs defined for executing day ahead operation are:
loads (trains) at DAO, it is critical to use a simple model of the 1) energy trading; 2) billing; and 3) global optimization.
electrical network. In our electrical network model, the sub- The main objective of energy trading is to buy and sell the
stations (ISST, RSST, and SST) are modeled as ideal voltage energy at the best price for the whole railway network located
sources in series with equivalent impedance. In this model, the in the domain of each IM. The billing calculates the real cost
impedance of the overhead line and catenary are considered in of the consumed energy and the optimization goal is to opti-
series and are modeled by one series impedance representing mize the operational cost and energy/power consumption of
the feeding section. The line is modeled as a  line model. the whole network during next day. This is the high-level
The loads, except the trains, in this representation are mod- optimization which is done centrally.
eled as constant power loads. Trains are modeled in power flow The minutes ahead HLUC is the local optimization which
calculations as current sources. The required information for calculates the optimum 15 min ahead power profile taking into
modeling a train as a moving load consists of its power profile account the reference 24 h power profile. This optimization is
and its position, which are assumed as input data. The effect done locally in each subnetwork.
of train displacement on the electrical model of distribution The real-time HLUCs comprises real-time data acquisi-
system is represented by changing the impedance of feeding tion, estimation, operation control, and actions implementation
sections to correspond to the distance of train to PCCs. In dealing with local subnetwork agents.
MAO, since the power flow should run in restricted area and The real-time data acquisition collects the real-time sta-
only for 15 min, the detail electrical network model is used tus of each subnetwork agent. The estimation aggregates the
by railway electrical network simulator. prediction of consumption/generation of each agent in the
next 15 min, which is needed for MAO. The operation con-
trol generates operational suggestions for each agent. Actions
III. REM-S A NALYSIS P HASE
implementation get operational suggestions from the operation
A. Use Case Analysis control, calculates the optimum way (i.e., real-time actions) to
The starting point for mapping REM-S concept to SG fulfill the suggestions by each agent. The optimization proce-
reference architecture is the use case analysis. Three oper- dures in all HLUCs are executed using the following set of
ational modes defined in the architecture concept based on hard constraints.
different level of optimization and three time scales are con- 1) Each train should reach its destination within a max-
sidered as use case clusters. Based on the defined operation imum window of acceptable delay agreed with the
of each cluster, high-level use cases (HLUC) were identified railway undertaking (RU).

Authorized licensed use limited to: b-on: Instituto Politecnico de Lisboa. Downloaded on July 25,2024 at 17:59:02 UTC from IEEE Xplore. Restrictions apply.
1168 IEEE TRANSACTIONS ON SMART GRID, VOL. 7, NO. 2, MARCH 2016

TABLE I
2) Maximum utilization of internal energy sources U SE C ASE C LUSTER , HLUC, AND P RIMARY U SE C ASE R ELATION
(e.g., renewables installed within the infrastructure) is
achieved.
3) Limits of use of the infrastructure (e.g., maximum power
of a given SST).
The unified modeling language (UML) use case diagrams,
based on the use case clusters and HLUCs definitions are
modeled in the enterprise architect (EA). The use case steps
and information exchange between HLUCs and other actors
for supporting REM-S objectives have been analyzed and are
modeled in EA as UML sequence diagrams. Fig. 4 displays
the UML use case diagram and sequence diagram of MAO as
a sample.

B. Function Layer
Based on the HLUCs, the primary functions are identified
to develop the SGAM function layer. The primary functions
are described in detail with specific information objects as
input and output for implementing use case objectives [9].
The primary functions are detailed enough to be mapped onto
a specific architecture. Table I shows the relation between
use case clusters, HLUCs, and primary functions, which are
modeled in the function layer of SGAM.
In the market zone, the energy trading and energy trading negotiation function, resolves the local lack of power with
estimation functions are defined to connect the railway net- local savings or the surpluses of the neighborhood subnet-
work to the public grid electricity market. These functions are works and vice versa. The other functionalities defined for
in contact with the electricity market in order to forecast the supporting local EMS in the minutes ahead time frame are
next day’s energy price and to buy/sell energy at the best price profile slicing, supervision, and deviation alert. The Deviation
from/to the electricity market. For doing this, energy trading alert_RTO relaunch MAO functions in the case of a major dis-
knows about the energy required by each connection point to ruption that cannot be resolved within the RTO occurs. In real
public grid, the estimated behavior of the market, the con- time, the control function compares the profile calculated by
straints from long-term contracts, the bidding strategies, the the MAO and the real-time status and makes the suggestions
electricity open sessions, and all related costs. for each agent in order to resolve the deviations.
In the enterprise zone, the billing function is considered. The station zone is the real-aggregation level for field level
Since the gathered data in REM-S is integrated from data [9], so aggregation of next 15 min forecasted situation
on-board, wayside, and coordination services, the billing of agents is prepared here. Furthermore in this zone, some
function can calculate the energy consumption for different real-time actions are produced in each agent based on control
railway subsystems and their components, and can conse- commands coming from local intelligence.
quently send its results to the public grid related actors (such In the field zone, the real-time data acquisition function
as utilities or energy suppliers) and railway related actors acquires the real-time status of the trains, DERs, ESS, ECs,
(IM and RU). SSTs, and RSSTs and the consumption measuring function
In the operation zone, several functions are defined for day reads the consumption of each PCC, train, and EC and stores
ahead operation, minutes ahead operation, and RTO modes. it in order to calculate the total consumption. The main point
For DAO, the power consumption/generation forecast of dif- of REM-S real-time data acquisition function or consumption
ferent trains, ECs, DERs, or any other loads will be gathered measuring or estimation function is that they are integrat-
from RU and IM and integrated in the control center of ing the data of different agents (DER, ESS, RSST, SST, and
REM-S. Based on these forecasts, the global optimization will DOEM) in local ISST. These data are used in ISST for local
prepare next day’s power consumption/generation schedules EMS and the required integrated data is sent to control center
of trains, ECs and DERs. The global EMS is supported by for global EMS. In this paper, the function layer model-
functions like audit, report, or mapping scheduling in this ing REM-S was done through the SGAM toolbox EA [13].
zone as well. The Deviation alert_MAO function is defined Fig. 5 displays the function layer of all DAO, MAO, and RTO
for triggering DAO whenever a major disruption that cannot integrated in one SGAM plane.
be resolved within the MAO occurs. In minutes ahead oper-
ation, the local optimization function calculates the optimum C. Business Layer
15 min power profile for the subnetwork according to the daily The business layer indicates which organizations and actors
plan received from control center. These optimum profiles are should participate in pursuing the business objectives (opti-
produced using the local intelligence of each subnetwork. This mizing energy consumption, power demand, and cost). Also
function, along with the power mismatch calculation and the the business processes that support the business objectives and

Authorized licensed use limited to: b-on: Instituto Politecnico de Lisboa. Downloaded on July 25,2024 at 17:59:02 UTC from IEEE Xplore. Restrictions apply.
KHAYYAM et al.: REM-S: CENTRALIZED–DECENTRALIZED AUTOMATION ARCHITECTURE 1169

The estimated price is used as initial energy price in

optimization procedure.
2) For each hour, the REM-S determines the following.
a) What amount of energy is purchased by means of
the long-term agreements.
b) What amount of the energy has to be offered in
each session of the spot market.
3) For each hour covered by each session, the REM-S splits
the energy into blocks of likelihood and assigns them
a price, according to the bidding strategy and to the
contractual agreements available. The result is called the
purchase/sale bid for each market session, which has to
be sent to the EMO.
When a major variation occurs in the planned operation, the
following decisions are taken:
1) determining which sessions of the spot markets can be
used to sell/buy the electricity depending on the hour at
which the variation occurs;
2) repeating the same process followed in the day time
horizon, but restricted to the actual time horizon.
And, once the above tasks are done, the estimated prices
have to be updated.

IV. REM-S A RCHITECTURE

The automation concept, functionalities, and business pro-
cesses of REM-S are defined in the previous section. In this
section, the design of component layer, information layer, and
communication layer essential to this automation concept is
introduced. By identifying these latter three layers, the defined
Fig. 5. DAO, MAO, and RTO function layer in SGAM plane.
functionalities are mapped to the physical architecture [9].

related functions are mentioned. The regulatory constraints are A. Component Layer
taken into account in this layer. The new functionalities defined for managing energy in
The REM-S business functions are energy trading esti- railway operation, should be executed by some components.
mation and energy trading. The energy trading estimation The component layer identifies the components, in the form
function is supported by three processes: 1) market price esti- of system, hardware, software, or interface, to implement the
mation; 2) optimization of energy supply; and 3) calculation intended functionalities, yielding the physical distribution of
of the actual price of energy. The energy trading function all participating components in REM-S architecture.
is supported by five processes: 1) market price estimation; The EBDM, which is the business actor of REM-S inter-
2) optimization of energy supply; 3) calculation of the actual facing with the electricity market, should be supported by
price of energy; 4) detection of electricity market opera- marketplace system and energy trading software.
tor (EMO) open sessions; and 5) construction of the bids. The billing function needs billing software to calculate
Electricity buyer decision maker (EBDM) is the REM-S actor the energy consumption of different components which are
responsible for executing energy trading estimation and energy integrated from different railway subsystems in REM-S.
trading business functions. For global EMS, the EMS/supervisory control and
These two functions are responsible to take the following data acquisition (SCADA) system is required in order to sup-
decisions in the 1–2 days horizon. port operational activities for dispatching energy at higher
1) In order to provide the right financial signals to the level of system in control center. Global optimization soft-
REM-S, an estimated price of the energy must be cal- ware (GOS) supports intelligent functions of REM-S in the
culated as the weighted mean of: a) the price of the control center and makes an optimum plan for the next day. It
energy bought/sold by means of contracts and b) the must have an RU server, an IM server and a DER, EMS,
price of the energy bought/sold in each session of the and VPP system to be responsible for gathering the next
electricity markets. The latter is known only after the day forecasting of timetables, power demands and energy
matching procedure is completed and the prices are generation.
published by the EMO. While these prices are not yet For local EMS, the distribution management sys-
available, a reasonable estimation is calculated by energy tem (DMS)/SCADA supports all operation activities at
trading estimation function according to historical data. each subnetwork to dispatch energy internally or to the

Authorized licensed use limited to: b-on: Instituto Politecnico de Lisboa. Downloaded on July 25,2024 at 17:59:02 UTC from IEEE Xplore. Restrictions apply.
1170 IEEE TRANSACTIONS ON SMART GRID, VOL. 7, NO. 2, MARCH 2016

neighbor subnetwork. The DMS hosts the ISST of each

subnetwork. It takes care of aggregating forecasted profiles
received from all agents (DER controllers, DOEMs, and
ECs). The controlling suggestions are followed by this actor
in ISST. Local optimization software (LOS) located at ISST
supports intelligent functions of REM-S in each subnetwork
area and makes an optimum plan for next 15 min.
Local SCADA is located at SSTs or RSSTs to implement
the actions receiving from ISST. DER controller, intelligent
electronic device, and DOEM have the same role in DERs,
EC components or ESSs, and trains.
Human–machine interface can be used in control center,
ISST, or other SSTs to monitor the energy flow and prepare
an interface for manual applications. The RTU is the interface
object of the control center and ISST components.
The router at control center is needed to realize communi-
cation between the different components (marketplace system,
EMS/SCADA, GOS, etc.) and the subnetworks, public grid,
and electricity market. In ISST, it is used to communicate
to neighborhood subnetworks and realize communication for
data acquisition and sending and receiving control commands.
The measuring instrument device meter measures the energy
consumption (they are already installed in most of railway
system components) and sends them to meter data concentra-
tor (MDC). MDC performs some preliminary analysis of data,
Fig. 6. DAO, MAO, and RTO component layer in SGAM plane.
such as bad data detection and elimination, and then sends
data to meter data management system (MDMS) to gather
all data required to calculate and report necessary informa- 2) Timetable Data: Timetable data to the RUs, related
tion (such as energy bill). Data storage is needed to save the modifications originating from the traffic management
measured data, optimization results, and all other information system (TMS) in real time (i.e., new arrival or departure
which is useful for reporting, supervision, etc. times, or passing time at waypoints).
All new components proposed for REM-S are located 3) Forecasted Trains Power Profile: Each RU must inform
on SGAM plane and modeled with SGAM toolbox [13]. the GOS about the forecasted energy consumption of the
Fig. 6 shows these components in SGAM plane. following day.
B. Information Layer 4) Real-Time Data From DOEM, ESS, ECs, DERs, and
SST/RSSTs: The exchanged data format between LOS
Based on the defined components and their functionalities, and GOS and the end components is also a subject
the information exchange is detailed here. It is modeled with of standardization. While local communication at LOS
UML sequence diagrams with EA to show the chronological level comes from the electricity/power field, new REM-S
sequence of information exchange. The information objects functions need additional standardized data models.
and data models are identified in order to allow an interopera- 5) Estimation for MAO From DOEMs: A consumption pro-
ble information exchange via communication means between file for the minutes ahead calculated by train and sent
components or actors [9]. to the LOS.
For high-level data (object oriented) exchange mecha- 6) Suggestions From Control to DOEMs: LOS opera-
nism, several middleware were analyzed yielding SOAP and tional suggestions and orders (e.g., temporary power
common object request broker architecture as candidates, restrictions, new arrival times, etc.) to all train
consistent with InteGRail project. mounted DOEM.
In REM-S, in some cases the information is exchanged Having in mind that Web services (SOAP) are chosen as
between two actors, in other cases it takes place under the the most appropriate way for transferring data and calling
umbrella of a unique actor. In the first case, there is a clear function remotely, it makes sense to use extensible markup
need for standardization (communication and application pro- language (XML) to define the proposed standard format of
files), while in the second case the standard could remain the aforementioned data types.
as a recommendation. For interoperability, the following In that sense, the adoption of railway modeling lan-
data must be standardized. guage (RailML),1 which is a XML-based solution for simpli-
1) Track Data: A centralized track data server (TDS) dis- fied data exchange between railway applications, is proposed
tributes the track information (the information exchange here. Although RailML covers many aspects of the railways
through the rail line) to the RUs using a standardized
format. 1 http://www.railml.org/

Authorized licensed use limited to: b-on: Instituto Politecnico de Lisboa. Downloaded on July 25,2024 at 17:59:02 UTC from IEEE Xplore. Restrictions apply.
KHAYYAM et al.: REM-S: CENTRALIZED–DECENTRALIZED AUTOMATION ARCHITECTURE 1171

Fig. 7. Detailed REM-S communication layer.

(e.g., infrastructure, vehicles, etc.), the novelty of the REM-S The REM-S communication layer comprises different
concepts requires the extension of its schemas with new tags. networks.
The different schemas of this language are fit for different rail- 1) The train on-board network (green dotted box in
way data. The main data structures defined by RailML are as Fig. 6) follows the IEC 61375 Standard series. In partic-
follows: ular, REM-S relies on this communication for the new
1) timetable that contain schedules; buses based on ECN, as these technologies provide great
2) infrastructure that contains the information about the advantages in terms of flexibility, modularity, cost effec-
route, the line characteristics, stations, etc; tiveness, and reusability. Train subsystems, including the
3) rolling stock that contains the rolling stock DOEM, are connected to it.
characteristics. 2) The IM intranet comprises all components and servers
required to manage the rail network. The relevant actors
C. Communication Layer connected to this intranet are: a) TDS; b) the DER,
For REM-S communication layer, mostly the existing com- EDS, and VPP system (DER EVS); c) the GOS; and
munication profiles (represented as IEC, European Committee d) a number of LOS units.
for Electrotechnical Standardization (CENELEC), or W3 stan- 3) The RU intranet includes several applications like tick-
dards) are used in both the energy and railway fields, although eting, train maintenance, and others. Amongst them, the
in some cases new profiles are needed to cover some function- energy forecaster system is a new component brought
alities. by REM-S.
Each link is analyzed ensuring interoperability to determine 4) The LOS intranet refers to the local area network
whether the profiles require standardization or a recommen- devoted to controlling the power system around an
dation is enough. Fig. 7 proposes the detailed architecture of electrical SST. Different components are connected to
the communication layer. The REM-S communication layer is this network for data exchange: meters, DER, DMS, etc.
based on four main standard families. It provides the pathway to distribute the final order to the
1) IEC 61375 (Train Communication Network): In case of controlled electrical subsystems and to collect low-level
REM-S, four parts are relevant: a) 61375-3-4 [Ethernet electrical parameters.
consist network (ECN)] and b) 61375-2-3,4,6 (commu- In the REM-S communication layer, several ways are
nication and application profiles and train to ground defined for exchanging data between above four internal
communication). networks. The train to ground communication links train
2) IEC 60870-5: Two relevant parts are 60870-5-101 (trans- onboard network and the IM intranet. It is based on the
mission protocols, companion standards especially for IEC 61375-2-6 Standard which contemplates a multitechnol-
basic telecontrol tasks) and 60870-5-104 (transmission ogy (GPRS, 3G, long-term evolution, WiFi, etc.) communica-
protocols, network access). tion with transparent handovers. In Fig. 7, both data exchange
3) IEC 61850: The REM-S related parts of this standard through WiFi hotspots (e.g., at depot, stations, etc.) and
is IEC 61850-6 (configuration language), IEC 61850-7 through the mobile telephony network are represented.
(basic communication structure for substation and feeder According to the architecture defined in IEC 61375-2-6 some
equipment), and 61850-8 (specific communication ser- components like the domain name system server and the
vice mapping). authentication, authorization, and accounting (AAA) server are
4) Simple object access protocol (SOAP). needed.

Authorized licensed use limited to: b-on: Instituto Politecnico de Lisboa. Downloaded on July 25,2024 at 17:59:02 UTC from IEEE Xplore. Restrictions apply.
1172 IEEE TRANSACTIONS ON SMART GRID, VOL. 7, NO. 2, MARCH 2016

The Internet is used to interface third parties from the IM actors. So, 45% of actors are new actors defined particularly
intranet. In particular, communication from and to the RUs and for REM-S.
the electricity market is realized through this common used 3) Integration of Existing Standards: Some standard gap
network. Obviously security measures should be deployed to or some standard modification can emerge in the proposed
avoid intrusions (e.g., firewalls, virtual private network, etc.), architecture which is described in Section IV-C.
where the AAA server can play an important role. When the 4) Hierarchical Level of the Architecture: For this index,
communication uses the telephony network the Internet also we should specify how many nodes or how much nominal
serves as the carrier for the exchanged information. power is managed from control center or from subnetworks.
As the many LOS could be placed in remote areas, differ- For example, one of our scenarios (the line between Paris and
ent means to provide communication to the IM network must Lyon) with 389 km railway is divided to five subnetworks.
be provided. Of course, it may happen that the IM physical There are 384 trains passing this line daily. Hence, each sub-
network is deployed as far as the LOS facilities, but if it were network manager agent is dealing with 389 moving loads and
not the case, the telephony network should be used. When 2–3 EC and 1 or 2 ESS. In this scenario, the control center
the mobile telephony network is used, the same infrastructure is only in contact with five subnetwork managers located in
deployed for the train to ground communication can be reused, five ISSTs.
as shown in Fig. 7.
Currently, the most commonly adopted procedure for
informing the train driver about delays and arrival times, and B. Architecture Robustness to Loss of Communication,
in general for the communication between the TMS and the Communication Delay, and Failure of
driver is the mobile phone. In this case, the driver has to Hardware and Software
manually update the DOEM using the data manager interface. Since in the hybrid architecture the control nodes (sub-
network manager as ISST) are significantly closer to the
V. E VALUATION OF A RCHITECTURE M ODEL loads (train, EC, etc.), the communication delays and loss of
data will decrease, while in the centralized architecture with
The EA UML tool, KPI evaluation, and analysis of abnor-
large data loads over long distances between control nodes
mal conditions are applied for assessing the architecture.
and loads, channel congestion is more likely to happen and
EA is used to check the interoperability of the layer con-
the number of bottlenecks may rapidly increase, leading to
nections between components and functions, communication
a lower level of architecture robustness. On the other hand
protocols, data models, and the feasibility of all use cases,
by defining the functions implemented in each device based
which confirms whether or not the architecture supports all
on the SGAM framework in EA, the effect of failure of such
the use cases.
devices is studied. The architecture passes “consistency check”
Some KPIs can be defined for the REM-S architecture, to
to ensure reliability by defining all types of information objects
conceptually address the following points.
needed for supporting functions in degraded operation sce-
narios along with necessary incremental reinforcements of
A. Characterization of Key Features a proposed infrastructure.
Quantifiable metrics are defined to compare key features of
REM-S architecture to other architectures [14].
1) Ratio Between Number of Actors Shared Between Use C. Increase in Production or Consumption
Cases and Total Number of Actors: This ratio is intended to Hosting Capacity
represent the effort of synthesis of actors. That means if the The hosting capacity is defined as the maximum amount of
actors work separately, the amount of their analysis effort and new production (e.g., new DERs in case of REM-S) or con-
its complexity will decrease. There are six business actors in sumption (e.g., new trains or new ECs in case of REM-S)
three different use cases (IM, railway operator, transmission that can be connected without endangering the reliability
system operator/distribution system operator, EMO, energy or quality for other customers [15]. Different performance
supplier, and grid owner). Because four of them (IM, railway indices, that are mostly related to power quality indices, can
operator, energy supplier, and grid owner) are shared in the use be selected for evaluating the hosting capacity of the network.
case, the ratio is 67%. For the centralized architecture, shared As an example, for phenomena like network overloading by
use cases would be similar, so the ratio remains 67%. This wind power penetration, a performance index like maximum
ratio for SGAM actors (the new actors defined particularly hourly value of current through related transformer or the
for REM-S) becomes 0% for REM-S architecture, because no maximum power flow of network [15] can be considered. In
shared actors are defined among the use cases. This ratio for REM-S architecture, these performance indices are defined
the centralized architecture will be 20%. The SGAM actors as objective functions in DAO and MAO. In both optimiza-
are forecast provider, GOS, EBDM, LOS, and DMS. In case tion procedures, we consider hourly and real-time power
of the centralized architecture, LOS will be the shared actor. demand optimization as one of the main goals and the voltage
2) Ratio Between Number of New Actors and Total Number and current standard ranges as hard constraints on the opti-
of Actors: This ratio represents the innovation brought by mization procedure; therefore the optimal solution will take
the new architecture. There are six business actors which are care of phenomena such as overload, overvoltage, or under
the stakeholders of REM-S and five new actors as SGAM voltage.

Authorized licensed use limited to: b-on: Instituto Politecnico de Lisboa. Downloaded on July 25,2024 at 17:59:02 UTC from IEEE Xplore. Restrictions apply.
KHAYYAM et al.: REM-S: CENTRALIZED–DECENTRALIZED AUTOMATION ARCHITECTURE 1173

D. Cost of Architecture node (in REM-S, from field nodes to control center through
The architecture cost is a summation of deployment cost subnetwork nodes in station zone). Tj is calculated as
and operation cost [16]. In this formulation, the installation 8lmF→S
cost is not considered as a deployment cost and it is assumed Tj =
F→S
. (6)
t
that the required hardware and software devices are existing
and that we want to check the difference between apply- Here t is the time period for transmitting information form
ing different energy management architectures to the system a field zone node to a station zone node and lm shows the
under consideration. Therefore, the deployment cost con- length of the message [16]. For calculating the total com-
sists of the costs of deploying data storage, processing, and munication cost for the whole architecture it is necessary to
communication capacity calculate both cost of transmitting data from field zone nodes
to station zone nodes and also from there to the operation
CD = CS + CC + CP . (1) zone node (control center). Hence, we can briefly formulate
this relation as
Here CD is the deployment cost, CS is the storage cost;  
CC = CC (i) + CC ( j). (7)
CP is the cost of processing units and CC is the cost of
i∈N j∈M
transceivers/receivers [16].
The operation cost is defined as the amount of energy According to (6) and (7), since the cost scales directly
consumed in a fixed time period (e.g., one month) with the length of the message (which is significantly longer
in centralized architecture), and as N, the number of sub-
CO = Etotal × fE . (2) networks in the station zone, is much smaller than M, the
number of nodes at field zone, (N << M), the communica-
Here, Etotal is the average energy required by all nodes tion cost in the hybrid architecture is significantly less than
for operating during the time interval and fE is the price of in centralized architecture. The conclusion is in line with the
energy [16]. pilot experiment [15] on the comparison of the communication
1) Storage Cost Calculation: The cost of storing data at bandwidth for different architectures.
node k is calculated as follows: 3) Processing Cost Calculation: The processing cost at
node k for each query is calculated as
CS (k) = Sk × fS (Sk ). (3)
CP (k) = nm × n × fp (k). (8)
Here fS models the price of storage and Sk is the total
amount of storage capacity, given by Here nm is number of messages, n is the number of oper-
ations required for each query and fp (k) is the function that
lm models the processing price [16]. According to [16] simula-
Sk = T × . (4)
τ tion results, the processing operations are significantly fewer
Here T is the time duration, τ is the sampling interval, in hybrid architecture compared to centralized architecture.
and lm shows length of message [16]. The cost storage for This difference can be interpreted by the difference between
all nodes can be calculated by (4). This means that to calcu- distributing the process effort between the field, station and
late the storage cost of whole architecture we should add the operation zone nodes compared to doing most of the processes
cost of the field zone nodes (meters), station zone nodes (sub- centrally in operation zone.
network nodes), and operation zone nodes (control center). 4) Total Energy Cost Calculation: The energy required at
Equation (4) shows that the storage cost has a direct rela- each node for operation is formulated by
tion to distance; this results more expensive storage cost in j→k j
Ej = EC + Er/w + Epj . (9)
the centralized architecture than in the hybrid architecture due
j→k j
i , and E are formulated below
to the greater distance between field zone nodes and control The EC , Er/w p
center. On the other hand, if we consider N as the number of
j→k
subnetworks and M as the number of nodes in the field zone, EC = esc lsc + ercj→k lrcj→k
j→k j→k
(10)
N should be much smaller than M(N << M); therefore adding j
Er/w = erj lrj + ewj lwj (11)
N nodes for developing the hybrid architecture has no signif-
icant effect on the cost of whole architecture. All in all, as it Epj = epj npj . (12)
is also proved by simulation in [15], the storage cost in the j→k
In the formulation above: EC is the energy required to
centralized architecture is more expensive than in the hybrid
send/receive information between j and k nodes according
architecture. j→k j→k j
to the length of transmitted message (lsc or lrc ), Er/w
2) Communication Cost Calculation: The communication
is the required energy for reading or writing information
cost for each node is calculated as follows:
from/to storage according to the length of the message that
j j j
CC ( j) = Tj × fC (Tj ). (5) has been read/written (lr or lw ), Ep is the energy consump-
tion of processing and is calculated by the energy required
j
Here fc models the price of bandwidth and Tj is for processing a byte of information (ep ) and the number of
j
data rate needed to transmit data from one node the other processed bytes (np ) [16].

Authorized licensed use limited to: b-on: Instituto Politecnico de Lisboa. Downloaded on July 25,2024 at 17:59:02 UTC from IEEE Xplore. Restrictions apply.
1174 IEEE TRANSACTIONS ON SMART GRID, VOL. 7, NO. 2, MARCH 2016

The total energy cost is calculated by summing the energy VI. C ONCLUSION
required for all nodes from the field zone to the operation zone. The new REM-S architecture is presented. Its development
It can be seen from (10)–(12) that the consumed energy has demonstrates achievements in three main domains: the map-
a direct relation to the message length and to the number of ping of the railway system onto the SG concept, the first step in
processed bytes and, because N << M, the cost of energy in the direction of harmonization of standards of SG and railway
the centralized architecture is much bigger than in the hybrid systems in the area of energy management, and the mapping
architecture. This conclusion is illustrated by simulation results and development of the new architecture for railway systems
in [16] as well. onto the reference architecture of the SG.
All in all, it can be concluded that in the calculation of The SG concepts in railway systems span the centralized–
architecture cost, the larger the number of nodes in the field decentralized automation architecture, the adoption of different
zone the more suitable is the hybrid architecture compared to time horizons (day ahead, minutes ahead, and real time) and
the centralized architecture. the creation of flexibility with DOEM, DER, EC, and ESS.
The adoption of the SGAM framework yields the interop-
E. Scalability erability of different layers and the interoperability with the
In this section, another performance metric is defined to rest of the SG system.
evaluate the hybrid architecture communication cost according The standardization analysis identifies which railway or SG
to traffic rate, bandwidth size, number of nodes, and distance standards are applicable in REM-S and which parts need
between nodes, showing how the scalability of architecture extension. In support of this, the authors plan to submit
would be improved by switching from a centralized architec- the REM-S use cases to CEN-CENELEC-ETSI and upload
ture to hybrid architecture. The total cost of communication them in the use case management repository. Some recom-
in a traditional centralized architecture is calculated as in (13), mendations are identified for TECREC especially in the field
and in hybrid architecture as of rolling stock with the ground communication and energy
related parts of data modeling.
TotalCostC = β λ̄D̄M + F0 (13)
⎛ ⎞
 2   13
 2 β 3 γβ D̄ + F̄ ⎠
TotalCostH = λ̄D̄M 3 ⎝γ D̄ + R EFERENCES
γβ D̄ + F̄ β [1] Europe 2020: Commission Proposes New Economic Strategy, Eur.
Comm., Brussels, Belgium, Mar. 2010.
(14)
[2] EU Energy Policy to 2050, Eur. Wind Energy Assoc., Brussels, Belgium,
Mar. 2011.
where λ̄ is the average traffic rate on field zone nodes, D̄ is the [3] Railway Handbook: Energy Consumption and CO2 Emissions, Int.
average distance between field zone nodes and the subnetwork Energy Agency Int. Union Railways, Paris, France, 2012.
manager, M is the total number of nodes in the field zone, β is [4] P. Chavali, P. Yang, and A. Nehorai, “A distributed algorithm of appli-
ance scheduling for home energy management system,” IEEE Trans.
unit cost of bandwidth distance product, γ is the bandwidth Smart Grid, vol. 5, no. 1, pp. 282–290, Jan. 2014.
needed for the information exchanged between a distributed [5] Z. Zhao, W. C. Lee, Y. Shin, and K. Song, “An optimal power
server and a centralized server, Fj is the cost of deploying scheduling method for demand response in home energy manage-
MDMS at location j and F̄ is the average deployment cost of ment system,” IEEE Trans. Smart Grid, vol. 4, no. 3, pp. 1391–1400,
Sep. 2013.
distributed MDMS [17]. [6] G. T. Costanzo, G. Zhu, M. F. Anjos, and G. Savard, “A system
Equation (13) shows that the total cost in centralized archi- architecture for autonomous demand side load management in smart
tecture scales linearly with the number of nodes, the average buildings,” IEEE Trans. Smart Grid, vol. 3, no. 4, pp. 2157–2165,
Dec. 2012.
data generation rate of loads, and the average distance between [7] S. Salina, M. Li, P. Li, and Y. Fu, “Dynamic energy management for
the loads and control center. In (14), the cost scales as x2/3 smart grid with distributed energy resources,” IEEE Trans. Smart Grid,
with the similar parameters. These two equations show that vol. 4, no. 4, pp. 2139–2151, Dec. 2013.
[8] J. Liu, X. Li, D. Liu, H. Liu, and P. Mao, “Study on data manage-
by increasing the traffic rate to nodes or the number of nodes, ment of fundamental model in control center for smart grid operation,”
the communication cost increases more rapidly in the central- IEEE Trans. Smart Grid, vol. 2, no. 4, pp. 573–579, Dec. 2011.
ized architecture compared to the hybrid architecture, which [9] CEN-CENELEC-ETSI: Smart Grid Reference Architecture, ETSI
Standard SG-CG/M490/C, 2012.
implies that the hybrid architecture is more scalable than the [10] CEN-CENELEC-ETSI: Smart Grid First Set of Standards, ETSI
centralized architecture. Standard SG-CG/M490/B, 2012.
[11] CEN-CENELEC-ETSI: Smart Grid Use Case Management Process,
ETSI Standard SG-CG/M490/E, 2012.
F. Analysis of Abnormal Conditions [12] S. D. J. McArthur et al., “Multi-agent systems for power engineer-
The main structure of REM-S architecture is developed ing applications—Part I: Concepts, approaches, and technical chal-
lenges,” IEEE Trans. Power Syst., vol. 22, no. 4, pp. 1743–1752,
based on normal operation condition and then it is modified to Nov. 2007.
ensure that it can support the following abnormal conditions. [13] SGAM Toolbox. (May 2, 2004). [Online]. Available:
1) The SST or energy storage fails. http://www.entrust.at/downloads/sgamtoolbox/ accessed on: 2/5/2014
[14] “Distribution automation concept,” RWTH Aachen Univ., Aachen,
2) A delayed train tries to catch up on its timetable. Germany, Tech. Rep. IDEAL Grid for All FP7 Project Deliverable D3.1,
3) Communication with some agents is down. 2014. [Online]. Available: http://ide4l.eu/results/
4) Temporary or Permanent Timetable is updated. [15] N. Etherden, “Increasing the hosting capacity of distributed
energy resources using storage and communication,” Ph.D. dis-
5) Temporary or Permanent speed limits are updated. sertation, Dept. Eng. Sci. Math., Luleå Univ. Technol., Luleå,
6) DERs generate more or less energy than forecasted. Sweden, 2012.

Authorized licensed use limited to: b-on: Instituto Politecnico de Lisboa. Downloaded on July 25,2024 at 17:59:02 UTC from IEEE Xplore. Restrictions apply.
KHAYYAM et al.: REM-S: CENTRALIZED–DECENTRALIZED AUTOMATION ARCHITECTURE 1175

[16] A. U. N. S. Narashiman, M. Vasirani, R. V. Prasad, and K. Aberer, Valerio Recagno received the Master degree in
“A cost-benefit analysis of data processing architectures for the smart electronics engineering in 1992 and the Ph.D.
grids,” in Proc. ACM Int. Workshop Wireless Mobile Technol. Smart degree in transport engineering and economics
Cities (WiMobCity), Philadelphia, PA, USA, 2014, pp. 91–100. from the University of Genoa, Genoa, Italy,
[17] J. Zhou, R. Q. Hu, and Y. Qian, “Scalable distributed communica- in 1997.
tion architectures to support advanced metering infrastructure in smart He has experiences as a Consultant and
grid,” IEEE Trans. Parallel Distrib. Syst., vol. 23, no. 9, pp. 1632–1642, a University Lecturer. In 2003, he joined
Sep. 2012. D’Appolonia Company, Genoa, where he is cur-
rently the Development Manager with the Transport
Engineering Division.
Mr. Recagno has been a member of the
Transportation Energy Committee of the Transportation Research Board of
Sara Khayyam (S’14) received the Master’s degree National Academy, USA, since 2012.
in electrical engineering from the Iran University of
Science and Technology, Tehran, Iran, in 2007. She
is currently pursuing the Ph.D. degree with RWTH
Aachen University, Aachen, Germany.
She was a Research Assistant in the field of
distribution networks design with Niroo Research
Institute, Tehran, for five years. Since 2012, she
has been a Research Associate with the Institute
for Automation of Complex Power Systems, E.ON
Energy Research Center, RWTH Aachen University.
Her current research interests include smart grids and energy management
systems.
Valeria Bagliano received the Master degree in
computer science engineering from the University
of Genoa, Genoa, Italy, in 2006.
In 2006, she joined D’Appolonia Company,
Ferdinanda Ponci (M’00–SM’08) received the Genoa, where she is currently a Senior Project
M.S. and Ph.D. degrees in electrical engineering Engineer with the Transport Engineering Division.
from Politecnico di Milano, Milan, Italy, in 1998 Since 2011, she has been involved in several projects
and 2002, respectively. related to energy efficiency in the transport sector,
In 2003, she joined the Department of Electrical especially in the railway domain.
Engineering, University of South Carolina,
Columbia, SC, USA, as an Assistant Professor
with the Power and Energy Research Group, where
she was tenured and promoted as an Associate
Professor in 2008. In 2009, she joined the Institute
for Automation of Complex Power Systems, E.ON
Research Center, RWTH Aachen University, Aachen, Germany, where she is
an Professor.

Javier Goikoetxea received the Engineering degree

from the University of Navarre, Pamplona, Spain, Antonello Monti (M’94–SM’02) received the M.Sc.
in 1997. (summa cum laude) and Ph.D. degrees in electrical
He was with the Department of Research and engineering from Politecnico di Milano, Milan, Italy,
Development, CAF Company, Beasain, Spain, where in 1989 and 1994, respectively.
he was involved in the most relevant technolog- He started his career in Ansaldo Industria, Milan,
ical developments of the company, such as the and then moved to Politecnico di Milano, as an
Control and Supervision Modular System Train Assistant Professor in 1995. In 2000, he joined
Control and Monitoring System (COSMOS TCMS), the Department of Electrical Engineering, University
the Intelligent Tilting Integrated System (TCMS), of South Carolina, Columbia, SC, USA, as an
and the OARIS high-speed train. He is currently the Associate and then a Full Professor. Since 2008, he
Director of Research and Development Projects (at both international and has been the Director of the Institute for Automation
national levels), and the coordination of technology supporting CAF’s Chief of Complex Power System, E.ON Energy Research Center, RWTH Aachen
Technology Officer. He was a Technical Coordinator of MERLIN, Brussels, University, Aachen, Germany. He has authored or co-authored over 200 peer-
Belgium. In the last months, he has been deeply involved in the preparation reviewed papers published in international journals and in the proceedings of
of the SHIFT2RAIL initiative and he will be the Technical Coordinator of its international conferences.
lighthouse project ROLL2RAIL, which starts in 2015. Dr. Monti is an Associate Editor of the IEEE S YSTEM J OURNAL.

Authorized licensed use limited to: b-on: Instituto Politecnico de Lisboa. Downloaded on July 25,2024 at 17:59:02 UTC from IEEE Xplore. Restrictions apply.