An Efﬁcient Simulator for Fault Detection and Recovery in Smart Grids

FDIRSY

Syrine Ben Meskina

, Narjes Doggaz

and Mohamed Khalgui

LISI, ISI, University Tunis-El Manar, Tunis, Tunisia

LIPAH, FST, University Tunis-El Manar, Tunis, Tunisia

LISI, INSAT, University of Carthage, Tunis, Tunisia

Keywords:

Smart Grid, Fault Categorization, Dominant Faults, Equivalent Faults, Local Resolution, Non-local Resolu-

tion, Multi-agent System.

Abstract:

This research paper deals with failures and faults in power smart grids. We propose an original multi-agent

approach for power system recovery based on fault classiﬁcation. For that, we propose the classiﬁcation

of faults as dominant or equivalent ones. This classiﬁcation has the advantage of optimizing the task of

power system recovery. To test and validate our approach, we develop a simulator, named FDIRSY (Fault

Detection, Isolation and Recovery SYstem). The experimental study showed that our approach ensures the

search for the best solution from the existing ones thanks to the use of mobile agents. These agents have

the advantage of evaluating all the existing alternatives while reducing the communication cost (in terms of

exchanged messages). We demonstrate that our approach is gainful in terms of required times, actions to be

performed as well as the faults to be resolved thanks to the proposed fault classiﬁcation.

1 INTRODUCTION

The new generation of power networks, ”smart grid”,

makes power networks intelligent through the main

following characteristics; multi-service communica-

tion, reliability, security and safety allowing a real-

time supervision. It should integrate the actions of

all types of users and consumers increasing the im-

portance of low voltage network automation. (Fang

et al., 2012) discusses the several actors and factors

contributing on the evolution of power grids in their

survey. Many types of problems may occur on power

grids engendering other faults by propagation. There

are many related works dealing with power recovery

systems. But the existing methods do not handle the

consequent failures and do not investigate the relation

between them. Each one of the detected failures is

handled separately.

In this paper, we are interested in developing an

original software system detecting the faults in power

networks, identifying, localizing and investigating the

occurred faults as well as the consequent ones. This

new simulator is based on the multi-agent approach

for power system recovery proposed in (Ben Meskina

et al., 2014). The decentralization of the control (one

controller agent per sub-grid) restricts the number of

the deployed agents in order to avoid the expensive

inter-agent communication. The use of mobile agents

in power grid present a novelty and allows an acceler-

ated investigation of all the connected neighbor sub-

grids to provide efﬁcient recovery. The number of

faults to be resolved is reduced thanks to the proposed

fault categorization. The use of extensible data bases

updated at run-time is advantageous in order to re-use

solutions yet found. In this paper, Section 2 summa-

rizes the most important works related to smart grid

recovery in the literature. In Section 3, we expose the

problems which are not considered in the literature

and present the characteristics and architecture of the

developed simulator FDIRSY. Then, in Section 4 we

detail its functioning and give the principal algorithms

of the several agents deployed in the multi agent sys-

tem(MAS). The last section illustrates the gain, using

our simulator, in terms of required time and commu-

nication.

2 STATE OF THE ART

There is a big amount of research works, in the lit-

erature, relative to several topics about smart power

grids. An important number of them work on failure

132

Ben Meskina S., Doggaz N. and Khalgui M..

An Efﬁcient Simulator for Fault Detection and Recovery in Smart Grids - FDIRSY.

DOI: 10.5220/0005245001320139

In Proceedings of the 5th International Conference on Pervasive and Embedded Computing and Communication Systems (PECCS-2015), pages

132-139

ISBN: 978-989-758-084-0

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

detection and localization, self-healing and system re-

covery... These related studies investigate these prob-

lems and propose several approaches based on differ-

ent concepts such as; Multi-Agent Systems (MAS),

Petri-Nets, smart micro-grids using Distributed En-

ergy Resources (DER), electric devices like Smart

Storage Devices (SSD)... (Chertkov et al., 2011) de-

velops an approach to efﬁciently identify the most

probable failure modes in static load distribution for

a given power network. (Calderaro et al., 2011) pro-

poses a method using Petri Nets in order to detect and

identify failures in transmission power grids. (Rus-

sell and Benner, 2010) presents some examples of the

types of incipient failures that can be detected from

substation electrical waveforms. (Oudalova and Fidi-

gattib, 2009) uses the concept of micro-grids using re-

newable energy, SSD and load controllers in order to

reduce the transmission loads and to resolve failures.

(McArthur et al., 2007a) and (McArthur et al., 2007b)

investigate and study the higher value-added by MAS

to power industry (using open MAS architectures and

distributed system platforms). (Jiang et al., 2014b)

propose a hierarchical MAS for the self-healing based

on IEC 61499/61850. (Jiang et al., 2014a) propose

an hierarchical multi-agent architecture for automatic

restoration in power grids. They associate an intelli-

gent agent to each electrical station in order to con-

trol its functioning. (Massoud and Wollenberg, 2005)

presents a modern infrastructure for power grids im-

proving their efﬁciency, reliability, and safety and

making them intelligent by, simply, integrating a soft-

ware module in some electric components. The goal

is to decentralize the control without waiting the re-

sponse of the central protection system in the case of

power network failures.

After these short observations on the literature, we

count a multitude of research works working on fail-

ure detection and power system recovery. In fact, the

majority of them handle only one type of faults. We

remark, also, that the related works do not investigate

the relation between the detected fault and the con-

sequent ones to resolve both of them. It is important

to establish a relation between the detected faults and

the engendered ones in order to deﬁne an optimizing

strategy solving the maximum of the occurred faults

or, even, all of them. In the other hand, we note that

the majority of the proposed approaches, in the liter-

ature, are based on MAS deploying a big number of

agents ((Ramchurn et al., 2011) assigns one agent to

each consumer). In fact, this makes the process of

inter-agent communication very expensive in terms

of required time and, thus, the procedure of search

for solutions performs too slowly. (Rahman et al.,

2007) proposes an efﬁcient approach for power sys-

tems self-healing based on specialized micro-grids.

They use 7 agent types making the communication

process costly. Another disadvantage of the proposed

MAS in the literature lies in the centralized control.

(Vyatkin et al., 2010) deploys only a unique one mas-

ter agent Facilitator Agent (FAG) and multiple Bus

Agents (BAGs) leading to a centralized control power

system and to a big number of BAGs. The investi-

gation of the history of the encountered and solved

problems in power grids is important to avoid the res-

olution of problems solved beforehand. (Pipattana-

somporn et al., 2009) used an inextensible fact bases.

3 SMART GRID SIMULATOR

In this section, we begin at ﬁrst by exposing the rea-

sons motivating the development of FDIRSY. Second,

we detail our multi-agent approach for fault detection

and recovery. Finally, we present, FDIRSY, our de-

veloped simulator.

3.1 Problems and Motivations

The robustness of electrical grids lies on their abil-

ity of managing and facing all eventual submerging

problems. We model a problem in a power network

by a fault or a fault set. In fact, a fault occurs be-

cause of violating one or more constraints knowing

that such fault may engender other consequent ones.

We proposed, in (Ben Meskina et al., 2014), a set of

operating conditions that must be respected in smart

power grids. In order to optimize the required time

for power system recovery, we look for optimizing

the cost of the resolution procedure while trying to

solve the maximum of the occurred faults. For that,

we proposed new deﬁnitions for dominant and equiv-

alent faults in (Ben Meskina et al., 2014). Dominant

faults correspond to faults engendering other ones in

the connected components (belonging to the voltage

level below or the same one with smaller rank). The

rank presents the order of a component apparition in

a power line. Equivalent faults correspond to faults

occurring on connected components belonging to the

same voltage level and having the same rank. These

new relations facilitate the failure recovery and al-

low the control and the reduction of the fault recovery

time.

In order to illustrate our fault categorization, we

consider, in Figure 1, a power sub-grid. Let us assume

that the emergency Medium Voltage Line eMV L1 is

activated and that the Medium Voltage Line MV L1 is

deactivated. Let us suppose, in addition, that there

is an instability voltage problem observed on MV L2.

AnEfficientSimulatorforFaultDetectionandRecoveryinSmartGrids-FDIRSY

133

We denote by F

the resulting fault occurred on the

Low Voltage Transformer LV T 1 which is propagated

to the connected devices. We denote, respectively, by

, F

and F

the faults, consequently, occurred

on the Low Consumers LC1, LC5, LC8, and on the

Medium Consumer MC1. Based on the new pro-

posed deﬁnitions, we categorize F

, F

and F

as faults

dominated by F

and both of F

and F

as equivalent

ones. The resolution of the dominant fault resolves all

the problem (including the dominated ones). When a

dominant fault can not be resolved, we proceed to the

resolution of the dominated faults after categorizing

them. On the other hand, the resolution of only one of

the equivalent faults can resolve all the problem. The

advantage is, particularly, observed in the case of mul-

tiple occurring faults, as we are focussed on resolving

only one or, at most, a subset of them. Thus, in this

example, the resolution of either F1 or F5 solves all

the problem. This strategy facilitates, accelerates and

guides pertinently the procedure of search for solu-

tion.

Figure 1: Fault F

engendering other faults (F

, F

and

) on the connected components

3.2 The Proposed Architecture

To achieve the development of our system, we use the

multi-agent paradigm based on distributed architec-

ture decentralizing the control hence the decomposi-

tion of the overall electric network into sub-nets. This

architecture is, essentially, composed of three types

of software agents which are Reconﬁguration Agent,

Mobile Agent and Data Base Agent. These agents

interact and collaborate together in order to maintain

the stability and the effective functioning of the power

grid. Our system detects and classiﬁes the occurred

faults to facilitate the system recovery and looks for

ﬁnding the best solution as we use a solution updated

at run-time contrary to the related works. The UML

activities diagram, illustrated by Figure 2, describes

the behavior of our simulator. This schema indicates

the used software entities composing our system and

summarizes the main considered functionalities.

• Reconﬁguration Agent (RAgent)

A Reconﬁguration Agent is responsible of the detec-

tion and the recovery of the faults occurred in the sub-

grid under its scope. We propose to associate only

one RAgent to each power smart sub-grid. When

it detects an anomaly due to a violated constraint,

it isolates ﬁrst the component or line responsible of

the anomaly. Then, it searches locally for a solution.

When no local solution is found, it requests the Data

Base Agent about a solution already found for the en-

countered problem. If no stored solution, the Recon-

ﬁguration Agent searches for a cooperative solution

from the neighbor sub-grids. For that, it creates a Mo-

bile Agent(s) to obtain information about the compo-

nents belonging to the connected sub-grids.

• Mobile Agent (MAgent)

A Mobile Agent is a software entity moving dynam-

ically, through electric lines, to visit the components

belonging to the neighbor smart sub-grids. It is cre-

ated and destructed by an RAgent. Its task consists

on collecting and communicating useful information

for a given search for fault recovery. When the vis-

ited component has more than one outgoing line (con-

nected to more than one components), the MAgent

is duplicated in order to analyze all the existing al-

ternatives or paths. The duplicated agents are called

clones and takes only one line (path) at a time. All

the clones operate simultaneously in order to decrease

the required time. Each component should be visited

at most once by the same Mobile Agent to ensure that

the resolution terminates.

• Data Base Agent (DBAgent)

The DBAgent is responsible of the management and

the storage of all the information relative to the whole

power grid structure. It contains, also, the history of

all the occurred faults and all information relative to

the problems (like problem type, failed components,

CPU...) as well as the corresponding solutions found

by MAgents. It may be requested by both Recon-

ﬁguration and Mobile Agents about information rel-

ative to the power system. It should update its knowl-

edge for each new solution found by Mobile Agents.

To avoid searching solutions for problems already re-

solved, the Reconﬁguration Agent may request the

Data Base Agent when it encounters a problem for

which it did not found a local solution.

PECCS2015-5thInternationalConferenceonPervasiveandEmbeddedComputingandCommunicationSystems

134

Figure 2: Activities diagram relative to the proposed MAS.

3.3 FDIRSY

To test, validate and evaluate our approach, we have

developed in Java, FDIRSY, a Fault Detection, Iso-

lation and Recovery SYstem. It provides the fol-

lowing services: (i) design and simulation of smart

grids composed of electrical components (genera-

tors, transformers, consumers) and lines (principal

and emergency ones) belonging to the high, medium

and low voltage levels, (ii) checking operating condi-

tions to detect, localize and identify problems, even-

tually, occurred in the simulated smart grid, (iii) iso-

lation of the detected problem to avoid propagation,

(iv) classiﬁcation of the occurred faults relative to the

encountered problem and (v) searching for solution to

recover the failed power grid.

4 IMPLEMENTATION

In this section, we describe the functioning of the pro-

posed simulator FDIRSY. We begin by detailing and

giving some algorithms relative to the tasks ensured

by RAgents followed by those ensured by MAgents.

4.1 RAgent Implementation

An RAgent is responsible of supervising the proper

functioning of the sub-grid under its scope. Let RA

the RAgent supervising the i

smart sub-grid denoted

by SSG

. RA

ensures the following tasks:

• Check Operating Constraints

FDIRSY allows RA

to investigate the healthy opera-

tion of SSG

. In fact, RA

checks the operating condi-

tions over SSG

; the activation, stability and ﬂowing

load constraints. Algorithm 1 provides the set of the

violated constraints (VC(SSG

)), below, the descrip-

tion of the several used functions:

- A(X): returns the activation state of an electrical

component or line X as a boolean value (true if ac-

tivated and false if deactivated),

- Freq(C): returns the frequency of an electrical com-

ponent C,

- Volt(L): returns the voltage value of an electrical line

AnEfficientSimulatorforFaultDetectionandRecoveryinSmartGrids-FDIRSY

135

- ReqL(C) and ReceivL(C): returns, respectively, the

required and received loads by a consumer C.

Algorithm 1: Checking Operating Constraints.

Require: SSG

Ensure: VC(SSG

)

VC(SSG

) ←

if (∃C ∈ SSG

\A(C) = f alse or Freq(C) 6=

De f aultFreq or C does not respect its capacities

or ReqL(C) < ReceivL(C)) then

add violated constraint(C) to VC(SSG

)

if (∃pL ∈ SSG

\A(L) = f alse or Volt(L) 6=

De f aultVolt or L does not respect its capacities)

then

add violated constraint(pL) to VC(SSG

)

• Identify Problem

When there is, even at least, one violated constraint

(VC(SSG

) 6=

0), RA

proceeds, in 2, to identify the

encountered problem ep(VC(SSG

)). In fact, it is

an important step to localize the failed components

( f cL(SSG

)) and lines ( f lL(SSG

)) in order to guide

the recovery procedure since each problem type is re-

solved differently. Moreover, it is a useful step to up-

date the data base. We note that each failed compo-

nent corresponds to a fault. It should be noted that

several faults can occur simultaneously following to

more than one violated constraint (one problem).

Algorithm 2: Problem Identiﬁcation.

Require: VC(SSG

)

Ensure: f cL(SSG

), f lL(SSG

), ep(VC(SSG

))

f cL(SSG

) ←

0, f lL(SSG

) ←

for each vc ∈ VC(SSG

)

switch (constraint type of vc)

case activation:

ep(VC(SSG

)) ← deactivation problem

case stability:

ep(VC(SSG

)) ← instability problem

case ﬂowing load:

ep(VC(SSG

)) ← problem on ﬂowing loads

end switch

update( f cL(SSG

), ep(VC(SSG

)))

update( f lL(SSG

), ep(VC(SSG

)))

end for

• Isolate Failed Components / Lines

Once the encountered problem is identiﬁed, RA

should isolate the failed components and lines

( f cL(SSG

) and f lL(SSG

)) -by deactivating them- in

order to avoid the failure propagation to the non-failed

connected components.

• Classify Faults

We detail, in Algorithm 3 how, RA

classiﬁes

the identiﬁed faults by dominance or by equiva-

lence ( f t( f cL(SSG

))). We note that the function

VoltLevel(X) returns the voltage level of an electric

component X (1, 2 or 3 for, respectively, the high,

medium or low voltage levels).

Algorithm 3: Fault Classiﬁcation.

Require: f cL(SSG

)

Ensure: f t f cL(SSG

)

min ← minimum VoltLevel( f cL(SSG

))

min

← {c ∈ f cL/VoltLevel(c) = min}

if (|C

min

| = 1) then

f t( f cL(SSG

)) ← dominance

dominant fault(( f cL(SSG

)) ← C

min

dominated faults(( f cL(SSG

)) ← ( f cL(SSG

) \

min

elsif (|C

min

| > 1) then

f t( f cL(SSG

)) ← equivalence

equivalent faults ( f cL(SSG

) ← C

min

• Search Local Solution

According to the identiﬁed type of the occurred faults

( f t( f cL(SSG

))), RA

is guided to resolve pertinently

the encountered problem based on the strategy of fault

management described in Section 3.1. In order to

optimize the search for solution as well as the net-

work route, RA

follows a ﬂexible strategy. It begins

by searching a local solution that consists of a de-

activated local emergency line (∈ SSG

). This line

should procure sufﬁcient load to supply the compo-

nent(s) outputted by Algorithm 3. If there exist more

than one solution, RA

choose the one providing the

bigger Remaining Load (RL = ReceivLoad(x) - Re-

qLoad(x)).

• Search Non-local Solution

When no local solution is found, RA

begins by look-

ing for a solution previously stored on the data base

by requesting the DBAgent about the encountered

problem relative to SSG

. In fact, our system may

have encountered and solved this problem before-

hand. If there is no stored solution, RA

looks for new

non-local solution from the other connected sub-grids

through MAgent(s).

4.2 MAgent Implementation: Non-local

Solution Search

In case of neither local solution nor stored solution,

creates a MAgent MA

on the failed component

C. MA

visits all the connected devices to C through

PECCS2015-5thInternationalConferenceonPervasiveandEmbeddedComputingandCommunicationSystems

136

the existing outgoing lines called paths in order to

collect useful information about; (i) Cumulative Re-

maining Load (CRL), (ii) Cumulative Priority (CPr)

and (iii) End of the taken Path (EoP is equal to true

if there is no longer component to be visited and false

otherwise). MA

is destructed when it reaches EoP

(there is no longer component to be visited) or when

the calculated CRL is negative (invalid sub-solution).

If MA

reaches its EoP with positive CRL, the taken

path presents a solution. Before the destruction, MA

notiﬁes its creator by sending a message containing

the collected information. Algorithm 4 describes how

operates. Thus, all the existing alternatives are

investigated since the created MAgent(s) visit(s) all

the neighboring power sub-grids.

Algorithm 4: MAgent Movement.

for each visited component Y do

calculate (CRL; CPr)

if (CRL(Y ) < 0) then Send Message to

(CRL,CPr)

elsif (@ outgoing lines from Y) then EoP ← true

Send Message to RA

(CRL,CPr, EOP)

else EOP ← false

send message to RA

(RL,CPr,EoP)

switch (number of outgoing lines from Y )

case =1:

Move on Z, the unique connected component,

to Y

call Mobile Agent Movement on Z)

case > 1:

clone on each connected component

for each clone do

call MAgent Movement

end for

end switch

end for

5 SIMULATIONS

In this section, we present an example of an electrical

grid including the three voltage levels (high, medium

and low). The ﬁrst contains 2 power generators and

4 lines. The second one comprises 4 transformers, 7

consumers and 18 lines including 5 emergency ones.

Finally, the third level involves 6 transformers, 17

consumers and 24 lines including 7 emergency ones.

We circled, in red, the electric components to which

FDIRSY assigned RAgents and we drawn the emer-

gency lines in yellow. We begin by presenting two

fault cases; the ﬁrst one is locally resolved while the

second requires a solution from the other connected

sub-grids. Then, we study the behavior of FDIRSY

in terms of required time, exchanged messages and

recovery rate by running on multiple injected prob-

lems. Figure 3 illustrates the simulation of the studied

smart grid by FDIRSY as well as the principal inter-

face of FDIRSY allowing user to design and parame-

terize a power grid (all types of electrical components

and lines).

• Local Resolution

Let us consider, in the ﬁrst case locally resolved, that

there is a voltage instability problem p on the LVL

going from LVT6 to LC13. The concerned RAgent

looks for the existence of solution to LC13 on which

the dominant fault occurs. The solution consists of

activating the emergency line going from LC12 to

LC13. The required time for resolving p, denoted by

TG(p), is smaller than 1.8 µs (Table 1). Thanks to the

proposed fault categorization, FDIRSY resolves only

the fault on LC13 (corresponding to the dominant

fault) instead of searching for solutions (for LC13 and

LC16). Thus, TG(p) is reduced. The recovery is per-

formed at 100% as all the occurred faults are resolved.

Table 1: Local Recovery for Dominant Fault on LC13

Task Description CPU

Detection

detects voltage pro-

1 ηs

blem on LV L

Isolation

deactivates the fai-

< 1 ηs

led LV L

Fault Dominant Fault on

1,6 ηsCategori- LC13 and domina-

zation ted fault on LC16

Local ∃ emergency line

1,76 µssearch between LC12 and

(onLC13) LC13

Resolution

activate the found

< 1 ηs

emergency line

TG(p)

p is a voltage insta-

< 1, 8 µsbility problem

• Non Local Resolution

Let us consider, in the second case, that there is a fre-

quency problem p on LV T 5 which is resolved thanks

to the instantiation of the MAgents. The new found

solution consists on activating the non-local emer-

gency line connecting LC17 to LC3. The required

time, T G(p), is smaller than 20 µs (2). It presents,

also, a comparison between CPUs when p occurs for

the ﬁrst time and for the second or n

time. The use

of data base updated at run-time makes our approach

gainful in terms of exchanged messages and CPUs.

The recovery is performed at 75% as there are 3 re-

paired components from 4 failed ones. In fact, we can

not propose a software solution for both frequency

AnEfficientSimulatorforFaultDetectionandRecoveryinSmartGrids-FDIRSY

137

Figure 3: An example of smart power grid simulated by FDIRSY.

and instability problems as they need a physical hu-

man intervention. Thus, we look for resolving, only,

the engendered faults (like in this case).

• Multiple Problem Resolution

In the last part of this section, we begin by estimating

the required time TG for recovery in the studied power

network. For that, we execute our system on multi-

ple problems at the same time and at different times.

In order to estimate the best and the worst TG, we

run FDIRSY with multiple problems over the small-

est and the bigger sub-grids at different times. These

sub-grids are, respectively, composed of (3 compo-

nents, 3 lines) and (6 components and 5 lines). In fact,

the best and worst T G(p) represent, respectively, the

time spent to recover locally the ﬁrst one and non-

locally the second one (respectively equal to 1, 45 ηs

and 21 µs). We run, also, FDIRSY over the studied

smart grid - which is composed of 12 sub-grids - on

multiple faults at the same time. Figure 4 illustrates

the T G spent to resolve p problems at the same time

(at most p = 8).

It is, also, important to investigate the communi-

cation process between MAgents, RAgents and the

DBAgent. For a given problem p at a time t; (i) if p is

locally resolved, there is 0 exchanged messages, (ii)

if p requires a non-local solution yet found, there is 1

exchanged message and (iii) if p requires a new non-

local solution, the number of exchanged messages is

equal to the number of the existing paths to which we

Table 2: Non-Local Recovery for Dominated Faults on

LC11, LC15 and LC17

Task

Occurrence of p

Detection

detect frequency pro-

1 ηs

blem on LV T 5 1 ηs

Isolation

deactivate the failed

< 1 ηsline between LV T 5 and

LC11 < 1 ηs

Fault dominant fault on LV T 5

4,8 µsCategori- and dominated faults on

zation LC11, LC15 and LC17

4,8 µs

0 local solution for both

3 ηsLocal of dominated and domi-

search nant faults 3 ηs

DBAgent

0 stored solution 2 µs 2,8 µs

Request

Search create MAgent on LC11

0snon local moving to the sub-grid

solution supplied by LV T 2 13 µs

Resolution

activate the found emer-

< 1 ηs

gency line < 1 ηs

TG(p)

p is a frequency pro-

< 7, 6 µs

blem on LV T 5 < 20 µs

add two messages (for the DBAgent request and for

updating data base).

PECCS2015-5thInternationalConferenceonPervasiveandEmbeddedComputingandCommunicationSystems

138

Figure 4: TG for resolving multiple problems.

6 CONCLUSION

In this paper, we propose an original approach for ef-

ﬁcient smart power grid recovery. In order to eval-

uate the proposed approach we develop a simulator

for smart grids. It ensures the detection of the faults,

the identiﬁcation of the encountered problem and the

localization of the failed electrical components (even

the consequent ones) thanks to the proposed fault cat-

egorization. It resolves the problems and searches the

existing solutions according to the deﬁned strategy.

The experimental study showed that our approach is

gainful in terms of faults to be resolved, CPU and

communication. In future works, we look for intro-

ducing a learning module in order to deduce new so-

lutions from other existing ones. We are, also, inter-

ested in large scale tests for larger power grids.

REFERENCES

Ben Meskina, S., Doggaz, N., and Khalgui, M. (2014). New

solutions for fault detections and dynamic recoveries

of ﬂexible power smart grids. In 9

International

Conference ICINCO in Informatics in Control, Au-

tomation and Robotics.

Calderaro, V., Hadjicostis, C. N., Piccolo, A., and Siano, P.

(2011). Failure identiﬁcation in smart grids based on

petri net modeling. IEEE Transactions on Industrial

Electronics, pages 4613 – 4623.

Chertkov, M., Pan, F., and Stepanov, M. G. (2011). Predict-

ing failures in power grids: The case of static over-

loads. IEEE Transactions on Smart Grid.

Fang, X., Misra, S., Xue, G., and Yang, D. (2012). Smart

grid - the new and improved power grid: A survey.

IEEE Communications Surveys and Tutorials.

Jiang, Z., Khalgui, M., Mosbahi, O., and Jaouadi, A.

(2014a). A novel hierarchical multi-agent architecture

for automatic restoration of smart grids. International

Journal of Control and Automation.

Jiang, Z., Mosbahi, O., and Khalgui, M. (2014b). A multi-

agent architecture for the self-healing of sgs based

on iec 61499/61850. Energy Education Science and

Technology Part A. Energy Science and Research.

Massoud, A. and Wollenberg, B. (2005). Toward a smart

grid: power delivery for the 21st century. In IEEE

Power and Energy Magazine, Minneapolis, MN,

USA.

McArthur, S. D. J., Davidson, E. M., Catterson, V. M.,

Dimeas, A. L., Hatziargyriou, N. D., Ponci, F., and

Funabashi, T. (2007a). Multi-agent systems for

power engineering applications - part i: Concepts, ap-

proaches, and technical challenges. In IEEE Transac-

tions on Power Systems.

McArthur, S. D. J., Davidson, E. M., Catterson, V. M.,

Dimeas, A. L., Hatziargyriou, N. D., Ponci, F., and

Funabashi, T. (2007b). Multi-agent systems for power

engineering applications part ii: Technologies, stan-

dards, and tools for building multi-agent systems. In

IEEE Transactions on Power Systems.

Oudalova, A. and Fidigattib, A. (2009). Adaptive network

protection in microgrids. International Journal of Dis-

tributed Energy Resources.

Pipattanasomporn, M., Feroze, H., and Rahman, S. (2009).

Multi-agent systems in a distributed smart grid: De-

sign and implementation. In IEEE/PES Power Sys-

tems Conference and Exposition, Adv. Res. Inst., Vir-

ginia Tech, Arlington, VA.

Rahman, S., Pipattanasomporn, M., and Teklu, Y. (2007).

Intelligent distributed autonomous power systems

(idaps). In IEEE Power Engineering Society General

Meeting, Adv. Res. Inst. of Virginia Tech, Arlington,

VA.

Ramchurn, S. D., Vytelingum, P., Rogers, A., and Jennings,

N. (2011). Agent-based control for decentralised de-

mand side management in the smart grid. In The 10th

International Conference on Autonomous Agents and

Multiagent Systems. International Foundation for Au-

tonomous Agents and Multiagent Systems.

Russell, B. D. and Benner, C. L. (2010). Intelligent systems

for improved reliability and failure diagnosis in distri-

bution systems. IEEE Transactions on Smart Grid.

Vyatkin, V., Zhabelova, G., Ulieru, M., and McComas, D.

(2010). Toward digital ecologies: Intelligent agent

networks controlling interdependent infrastructures.

In First IEEE International Conference on Smart Grid

Communications (SmartGridComm).

AnEfficientSimulatorforFaultDetectionandRecoveryinSmartGrids-FDIRSY

139