brainstorming

download brainstorming

of 19

description

Al based

Transcript of brainstorming

  • Applied Soft Computing 35 (2015) 291309

    Contents lists available at ScienceDirect

    Applied Soft Computing

    j ourna l h o mepage: www.elsev ier .com/ locate /asoc

    Securit ngusing h ing

    BelkacemDepartment of

    a r t i c l

    Article history:Received 10 SeReceived in reAccepted 24 JuAvailable onlin

    Keywords:Optimal poweFirey algorithPattern searchBrainstormingObjective funcSVC

    em prposetive otive ithmled anposectivepensaviatiohe p

    two practical test systems IEEE 14-Bus and IEEE 30-Bus. Simulation results conrm the effectiveness ofthis hybrid strategy for solving the security optimal power ow.

    2015 Elsevier B.V. All rights reserved.

    1. Introdu

    The secutool for powobjective ofof power systate of mulrelated to aconstraintsity are threpower systethese objecperformancconrmed tplex multi strategy to under normhas been wmization te(LP, NLP) [2

    CorresponE-mail add

    http://dx.doi.o1568-4946/ ction

    rity optimal power ow (OPF) is the most importanter system planning, operation and control. The main

    OPF tool which considered as an important sub problemstem planning is to determine the optimal operationti dispersed units by optimizing multi control variables

    specied objective function considering multi practical. Active power loss, voltage deviation and voltage stabil-e important tasks which interest expert specialized inm planning operation and control. The optimization oftive functions affect directly and indirectly the dynamices of practical power system [1]. Many research resultshat optimization of these objective functions is a com-objective problem which require robust and exibleenhance the performance of practical power systemsal and at critical situations. The standard OPF problemidely solved using several classical mathematical opti-chniques such as linear and non linear programming], gradient based method [3], quadratic programming

    ding author. Tel.: +213 665084111.ress: [email protected] (B. Mahdad).

    (QP) [4], Newton-based method [5,6], and interior point meth-ods [7]. A recent survey of the major contributions related tothis category based mathematical methods is presented in [8].Based on a large number of papers proposed to solving variousproblems related to power system planning, operation and con-trol, we can conclude that the deterministic optimization methodsconverge to local optimal solution and can not guarantee globalsolution when considering practical generator constraints (Pro-hibited zones, valve point effect, and multi-fuel options) due tosensibility to initial conditions and to the form of the objectivefunction. These major drawbacks have contributed to the devel-opment of new stochastic based optimization techniques suchas, evolutionary programming [9], improved evolutionary pro-gramming [10], enhanced genetic algorithm (EGA) [11], improvedgenetic algorithms (IGA) [12], adapted genetic algorithm (AGA)[13], particle swarm optimization (PSO) [14,15], differential evo-lution (DA) [1619], improved harmony search algorithm [20],Imperialist competitive algorithm (ICA) [21], gravitational searchalgorithm (GSA) [22], articial bee colony algorithm (ABC) [23,24],rey algorithm (FFA) [25,26], brain storm optimization algorithm(BSA) [27], Application of brainstorming strategy for enhancementdynamic teaching based LMD education system [28]. In generalview, we can conclude that the main difference between thesemethods is related to the robustness of their mechanism search

    rg/10.1016/j.asoc.2015.06.0372015 Elsevier B.V. All rights reserved.y optimal power ow considering loadiybrid FFAPS assisted with brainstorm

    Mahdad , K. SrairiElectrical Engineering, Biskra University, Algeria

    e i n f o

    ptember 2014vised form 2 April 2015ne 2015e 2 July 2015

    r owm

    processtions

    a b s t r a c t

    This paper presents a new power systpattern search algorithm (PS). The putotal voltage deviation, with the objecpower system security. A new interacproposed that allows FFA and PS algorStatic VAR compensator (SVC) is modeing the voltage stability index. The provariables at the same time, namely, aand the reactive power of shunt comtotal power loss and total voltage deand competitive search mechanism. T margin stability rules

    lanning strategy which combines rey algorithm (FFA) with is minimizing total fuel cost, total power loss and reducingf enhancing the loading margin stability and consequently theand simple mechanism, inspired in brainstorming process, iss to explore new regions of the search space. In this study thed integrated in an efcient location which is chosen consider-

    d algorithm is interactive and tries to optimize a set of control power generations, voltage of generators, tap transformers,tors to optimize three objective functions such as: fuel cost,n. These variables are optimized using a exible interactiveroposed planning strategy has been examined and applied to

  • 292 B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309

    determined by the number of parameters to be adjusted, an opti-mization method with simple and interactive mechanism searchand with less parameter to control is preferred.

    Recently, many research results clearly conrmed that eachglobal optiHybrid metion to comhybrid metfor solvingpower systeniques are,modied teferential ev[31], hybridalgorithm (imperialist SQP) algori[35], new tive algorithoptimizatiosive teachin[39] and hiliterature rhow adjustbetween vaances of eatime. The mmethods is nication prexploration

    Flexible Hingorani atechnologytechnologysource prodgin stabilitydevices suctrollers (TCthe electricpower systgies have bow considmulti objecmulti FACTand Nageshoptimal powow controtion algoritconsidering[47] a hybadapted anfor congestproposed fosecurity.

    Firey ation methodapplied by to power syis preferredparametersevolutionaring many prcontrol. Thethe rules pprocess wauals workin

    is rstly developed and formulated as an optimization problem byShi in 2011 [27]. To date, the BSO algorithm has not been appliedintensively to solving the complex problems related to power sys-tem planning and control, very recently in [50], Rezaee Jordehi

    ed an efcient algorithm based brain storm optimizationhm (rove viatioant athis rd ore ad

    experedmpled cmued inant o

    loss SVCpropgencstemering

    urity

    urity systetimave fustraitructs:

    izatit denal ithmybridsted nizedods.r ows in of poof coithmctive

    opttiond indmunial anqualiess d.

    his wA cone the

    mecwo md fromization method has its advantages and drawbacks.thods considered as an alternative and robust solu-bine different methods. In the recent literature varioushods have been proposed and applied with success

    many complex and combined problems related tom planning, operation and control, some of these tech-

    Evolving ant direction differential evolution [29], Aachinglearning based optimization [30], hybrid dif-olution (DE) with particle swarm optimization (PSO)

    fuzzy particle swarm optimization and NedlerMeadHFPSO-NM) [32], chaotic improved PSO [33], hybridcompetitive-sequential quadratic programming (HIC-thm [34], A modied shufe frog leaping algorithmmodied and hybrid modied imperialist competi-ms [36], adaptive biogeography based predatorpreyn technique [37], self-evolving brain-storming inclu-glearning-based algorithm [38], A hybrid GAPSSQPerarchical adaptive PSO [40]. As well described in theeview, the structure of these methods is based oning dynamically their parameters and on combinationrious methods to exploit efciently the best perform-ch method to achieve the global solution at a reducedain idea introduced in the mechanism search of theseto provide an intelligent adaptive and exible commu-ocess to guide efciently the processes search during

    and exploitation stages.AC transmission Systems (FACTS) rst introduced bynd Gyugyi [41], is relatively considered as one of recent

    installed in many practical electrical networks. FACTS is an alternative solution to ensure balance betweenuction and power demand particularly at loading mar-. In recent years many types of controllers based FACTSh as: shunt Controllers (SVC, STATCOM), series Con-SC, SSSC) and hybrid Controllers (UPFC) integrated inity market to improve the performances of the practicalem [42]. In the literature many techniques and strate-een proposed to enhance the security optimal powerering the integration of FACTS devices [43]. An efcienttive strategy based differential evolution consideringS technology is presented in [44]. Venkateswara Rao

    Kumar [45] proposed a BAT search algorithm baseder ow for generation reallocation with unied powerller, in [46] a multi-objective adaptive clonal selec-hm is proposed for the solving optimal power ow

    multi-type FACTS devices and load uncertainty, inrid bacterial foraging and NelderMead algorithm isd applied for optimal location of series FACTS devicesion management, and in [48] a strategy based PSO isr dynamic control of UPFC to enhance power system

    lgorithm is one of the recent metaheuristic optimiza-s proposed by author in [25], this method adapted andresearchers to solve many complex problems relatedstem planning operation and control. In this work FFA

    for its simplicity to adapt and program and has a few to adjust. Pattern search algorithm is also an efcienty technique, adapted and applied with success to solv-actical problems related to power system operation and

    brainstorming concept rst introduced by Osborn [49],roposed by Osborn demonstrated that brainstormings more effective in generating new ideas than individ-g. The adaptation of brainstorming principle and rules

    proposalgoritto impage deimport

    In standarules acess, toconsidto accoproposble comproposimportpowerof oneof the convertest syconsid

    2. Sec

    Secpowerthe opobjectiity conbasic sfollow

    Optimegy, the algoror hadjuorgameth

    Poweworkrole tion algor

    Objeto bedeviamize

    Comglobtion procused

    In tthe FFachievsearchthese tinspireBSOA) for optimal location and control of FACTS devicesvarious objective functions such as power loss and volt-n and overload minimization, the results achieved arend competitive.work and to enhance the performances of the twoptimization methods (FFA and PS), the brainstormingapted and used to assist FFA and PS during search pro-loit with efciently the search space. In this study, FFA is

    as a global optimization technique and the PS is adaptedish the task of local search, the main particularity of theombined optimization method is related to the exi-nication rules inspired from brainstorming process. Theteractive hybrid approach applied to optimize threebjective functions: reduce the total fuel cost, the total

    and voltage deviations by considering the integration compensator at efcient location, The performancesosed planning strategy in term of solution quality ande characteristic have been tested on the two practicals, IEEE 14-Bus and IEEE 30-Bus at normal condition and

    loading margin stability.

    optimal power ow

    optimal power ow is a fundamental sub-problem ofm planning, the main objective of this strategy is to ndl settings of control variables in terms of one or morenction while satisfying operating equality and inequal-nts [1]. Fig. 1 shows the strategy of security OPF, theure of security OPF may be presented and described as

    ion method: represent the core of the optimization strat-pends on the complexity of the problem to be solved,results achieved depend on the particularity of the

    used such as standard algorithms (GA, PSO, ABC), method, how the initial parameters are chosen andduring search process, how the control variables are, and how the global method is hybridized with other

    (PF) based NewtonRaphson algorithm: this second taskcoordination with the optimization method, the mainwer ow is to determine the state variables and viola-nstraints. The equality constraints are veried using PF.functions: in general, the well known objective functionsimized are: fuel cost, emission, power losses, voltage, and voltage stability, these objective functions are opti-ividually or simultaneously.cation process: is related to the interaction betweend local optimization methods to enhance the solu-ty and convergence characteristics. The communicationepends on the global and local optimization methods

    ork, we have used two simple optimization methods,sidered as a global optimization method adapted to

    exploration phase and the PS considered as a localhanism coordinated to enhance the exploitation search,ethods are coordinated using exible practical rules

    m brainstorming process.

  • B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309 293

    jectiv

    In generconsideringusing the fo

    Minimize J

    Subject t

    g(x, u) = 0

    h(x, u) 0

    where J is thtives. g is thsecurity ineables, u repstate vectortudes VL, slapower Qg.

    x = [(1,...,NThe cont

    generator tedynamic co

    u = [Pg(1,...,

    2.1. Objecti

    2.1.1. MinimMinimiz

    considered operation. T

    presws:

    = m

    NG iFig. 1. Strategy of security multi ob

    al the basic mathematical formulation of security OPF one or more objective function may be representedllowing generalized expressions:

    i(x, u) i = 1, . . ., Nobj (1)

    o:

    cost exas follo

    J1(x, u)

    where(2)

    (3)

    e ith objective function, and Nobj is the number of objec-e equality constraints, h is the system operation andquality constraints. x represents the vector of state vari-resents the vector of control variables. In general, the

    includes bus voltage angles , load bus voltage magni-ck bus real power generation Pgs and generator reactive

    L), VL(1,...,NL), Pgs, Qg(1,...,NG)]T (4)

    rol variable vector consists of real power generation Pg,rminal voltage Vg, shunt capacitors/reactors Bsh, shuntmpensators (SVC) Bsvc and transformers tap ratio t.

    NG1), Vg(1,...,NG), Bsh(1,...,Nsh), Bsvc(1,...,Nsvc), t(1,...,Nt)]T

    (5)

    ve functions

    ization of total fuel costation of fuel cost is an important task for expert andas an important index for power system planning andhe simplied objective function f is the total generation

    generation ith generato

    2.1.2. MinimThe obje

    total voltagequation:

    J2(x, u) = m

    where Vref i

    2.1.3. MinimThe obje

    to be optim

    J3(x, u) = m

    2.2. Constra

    2.2.1. EquaThe equ

    equilibriume OPF.

    sed as a quadratic function, mathematically formulated

    in fi = min(

    NGi=1

    (ai + biPgi + ciP2gi))

    (6)

    s the number of thermal units, Pgi is the active power

    at unit i and ai, bi and ci are the cost coefcients of ther.

    ization of voltage deviationctive function associated to the minimization of thee deviation at all load buses is described by the following

    in VD = min(

    iNL|Vi Vref |

    )(7)

    s the desired voltage at all load buses.

    ization of power lossesctive function associated to the total active power lossized is formulated by the following equation:

    in Ploss = min(

    NLk=1

    gk[(tkVi)2 + V2j 2tkViVj cos ij]

    )(8)

    ints

    lity constraintsality constraints g(x) of the OPF reect the physical

    of active and reactive power between production and

  • 294 B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309

    demand at all buses. In other term, the principle of energy conser-vation must be veried at each bus. The following two generalizedequations reect the equilibrium and conservation of active andreactive power at all buses:

    Pgi Pdi V

    Qgi Qdi

    where N is ttive power power demrespectivelyj respectiveadmittance

    2.2.2. InequThese ar

    ational powcompensatothese constfollows:

    2.2.2.1. Gen

    Limits on

    Vmingi Vg Limits on

    Pmingi Pg Limits on

    Qmingi Qg

    2.2.2.2. Tra

    Limits on

    tmini ti

    2.2.2.3. FAC

    Parametetheir limi

    Xmin XF

    2.2.2.4. Secsumer instaTo achieve tbuses) and tconstraints

    Sli Smaxli ,

    VminLi VLi

    where NBRof loading b

    2.2.3. Constraints managementIn order to take in consideration the violation of security con-

    straints, a modied tness function is adapted using the penaltyfunction. The inequality constraints are immerged within the

    ve funtedsion:

    u) =

    =

    (PGs

    v, e occtudestion ible ndlin{

    VL

    VL{

    Q

    Q{PmG

    PmG

    Smaxbri

    adin

    loadlity ocritic

    nddinaltage

    load poinentatation

    P Q

    Pbase Qbaseding

    ltage

    to systor caonsenormancei

    Nj=1

    Vj(gij cos ij + bij sin ij) = 0 (9)

    Vi

    Nj=1

    Vj(gij sin ij bij cos ij) = 0 (10)

    he number of buses, Pgi, Qgi are the active and the reac-generation at bus i; Pdi, Qdi are the real and the reactiveand at bus i; Vi, Vj, the voltage magnitude at bus i, j,; ij is the phase angle difference between buses i andly, gij and bij are the real and imaginary part of the

    (Yij).

    ality constraintse the sets of physical constraints associated to all oper-er system elements (generator, transformer, lines, andrs), from production source to load bus. Generally,raints are organized and mathematically expressed as

    erator constraints.

    the generator bus voltage magnitude:

    i Vmaxgi , i = 1, 2, . . ., NPV (11)

    the output of active power generation:

    i Pmaxgi , i = 1, 2, . . ., NPV (12)

    the output of reactive power generation:

    i Qmaxgi , i = 1, 2, . . ., NPV (13)

    nsformer constraints.

    the tap ratio (t) of transformer.

    tmaxi , i = 1, 2, ..., NT (14)

    TS controllers constraints.

    rs of shunt FACTS controllers must be restricted withints.

    ACTS Xmax (15)

    urity limits. Energy produced must be delivered to con-ntaneously and within the admissible security values.his task, the constraints on voltage at loading buses (PQransmission line loadings are considered. These security

    are expressed as follows:

    i = 1, 2, . . ., NBR (16)

    VmaxLi , i = 1, 2, . . ., NPQ (17)

    and NPQ are the number of branches and the numberuses respectively.

    objectiaugmeexpres

    Jmod(x,

    Penalty

    + P

    wheretion armagnigeneraadmissthe ha

    V limLi =

    Q limGi =

    PlimGs =

    Slimbri =

    2.3. Lo

    Thethe abiunder used toin coorand vo

    Thetem torepresthe rel

    Pnew =Qnew =where

    Pnew, Qnew,: loa

    2.4. Vo

    DuepowerOperatity to cunder disturbnction. The mathematical formulation of the proposed objective function can be expressed by the following

    J(x, u) + Penalty (18)

    v

    NLi=1

    (VLi V limLi )2 + Q

    NGi=1

    (QGi Q limGi )2

    PlimGs )2 + br

    NBRi=1

    (Sbri Slimbri )2

    (19)

    Q, P and br are penalty factor terms, if limits viola-urred at dependent variables such as: load bus voltage, real power generation at slack bus, reactive poweroutput and transmission loading, they are xed to theirlimits (max or min). The following expressions illustrateg of constraints violations:

    mini

    if VLi < VminLimaxi

    if VLi > VmaxLi(20)

    minGi

    if QGi < QminGimaxGi

    if GGi > QmaxGi(21)

    ins if PGs < P

    minGs

    axs if PGs > P

    maxGs

    (22)

    if Sbri > Smaxbri (23)

    g margin stability

    ing margin stability is an important index which reectsf the system to deliver dynamically power to consumeral situations. In this study loading margin stability is

    the best location of SVC controller and also optimizedtion with other objective functions such as power loss

    deviation.ing margin stability is determined by pushing the sys-t of collapse by load increase at all load bus. Based onion shown in Fig. 2, the following equation describes

    between load increase and loading margin stability.

    base (24)

    base (25)

    : the new and basic active power demands: the new and basic reactive power demands

    margin stability

    stability index

    economical and environment constraints, practicalem must to operate close to their admissible limits.n estimate the reliability of a power system by its abil-rve constantly acceptable voltage delivered to consumeral condition and also at critical situations (load increase,s) [51]. L-index is one of the efcient voltage stability

  • B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309 295

    index propobus indicatbus; L-indesponding toequations dage stability

    First, thebuses expre[

    IL

    IG

    ]= [Yb

    where YLL, YSecond, [

    VL

    IG

    ]= [H

    where H maHGL and HGGPV buses an

    The H m

    [H] =[

    ZL

    YGL

    Finally, tfollows:

    Lj =1

    NGi=

    For stabstable zone

    3. Modelin

    The statidevices largin Fig. 3, thparallel Thyhigh voltagtogether widynamicallyically capac

    howle of

    SVCcontrons d

    jBSVC

    reased a

    BSVCi

    tima

    he litn, siz

    stud efcion oering

    folloed lo

    ticaRun stabiSort valueFig. 2. Loading margin stability.

    sed by Kessel and Glavitch [52]. L-index of a speciedes the proximity of voltage collapse condition of thatx varies between two extreme values, 0 and 1 corre-

    no load and voltage collapse situation. The followingescribe the mathematical formulation of L-index volt-.

    buses are separated into two parts: PV buses and PQssed as follows

    us] .

    [VL

    VG

    ]=[

    YLL YLG

    YGL YGG

    ].

    [VL

    VG

    ](26)

    LG, YGL and YGG are sub matrix of Ybus.the following transformation can be written:

    ] .

    [IL

    VG

    ]=[

    HLL HLG

    HGL HGG

    ].

    [IL

    VG

    ](27)

    trix is generated by the partial inversion of Ybus, HLL, HLG,are sub matrix of H, are voltage and current vector of

    d PQ buses, respectively.atrix is given as follows:

    L ZLLYLGZLL YGG YGL.ZLLYLG

    ](28)

    he L-index corresponding to each bus is expressed as

    Fig. 4 sprincip

    Theshunt equati

    ISVC = The

    expres

    QSVCi =

    3.1. Op

    In tlocatioIn thisexploitreductconsid

    Thepropos

    1. Iden1.1

    1.2 1

    HLGji ViVj

    j = 1, 2, . . ., NL (29)le situations the value of Lj at each bus must be in the

    Lj < 1.

    g of static VAR compensator (SVC)

    c VAR controller (SVC) [1,57] is one of the shunt FACTSely exploited and installed in the world. As well showne basic structure of SVC consists of connected anti-ristors to provide controllability. Air core reactors ande AC capacitors are the reactive power elements usedth the Thyristor valves. The SVC has the ability to control

    the voltage at critical buses by exchanging dynam-itive or inductive reactive power with the network.

    cons2. Select th

    buses NB3. Identify

    example4. Locate th5. Locate th

    bus C is SVC is sh

    4. Overvie

    Firey rmetaheurisShe Yang aadapted anbinatorial oFig. 3. Basic circuit schema of SVC.

    s the SVC steady-state model representation and basic SVC.

    model used in this study is based on representing theoller as variable susceptance; the following simpliedescribe the susceptance SVC model.

    V (30)

    ctive power QSVCi

    exchanged with the bus i can bes:

    V2i (31)

    l location strategy of SVC controllers

    erature various techniques were proposed for optimaling and control of multi type of FACTS devices [1,57].y a simple and exible location strategy is proposed toiently the integration of SVC controller to maximize thef power losses and voltage deviation under normal and

    loading margin stability.wing steps summarize in brief the methodology of thecation strategy.

    tion of critical buses based on voltage stability index.continuation power ow to determine loading marginlitythe buses based on their minimal voltage stability index. The buses with low voltage stability index value are

    idered as the rst best candidate for SVC location.e weak bus with low voltage stability index from all, based on Fig. 5, k is the candidate bus.all branches associated to the rst candidate bus: for

    i-j-m as well shown in Fig. 6.e branch associated to the related weak bus.e SVC compensator in the middle of the selected branch,the best location. The owchart for optimal location ofown in Fig. 7.

    w of FFA

    esearch algorithm is one of the new Biology inspiredtic algorithms which have recently introduced by Xin-t Cambridge University in 2007 [25]. FFA has beend applied with success to solving many complex com-ptimization problems FFA [26]. The main advantages of

  • 296 B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309

    Fig. 4. Steady-state model representation and basic principle of shunt compensation.

    the rey asubgroups, enhance thwith anotheFFA are premechanismrules which[26,53,54],

    1. All reies are unisex, the sex of reies has no effect on theirmovement in the environment, they will move only toward moreattractive and brighter ones.

    2. The degree of attractiveness of a rey is proportional to itsbrightness which decreases as the distance from the other reyincreases due to the fact that the air absorbs light. If there is nota brighter or more attractive rey than a particular one, it willthen move randomly [26,53].

    level of light intensity of a rey is determined by the valuee ob

    ctive basbetw

    = 03. Theof th

    AttraTh

    tion

    (r) Fig. 5. Optimal location of SVC: step 1.

    lgorithm is its ability to divide initial population intothis pertinent propriety is exploited in this study toe robustness of the mechanism search in coordinationr local search algorithm. The basic steps of the standardsented in the schematic diagram shown in Fig. 8. The

    search of FFA has three particular idealized interactive are inspired from ashing characteristics of real reiesthe particularity of these rules described as follows:

    where r isattractivecontrols t

    DistanceIn the s

    positions

    rij = ||xi

    where xi,kd is the n

    MovemeThe mo

    tive re

    xt+1i

    = xti

    Fig. 6. Steps for optimal location of SVC Comjective function related to a specied problem.

    enessic mathematical formulation of the attractiveness func-een reies is expressed by the following equation:

    exp(rm), with m 1, (32)

    the distance between any two reies, 0 is the initialness at r = 0, and is an absorption coefcient whichhe decrease of the light intensity.

    earch space, the distance between two reies i and j at xi and xj can be dened by the following relation:

    xj|| =

    dk=1

    (xi,k xj,k)2 (33)

    is the ith component of the spatial coordinate xi, xj andumber of dimension.ntvement of a rey i which is attracted by a more attrac-y j is given by the following equation [53,54].

    + 0. exp(r2ij ) (xj xi) + (rand 0.5), (34)pensator.

  • B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309 297

    where theond termintensity random mones. Ii anthat is assbe evaluadescriptio

    ParameteLike ma

    initial valgreatly th

    Fig. 9. Basic pseudo code for the proposed rey algorithm.

    ithm. These parameters depend on the nature and particu- of t

    [0, 1omp= 1;e att

    rvie

    pattccesy coFig. 7. Basic ow chart for optimal location of SVC.

    rst term is the current position of a rey, the sec- is used for considering a reys attractiveness to lightseen by adjacent reies, and the third term is for the

    algorlarity

    the c

    0Th

    5. Ove

    Thewith suto manovement of a rey in case there are not any brighterd Ij are two variables which reect the light intensityociated with a specied tness function of particles toted. The pseudo code depicted in Fig. 9 gives a briefn about the standard rey algorithm.rs settingsny metaheuristic optimization methods, choosing theues of parameters is an important task which affectse efciency of results and convergence behaviors of the

    global optimadvantage oputationallto the initiato be used optimizatio

    As well iing a set of p[1 0], [0

    Fig. 8. Mechanism search of the rey alghe problem to be solved.]: is a randomization parameter determined based onlexity of the problem to be solved.

    ractiveness or absorption coefcient = 1.

    w of pattern search (PS)

    ern search (PS) is an evolutionary algorithm applieds for solving a variety of optimization problems relatednstrained complex research areas. Compared to manyization methods, pattern search (PS) method has thef being very simple in concept, easy to implement, com-y competitive, the drawback of this algorithm is relatedl conditions, for this reason this method is very suitableas a local search routine in coordination with a globaln method [39,57].llustrated in Fig. 10b, the algorithm starts by establish-oints called mesh, around the initial point X0. [1 0], [0 1],1], the determination of the next position representedorithm.

  • 298 B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309

    gorith

    by directiofunction. Dothers direcbe found in

    The basiIn this studtion factorsnumber of complexityand dynamsearch proc

    6. The pro

    The matrapping intration and PS algorithmconsidered mal regionsfunctions (fied stage nd new soto generatixed shuntcesses is rereached.

    6.1. Method

    As well ithe method

    1. Identica2. Specify t

    F2 (poweoptimal s2.1 Activ

    variaF3

    2.2 Savetive f

    optimcontrVoltaSave

    Restrattrol follobinat FFA

    optixedmizeTap tSave

    optimtranss areReacSaveFig. 10. Flow chart of the PS al

    n and step size is based on the evaluation of tnessetails description and graphic representation of PS andt search methods for unconstrained optimization may

    [36,57].c research steps of PS algorithm are shown in Fig. 10a.y, the mesh size and the mesh expansion and contrac-

    are selected as 1, 2 and 0.5 respectively; the maximumgeneration is taken between 50 and 150 based on the

    of the problem. The vector of initial points X0 is selectedically updated based on communication process duringess.

    posed dynamic power system planning strategy

    in task of this proposed hybrid method is to avoido a local optimum by dynamically balancing the explo-exploitation capability of the two standard FFA ands. Initially only active power of generating units are

    and optimized using FFA to explore the best sub opti- by identifying critical units considering three objectiveuel cost, power loss and voltage deviation). At a spec-the partial solution found are communicated to PS to

    3. Theing 3.1 3.2

    4. Theare opti4.1 4.2

    5. Thetap able5.1 5.2 lution by optimizing two control variables associatedng voltages, tap transformers and reactive power of

    compensators and SVC Controllers. The search pro-peated until a number of minimum specied trails are

    ology of the proposed combined algorithm

    llustrated in Fig. 11, the following points describe brieyology of the proposed combined algorithm.

    tion of critical buses.he objective function to be optimized: F1 (fuel cost),r loss) or F3 (voltage deviation). Generate initial sub-olution, this task achieved using FFA1e power generation is considered as the rst controlble to be optimized to minimize individually F1, F2 and

    the sub-optimal solution corresponding to each objec-unction.

    6.2. Coordi

    6.2.1. OverThe pro

    reasoning usummarizetion betweebrainstormproposed bmore effectThe Osbornterm qualitipate and gindividualsexible anding rules arbetween diVery recentthe followilows:m.

    al active power generation is xed: Run PS, the follow-ol variables are to be optimized:ge of generation units

    the new best solutionmark: the main advantage of the proposed exibleegy is that expert can choose the sequence of con-variables associated to each optimization stage, thewing description are a sample of basic and possible com-ion.1, FFA2, FFA3 or FFA2, FFA3, FFA1 or FFA3, FFA1, FFA2

    mal active power generation and voltage of generators: Run FFA2: the following control variables are to bed:ransformer,

    the new best solutional active power generation and voltage of generators,

    former are xed: Run FFA3: the following control vari- to be optimized:tive power of SVC compensator.

    the new best solution.nation based on brainstorming process

    viewposed interactive hybrid method is based on humansing brainstorming process; the following descriptions the particularity and importance of the communica-n experts to achieve the desired optimal objectives. Theing process rst introduced by Osborn [49], the rulesy Osborn demonstrated that brainstorming process wasive in generating new ideas than individuals working.s rules have the particularity to generate new ideas iny and quantity, it gives chance to all individuals to partic-enerate original idea. Experience conrmed that some

    have excellent idea, but needs to be exploited using interactive mechanism. In this study the brainstorm-e adapted and applied at specied stages to coordinatefferent solutions found using FFA and PS algorithms.ly, many optimization algorithms developed based onng basic rules named Osborns rules formulated as fol-

  • B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309 299

    1. Focus on to particiand prod

    2. Suspendigenerateviduals pideas witevaluatio

    3. Welcomefacilitatevariety o

    4. Go to quacombine

    In indusdifcult prosolution prosons (experbrain stormin term of s

    Fig. 12 ilnication baprocess unAs well claeach globalA perfect soperformancbeing applipractical reFig. 11. Mechanism parallel search using FFA and PS assisted

    quantity [49]: This rule gives chance to all individualspate and to generate new ideas which allow generatingucing a greater number of ides.ng judgment [49]: In brainstorming, criticism of ideasd should be put on hold by suspending judgment, indi-articipating will have the chance to generate unusualhout judgment. Criticism is reserved for a later stage forn.

    unusual ideas [49]: The particularity of this rule is to and allows working in an environment with a goodf ideas, unusual ideas are welcomed.ntify [49]: A strong evaluation mechanism selects ands ideas to form good idea.

    trial area, experience conrmed that when we face ablem which every single expert cannot solve or theposed fail to attain the desired objective, a group of per-ts), especially with different experience, get together to, the problem can usually be solved with high efciencyolution quality and time required [23].lustrates the basic principle of the interactive commu-sed on brainstorming process. During communication-expectable intelligence and original idea can occur.ried and conrmed in many recent research results,

    optimization method has advantages and drawbacks.lution is to hybrid these methods to achieve the beste. The optimization based on brainstorming processed to solving many complex problems related to manysearch areas. Details description about brainstorming

    mechanismin using thethe main iduniversity iintroducedrmed the ienhance th

    6.2.2. BrainThe ada

    developed 2011 [27]. Tsively to soplanning anproposed aand controlsuch as pozation. As wBSOA propooperations.

    Initializatthe initiaof searchbrainstor

    Clusteringclusters (K-means s with brainstorming rules.

    search can be found in [38,50]. Our rst contribution brainstorming concept was rst investigated in [28],ea proposed is to enhance the teaching quality in ourn Algeria within the educational system named LMD

    recently in our universities. Teaching experience con-mportance of introducing the brainstorming process toe educational level of our students.

    storm optimization algorithmptation of brainstorming principle and rules is rstlyand formulated as an optimization problem by Shi ino date, the BSO algorithm has not been applied inten-lving the complex problems related to power systemd control, very recently in 2015, Rezaee Jordehi [50]n efcient algorithm based BSOA for optimal location

    of FACTS devices to improve various objective functionswer loss and voltage deviation and overload minimi-ell depicted in Fig. 13, the basic mechanism search ofsed by Shi [27] is formulated based on the following 4

    ion: like the majority of global optimization techniques,l population is generated randomly in feasible region

    space. In BSOA, individuals are analogous to ideas inming.: this operation divides the population into differentbrainstorming groups) according to individual features.trategy is used for clustering [50].

  • 300 B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309

    Fig. 12. Interactive communication based on brains

    Fig. 13. Basic mechanism search of BSOA.

    Cluster ceideas of bbility P1, aindividuaindex to b

    Individuauals (ideaindex P2 ithe follow

    Xid(t) = Xwhere N(and standand is up

    S(t) = log

    where tmrespectiv

    The operrepeated

    6.2.3. PropoIn this w

    and PS durifunction F1ability P1 drelated to ocommunicabation operexploitationexible hybbetween di

    1. For thePS.torming process.

    nter perturbation: cluster centers are analogous to bestrainstorming groups, during this stage and with proba-

    cluster center is selected and is replaced with a randoml [50]. P1 is control parameter and considered as analance between exploration and exploitation capability.l perturbation: in order to improve the quality of individ-s) during search process, a second specied probabilistics introduced to create perturbation to individuals usinging equation:

    id(t) + S(t)N(0, 1) (35)

    0, 1) represents Gaussian distribution with mean zeroard deviation of one. S(t) is the perturbation step sizedated using the following equation:

    sig(

    (tmax/2)tK

    ) r (36)

    ax and t represent the maximum and current iterationely, K is a scale factor.ations of the mechanism search are evaluated anduntil a number of maximum generations is reached.

    sed brainstorming rulesork, the brainstorming rules are used to assist FFA

    ng search process, for each specied principal objective, the individuals of population are perturbed with prob-ened by experience, the sub optimal vector decisionther secondary objective functions such as F2, and F3 areted to the principal objective function F1, this pertur-ation is important to balance between exploration and

    capability. Fig. 14 shows the owchart of the proposedrid method. The following steps describe the interactionfferent methods during search process.

    next trail, communicate the best solution found to

  • B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309 301

    Fig.

    2. Brainstowithin operatiosuch asworst idate diveCommu

    At spetrol varfunctionsolutionsidered

    0.92

    0.94

    0.96

    0.98

    1

    1.02

    1.04

    1.06

    .u)Vol

    atge

    Mag

    nitu

    des

    (p 14. Flowchart of the proposed interactive hybrid method.

    Fig. 15. Perturbation during brainstorming process.

    rming process: The basic idea introduced and adaptedthe brain storming process is based on perturbationn, the best idea found from other objective functions

    F2, and F3 at specied stages which considered as theeas from the objective function F1 are exploited to cre-rsity in search space to improve the near global solution.nication process at specied stages is shown in Fig. 15.cied stage (iteration), communicate the optimal con-

    iables (Pgi, Vgi, Ti, Qci) corresponding to other objectives (F2, F3) to PS, it is important to clarify that the best

    found considering for example F2 (power loss), con- as the worst solution when considering the objective

    Fig.

    functioexplore

    3. Generaand woa speci

    4. Conver4.1 For eac

    stopped4.1 The pro

    achieve

    7. Case stu

    In this sdeveloped iThe propossystems. Onbe retrievedinvestigateteristics corgiven.

    7.1. Test sy

    This test1-2-3-6-8, one shunt Vthe total lothe values limits for acTable 1

    Case 1: FThe tota

    ing total p0.3479 p.u. the SVC comvoltages at

    Case 2: PTable 2

    function vacondition (Pwith the pthe results age deviatiexchanged line 13-14 ivoltages at 0 5 10 15Number of Buses

    16. Distribution of optimized voltages at all buses: case 1.

    n F1 (fuel cost optimization), this stage is important to new feasible regions (ideas) during process search.te new database: Save the new solution (best solutionrst solution): repeat the communication processes untiled Ntrail.

    genceh algorithm (partial solution), the search process

    if a specied number of generation is reached (Gmax).cesses search stopped if a specied number of trails ared.

    dies and numerical results

    ection, the proposed interactive planning strategy isn the Matlab programming language using 7.8 version.ed approach applied to many electrical practical test

    the standard IEEE 14-Bus and IEEE 30-Bus, data can from [55,56], in this study only one SVC controller is

    d. Due to paper length limit, the convergence charac-responding to the different objective functions are not

    stem 1: IEEE 14-Bus

    system consists of 5 generating units located at buses:two transformers located at lines 4-7, 4-9 and 5-6 andAR compensation installed at bus 9. At normal conditionad demand to satisfy is 259.0 MW, the generator data,of fuel cost coefcients, the minimum and maximumtive and reactive power of generating units are given in

    uel cost minimization

    l fuel cost achieved is 8078.63 $/h, the correspond-ower loss and voltage deviation are: 9.221 MW, andrespectively. The optimized reactive power delivered bypensator is 5.98 MVAR. The distribution of optimized

    all buses is shown in Fig. 16.ower loss minimizationshows the optimized control variables and objectivelues with real power loss minimization at base loadD = 259.0 MW). The minimum real power loss obtainedroposed strategy is 0.634 MW which is better thanfound on case 1 (9.221 MW), the corresponding volt-on achieved is 0.3816 p.u. The optimal reactive powerby the SVC and the network installed in the middle of thes 5.93 MVAR. For this case the distribution of optimizedall buses is shown in Fig. 17.

  • 302 B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309

    Table 1Technical admissible paramaters of generators and the fuel cost coefcients: IEEE 14-Bus.

    Bold font with red color has been used to indicate the limits of control variables.

    Table 2Optimized control variables for IEEE 14-bus system: cases 2 and 3.

    Bold font with blue color has been used to indicate best results of the proposedplanning strategy.

    0 5 10 150.92

    0.94

    0.96

    0.98

    1

    1.02

    1.04

    1.06

    Vol

    atge

    Mag

    nitu

    des (

    p.u)

    Number of Buses

    Fig. 17. Distribution of optimized voltages at all buses: case 2.

    0 0.92

    0.94

    0.96

    0.98

    1

    1.02

    1.04

    1.06

    Vol

    atge

    Mag

    nitu

    des (

    p.u)

    Fig.

    Case 3: VIn this c

    compared trespondingpower exch13-14 is 6 Mfor minimuobjective fusator are shat all buses

    Case 4: Pimization

    To test egy, the towith increatotal powerity = 1.6 p.ucorrespondinjected at security counits, branlimits. Fig. buses, Tablsponding stin all brancvariables an

    Table 3The best results for three optimized objective functions: IEEE 14-Bus: PD = 259 MW.

    Bold font with blue color has been used to indicate best results of the proposed strategy.5 10 15Number of Buses

    0.065

    18. Distribution of optimized voltages at all buses: case 3.

    oltage deviation minimizationase the total voltage deviation is reduced to 0.0465 p.u.o case 1 (0.3479 p.u.) and case 2 (0.3816 p.u.), the cor-

    total power loss is 1.2516 MW. The optimal reactiveanged by the SVC installed in the middle of the lineVAR. Table 2 shows the optimized control variables

    m voltage deviation. The best results of three optimizednctions considering the installation of one SVC compen-own in Table 3. The distribution of optimized voltages

    is shown in Fig. 18.ower loss minimization and loading margin stability max-the feasibility and robustness of the proposed strat-tal active power loss is minimized in coordinationsing load to the limit of margin stability. The optimal

    loss achieved is 3.0912 MW at loading margin stabil-. which correspond to the new load PD = 414.4 MW, theing voltage deviation is 0.2672 p.u. The reactive powerthe middle of branch 13-14 by the SVC is 6 MVAR. Allnstraints in term of reactive power limit of generatingches loading and voltages are within their admissible14 shows the distribution of voltage magnitude at alle 4 shows the optimized control variables, the corre-ate variables such as, active and reactive power transitshes are depicted in Table 5. As we can see all controld state variables are within their admissible limits.

  • B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309 303

    Table 4Optimized control variables for IEEE 14-bus system: cases 4 and 5: PD = 414.4 MW.

    Bold font with blue color has been used to indicate best results of the proposedstrategy.

    Case 5: Vity maximiz

    Voltage ering energsecurity limengineer. I0.0890 p.u.,point of brapower loss achieved wshows the othe correspotransits in atages at all b

    Fig. 19. Distribution of optimized voltages at all buses: cases 4 and 5: Loading marginstability = 1.6 p.u.

    show the repartition of power transit in branches, a comparativestudy in term of solution quality is depicted in Table 7. It is impor-tant to note that all security constraints are satised.

    7.2. Test system 2: IEEE 30-Bus: without considering SVCcontroller

    This second test system consists of 6 generating units locatedes: 1-12 d atl contor daximre giic pllti ompoe1: Me2: Me3: Me1: M

    Table 5Active power t

    Bold font witholtage deviation minimization and margin loading stabil-ationmagnitude is an important power quality index; deliv-y to consumer at critical situation within admissibleit is an important and difcult task for expert and

    n this test case the voltage deviation minimized at the optimal reactive power injected by SVC at the mid-nch 13-14 is 4.92 MVAR, the corresponding total activeis 3.6594 MW. The maximum loading margin stabilityithout violation of security constraint is 1.6 p.u. Table 4ptimal control variables corresponding to this test case,nding state variables such as, active and reactive powerll branches are depicted in Table 6. The optimized vol-uses for cases 4 and 5 is shown in Fig. 19. Tables 5 and 6

    at bus6-10, 4installenormageneraand munits adynamthe muthree i

    CasCasCasCasransit in branches: case 4: PD = 414.4 MW/loading margin stability: LF = 1.6 p.u.

    red color has been used to indicate the maximum limits of power transit in branches.-2-5-8-11-13, four transformers located at lines 6-9,and 28-27 and nine shunt VAR compensation buses

    buses 10, 12, 15, 17, 20, 21, 23, 24 and 29 [51,52], atdition the total load demand to satisfy is 283.4 MW. Theata, the values of fuel cost coefcients, the minimumum limits for active and reactive power of generatingven in Table 8 and available in [55,56]. The proposedanning strategy based FFAPS has been applied to solvebjective OPF. In this study and for fair comparison onlyrtant cases are considered:inimization of total generation fuel costinimization of total power lossinimization of total voltage deviationinimization of total generation fuel cost

  • 304 B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309

    Table 6Active power transit in branches: case 5: PD = 414 MW/loading margin stability = 1.6 p.u.

    Bold font with red color has been used to indicate the maximum limits of power transit in branches.

    Table 7The best results for optimized objective functions: IEEE 14-Bus: PD = 414 MW: LF = 1.6 p.u.

    Bold font with blue color has been used to indicate best results of the proposed strategy.

    Table 8Technical admissible parameters of generators and the fuel cost coefcients: IEEE 30-Bus.

    Bold font with red color has been used to indicate the limits of control variables.

    Fig. 20. Convergence characteristic for fuel cost minimization at the rst stage: global search achieved using FFA.

  • B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309 305

    Table 9Optimal setting of control variables: cases 13.

    Bold font withBold font with

    Fig. 21. Compintermediary s

    The totaing total po1.9142 p.u. compared tcompared terature. To mechanismrepresentedstage the toout consideconsideringbased on neintermediarstage, the tovalue found blue color has been used to indicate best results of the proposed strategy. red color has been used to indicate the limits of control variables.

    arison of convergence characteristic for fuel cost minimization attage.

    l fuel cost achieved is 798.9857 $/h, the correspond-wer loss and voltage deviation are: 8.4855 MW, andrespectively. The fuel cost achieved is highly reducedo the base case (power ow solution), and also improvedo many recent optimization methods cited in the lit-illustrate the principle of the proposed new combined

    search, the convergence characteristic of the fuel cost is in three stages as well shown in Figs. 2022, in the rsttal fuel cost optimized is 801.4326 $/h using FFA with-ring the brainstorming process, at specied trail and

    the practical rules generated by brainstorming rulesw updated database, the total fuel cost achieved at thisy stage is reduced to 800.6193 $/h, nally at the lasttal fuel cost optimized is reduced to 798.9857 $/h, this

    using local search based PS algorithm. It is important

    Fig. 22. Convesearch achieve

    to clarify thconstraintssecurity vaables foundoptimizatioother recenTable 10.

    Case2: MThe tota

    tive functiocorrespondacceptable.total activesumer withrgence characteristic for fuel cost minimization at nal stage: locald using PS.

    at the optimal fuel cost achieved without violation of, all security state and control variables are within theirlues. Table 9 shows results of optimized control vari-

    for three cases compared to the base case (withoutn). Comparison results of the proposed approach witht metaheuristic optimization methods are tabulated in

    inimization of total power lossl cost and the total power losses are two conict objec-ns. In the rst case, the total power losses generateding to the minimization of total fuel cost may not be

    There for the goal of this second case is to minimize the power loss by considering delivering energy to con-in the security limits. The convergence characteristic of

  • 306 B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309

    Table 10Comparison results of the proposed approach with other metaheuristic optimization methods: IEEE 30-Bus: cases 13.

    Bold font with blue color has been used to indicate best results of the proposed strategy.Bold font with red color has been used to indicate infeasible results.

  • B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309 307

    Fig. 23. Convenear global sea

    Fig. 24. Convestage.

    Fig. 25. Convesearch achieve

    the total poFigs. 2325.from the in

    The bestis 2.9162 Merences. Alssecurity conwith other Table 10.rgence characteristic for power loss minimization at the rst stage:rch achieved using FFA.

    rgence characteristic for power loss minimization at intermediary

    rgence characteristic for power loss minimization at nal stage: locald using PS.

    wer loss is represented in three stages, as well shown in As we can see the total power loss optimized is reduceditial stage to the last stage.

    total power loss achieved using the proposed strategyW which is better than many results cited in recent ref-o, the total power loss optimized without violation anystraints. Comparison results of the proposed approachmetaheuristic optimization methods are tabulated in

    Fig. 26. Convestage: near glo

    Fig. 27. Convediary stage.

    Fig. 28. Convstage: local se

    Case2: MVoltage

    ideal value case is the iload buses bthe ideal de0.1114 p.u.,convergencsented in ththe total vorgence characteristic for voltage deviation minimization at the rstbal search achieved using FFA.

    rgence characteristic for voltage deviation minimization at interme-

    ergence characteristic for voltage deviation minimization at nalarch achieved using PS.

    inimization of total voltage deviationproles at load buses must be within a desired value, theto be satised is 1 p.u. Therefore, the goal of this thirdmprovement of the distribution of voltage proles at ally minimizing the voltage deviation of PQ buses aroundsired value. The best total voltage deviation achieved is

    which is better compared to other recent methods. Thee characteristic of the total voltage deviation is repre-ree stages, as well shown in Figs. 2628. As we can seeltage deviation optimized is improved during the search

  • 308 B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309

    Fig. 29. Distristability = 1 p.u

    process. It ivoltage devWe can notimproved a

    8. Verica

    In the resolution qua category affect the foptimizatiopublished rtopology ofwell knownby injectingthe proposeNewtonRaactive powstraints of ctransformepensator anon details ocorrectionsin Table 10.

    9. Discussi

    The follospective of

    The optimcorresponof solutio

    The exibmodify thof the pro

    The braingence andspecied search sppractical

    The user cing the prmethods solved. It posed alg

    also important to clarify that there are no parameters to adjust,which considered as an advantage, so the brainstorming processvery suitable to working on parallel and in coordination withone or more metaheuristic methods. The formulation of rules

    nds to beis stution ct tom atropo

    betwhancropo

    orkin rela

    nclu

    ew pd PSto mion cned ads arn usng practitntroltly t

    of pf cosstiga

    loss y enng st4-Buserveic plowe

    stabpracttionng pos valstra

    wled

    authbution of optimized voltages at all buses: cases 1-2-3: Loading margin.

    s also important to note that minimization of the totaliation is achieved considering all security constraints.ice from Fig. 29, that the voltage prole has been greatlyround the desired value (1 p.u.).

    tion of limit violations

    cent literature many papers proposed to enhance theality of practical security OPF problems, unfortunatelyof results published recently are not well veried, thiseasibility, efciency and robustness of the proposedn methods. In this study the quality of some resultsecently are checked and veried considering the same

    electrical network and their associated constraints. As, the verication methodology is simple and it consists

    the optimal decision vector (vector control) found byd methods to the standard power ow program basedphson algorithm to determine all state variables, theer of slack generator and to verify the violation con-ontrol variables such as voltages of generating units, tapr, reactive power injected by conventional shunt com-d susceptance of specied SVC. In this section and basedptimal results given by in different papers, remarks and

    regarding the quality of certain methods are presented

    ons and perspectives

    wings points summarize the particularity and the per-the proposed practical planning strategy:

    ized results found for the two practical power system

    depelem

    In thdinasubjesyste

    The ption to en

    The pto wlems

    10. Co

    A nFFA anposed deviatcombimethosolutiostormiand prand coefcienin termterm ois invepowerstabilitplanniIEEE 1It is obdynamtotal pmarginposed applicaplannisuch athe con

    Ackno

    Theding to all objective functions are competitive in termn quality and convergence characteristic.ility of the proposed hybrid method allows expert toe algorithm based on the particularity and complexityblem.storm process proposed inspired from human intelli-

    reasoning rules introduced to assist the FFA and PS atstages to exploit and explore new feasible region in theace, these rules are not generalized and depends on theproblems to be solved and experiences of users.an easily replicate the proposed algorithm by introduc-oposed rules in coordination with other metaheuristicconsidering the particularity of the problem to beis important to not that the user can improve the pro-orithm by introducing also his own practical rules. It is

    editor and tand recommpaper. Real

    References

    [1] Belkacemunder co(2013) 53

    [2] H. DommSyst. PAS

    [3] R. Mota-penalty-f(1986) 31

    [4] R.C. Burchow, IEEE

    [5] D.I. Sun, Newton aon experience of user and the particularity of the prob- solved.dy only one SVC is investigated. Integration and coor-

    between many types of FACTS devices is an important enhance the loading margin stability of practical power

    critical situations.sed strategy is adaptable for integrating and coordina-een many types of FACTS devices such as TCSC and UPFCe the dynamic power system at critical situations.sed planning strategy is competitive and very suitableg in multi machine, so its ability to solving large prob-ted to power system planning and control.

    sion

    ractical power system planning strategy based hybrid algorithm assisted with brainstorming rules is pro-inimize total fuel cost, total power losses and voltageonsidering loading margin stability. In this proposedpproach the performances of the standard FFA and PSe improved by adjusting dynamically their sub-optimaling interactive reasoning rules inspired from brain-ocess. The proposed planning strategy allows to expertioner specialized in power system planning, operation

    to examine and achieve three important tasks: exploithe installation of FACTS devices, locate the weak regionsower quality, and identify critical production units int and security. In this study only one SVC controllerted and integrated at critical location to optimize totaland total voltage deviation considering loading marginhancement. The performance of the proposed practicalrategy has been tested with two practical test systemss and IEEE 30-Bus considering all security constraints.d from simulation results that the proposed combinedanning strategy is capable to reduce the total fuel cost,r loss and voltage deviation at an important loadingility. Due to the exibility and robustness of the pro-ical planning strategy, authors will strive to enlarge the

    of the proposed strategy for solving the large dynamicwer system considering practical security constraints

    ve point effect, prohibited zones, ramp rate limits andints associated to the integration of multi FACTS devices.

    gement

    ors gratefully acknowledge the Editor in chief, associatehe anonymous reviewers for their pertinent commentsendation to improve the quality and accuracy of our

    ly appreciated with great respect.

    Mahdad, K. Srairi, A study on multi-objective optimal power owntingency using differential evolution, J. Electr. Eng. Technol. 8 (1)63.el, W. Tinny, Optimal power ow solution, IEEE Trans. Power Appar.

    87 (10) (1968) 18661876.Palomino, V.H. Quintana, Sparse reactive power scheduling by aunction linear programming technique, IEEE Trans. Power Syst. 1 (3)39.ett, H.H. Happ, D.R. Vierath, Quadratically convergent optimal power

    Trans. Power Appar. Syst. 103 (11) (1984) 32673276.B. Ashley, B. Brewer, A. Hughes, W.F. Tinney, Optimal power ow bypproach, IEEE Trans. Power Appar. Syst. 103 (10) (1984) 28642875.

  • B. Mahdad, K. Srairi / Applied Soft Computing 35 (2015) 291309 309

    [6] A. Santos Jr., C.G.R.M. Da, Optimal power ow solution by Newtons methodapplied to an augmented Lagrangian function, IEE Proc. Gen. Trans. Distrib.142 (1) (1995) 3336.

    [7] X. Yan, V.H. Quantana, Improving an interior point based OPF by dynamicadjustments of step sizes and tolerances, IEEE Trans. Power Syst. 14 (2) (1999)709717.

    [8] S. Frank, I. Steponavice, S. Rebennak, Optimal power ow: a bibliographicsurvey I, formulations and deterministic methods, Int. J. Energy Syst. (Springer-Verlag) 3 (3) (2012) 221258.

    [9] J. Yuryevich, K.P. Wong, Evolutionary programming based optimal power owalgorithm, IEEE Trans. Power Syst. 14 (4) (1999) 12451250.

    [10] W. Ongsakul, T. Tantimaporn, Optimal power ow by improved evolutionaryprogramming, Electr. Power Co. Syst. 34 (1) (2006) 7995.

    [11] A.G. Bakirtzis, P.N. Biskas, C.E. Zournas, V. Petridis, Optimal power ow byenhanced genetic algorithm, IEEE Trans. Power Syst. 17 (2) (2002) 229236.

    [12] L.L. Lai, J.T. Ma, R. Yokoyama, M. Zhao, Improved genetic algorithms for optimalpower ow under normal and contingent operation states, Int. J. Electr. PowerEnergy Syst. 19 (5) (1997) 287292.

    [13] A.F. Attia, Y.A. Al-Turki, A.M. Abusorrah, Optimal power ow using adaptedgenetic a40 (11) (2

    [14] R.C. Eberhand resso

    [15] M.A. AbidPower En

    [16] R. Storn, Scheme fputer Scie

    [17] B. Mahdafor solutiConferen

    [18] M. Varadadifferenti

    [19] A.A. Abouevolution

    [20] N. Sinsuption usin2364237

    [21] A.J. Ghanow base(2011), Pa

    [22] S. Dumanitational 8795.

    [23] D. KaraboTechnicalEngineeri

    [24] M. Rezaeiobjective(2013) 21

    [25] X.S. Yang[26] X.-S. Yan

    convex eComput.

    [27] Y. Shi, Branational C303309.

    [28] B. Mahdadynamic in LMD sy

    [29] K. Vaisakwith non

    [30] A. Shabanbased opEnergy Co

    [31] K. Gnanamhybrid diPower En

    [32] Mahmouand contoptimizat(2014) 62

    [33] G. Chen, L. Liu, P. Song, Y. Du, Chaotic improved PSO-based multi-objectiveoptimization for minimization of power losses and L index in power systems,Int. J. Energy Convers. Manag. (Elsevier) 86 (2014) 548560.

    [34] M.J. Morshed, A. Asgharpour, Hybrid imperialist competitive-sequentialquadratic programming (HIC-SQP) algorithm for solving economic load dis-patch with incorporating stochastic wind power: a comparative study onheuristic optimization techniques, Int. J. Energy Convers. Manag. (Elsevier) 84(2014) 3040.

    [35] T. Niknam, M.R. Narimani, M. Jabbari, A.R. Malekpour, A modied shufe frogleaping algorithm for multi-objective optimal power ow, Energy 36 (11)(2011) 64206432.

    [36] M. Ghasemi, S. Ghavidel, M.M. Ghanbarian, H. Reza Massrur, E. Mahboubi-Moghaddam, Multi-objective optimal power ow using new modied andhybrid modied imperialist competitive algorithms: a comparative study, Inf.Sci. (Elsevier) 281 (2014) 225247.

    [37] A. Ananthi Christy, P. Ajay, D. Vimal Raj, Adaptive biogeography basedpredatorprey optimization technique for optimal power ow, Electr. PowerEnergy Syst. (Elsevier) 62 (2014) 344352.

    [38] K.R. Krishnanand, S.M.F. Hasani, B.K. Panigrahi, S.K. Panda, Optimal power owtion u

    orithmarbin,Alsume pow10) 17ahdag em. Hinglexiblcha, Cng andGitizaices u(2009ahda

    sideri89.enkatorithmer Enrinivam for

    load Rahmnt by terial 68.ahdaC to e322. Osbo

    Solviezaeem fortems, .E.H. Bching14) 49essel,ns. Poposto

    the 3.ahda

    ng Me, Intec

    IEEE merm

    simuahda

    ding cnag. (Elgorithm with adjusting population size, Electr. Power Compon. Syst.012).art, Y. Shi, Particle swarm optimization: developments, applicationsurces, in: Proc. 2001 Congr. Evolutionary Computation, vol. 1, 2001.o, Optimal power ow using particle swarm optimization, Electr.ergy Syst. 24 (2002) 563571.K. Price, Differential Evolution A Simple and Efcient Adaptiveor Global Optimization Over Continuous Spaces, International Com-nce Institute, Berkeley, CA, 1995, Tech. Rep. TR-95-012.d, K. Srairi, Differential evolution based dynamic decomposed strategyon of large practical economic dispatch, in: 10th EEEIC Internationalce on Environment and Electrical Engineering, Italy, 2011.rajan, K.S. Swarup, Solving multi-objective optimal power ow usingal evolution, IET Gener. Transm. Distrib. 2 (5) (2008) 720730.

    El Ela, M.A. Abido, S.R. Spea, Optimal power ow using differential algorithm, Electr. Power Syst. Res. 80 (2010) 878885.han, U. Leeton, T. Kulworawanichpong, Optimal power ow solu-g improved harmony search method, Appl. Soft Comput. 13 (2013)4.izadeh, G. Mokhtari1, M. Abedi, G.B. Gharehpetian, Optimal powerd on imperialist competitive algorithm, Int. Rev. Electr. Eng. 6 (June)pers Part B., U. Guvenc, Y. Sonmez, N. Yorukeren, Optimal power ow using grav-search algorithm, Int. J. Energy Convers. Manag. (Elsevier) 59 (2012)

    ga, An Idea Based on Honey Bee Swarm for Numerical Optimization, Report-TR06, Erciyes University of Engineering, Faculty of Computerng Department, 2005.

    Adaryani, A. Karami, Articial bee colony algorithm for solving multi- optimal power ow problem, Electr. Power Energy Syst. (Elsevier) 539230., Nature-Inspired Metaheuristic Algorithms, Luniver Press, UK, 2008.g, S.S.S. Hosseini, A.H. Gandomi, Firey algorithm for solving non-conomic dispatch problems with valve loading effect, Appl. Soft12 (2012) 11801186.in storm optimization algorithm, in: Proceedings of the Second Inter-onference, ICSI 2011, Part I, Chongqing, China, June 1215, 2011, pp.

    d, K. Srairi, Application of brainstorming strategy for enhancementteaching under LMD education system, in: First National Conferencestem, Proceeding Medea Conf., Algeria, 2007.h, L.R. Srinivas, Evolving ant direction differential evolution for OPF-smooth cost functions, Eng. Appl. Artif. Intell. 24 (2011) 426436.pour-Haghighi, A.R. Sei, T. Niknam, A modied teachinglearningtimization for multi-objective optimal power ow problem, Int. J.nvers. Manag. (Elsevier) 77 (2014) 597607.bal, C.K. babulal, Maximum loadability limit of power system using

    fferential evolution with particle swarm optimization, Int. J. Electr.ergy Syst. 43 (2012) 150155.d Joorabian, Ehsan Afsan, Optimal power ow under both normalingent operation conditions using the hybrid fuzzy particle swarmion and NedlerMead algorithm (HFPSO-NM), Appl. Soft Comput. 143633.

    solualgI, H

    [39] J.S. solv(20

    [40] B. Merin

    [41] N.Gof F

    [42] E. Aelli

    [43] M. dev50

    [44] B. Mcon75

    [45] B. ValgPow

    [46] B. Srithand

    [47] M. mebac57

    [48] B. MUPF315

    [49] A.Flem

    [50] A. Rrithsys

    [51] H.Rtea(20

    [52] P. KTra

    [53] T. Aing12

    [54] B. MUsiKG

    [55] The[56] Zim

    tem[57] B. M

    loaMasing self-evolving brain-storming inclusive teachinglearning-based, in: Proceedings of the 4th International Conference, ICSI 2013, Part

    China, June 1215, 2013, 2013, pp. 338345.ait, J.K. Sykulski, A.K. Al-Othman, A hybrid GAPSSQP method toer system valve-point economic dispatch problems, Appl. Energy 87731781.d, K. Srairi, Hierarchical adaptive PSO for multi-objective OPF consid-issions based shunt FACTS, in: IEEE Conference, 2012, pp. 13371343.orani, L. Gyugyi, Understanding FACTS, in: Concepts and Technologye AC Transmission Systems, IEEE Computer Society Press, 1999.. Fuerte-Esquivel, H. Ambiz-Perez, C. Angeles-Camacho, FACTS Mod-

    Simulation in Power Networks, John Wiley & Sons, 2004.deh, M. Kalantar, A novel approach for optimum allocation of FACTSsing multi-objective function, Int. J. Energy Convers. Manag. (Elsevier)) 682690.d, K. Srairi, Solving multi-objective optimal power ow problemng wind-STATCOM using differential evolution, Front. Energy 7 (2013)

    eswara Rao, G.V. Nagesh Kumar, Optimal power ow by BAT search for generation reallocation with unied power ow controller, Electr.ergy Syst. (Elsevier) 68 (2015) 8188.sa Raoa, K. Vaisakh, Multi-objective adaptive clonal selection algo-

    solving optimal power ow considering multi-type FACTS devicesuncertainty, Appl. Soft Comput. 23 (2014) 286297.at-Allah Hooshmand, J. Morshed, M. Parastegari, Congestion manage-determining optimal location of series FACTS devices using hybridforaging and NelderMead algorithm, Appl. Soft Comput. 28 (2015)

    d, T. Bouktir, K. Srairi, Strategy based PSO for dynamic control ofnhance power system security, J. Electr. Eng. Technol. 4 (3) (2009)

    .rn, Applied Imagination: Principles and Procedures of Creative Prob-ng, third revised ed., Charles Scribners Sons, New York, NY, 1963.

    Jordehi, Brainstorm optimisation algorithm (BSOA): an efcient algo- nding optimal location and setting of FACTS devices in electric powerElectr. Power Energy Syst. 69 (2015) 4857.ouchekara, M.A. Abido, M. Boucherma, Optimal power ow usinglearning-based optimization technique, Electr. Power Syst. Res. 11459.

    H. Glavitsch, Estimating the voltage stability of a power system, IEEEwer Deliv. 1 (3) (1986) 346354.lopoulos, A. Vlachos, Application of the rey algorithm for solv-economic emissions load dispatch problem, Int. J. Comb. (2011)

    d, Power Quality Management Based Security OPF Considering FACTStaheuristic Optimization Methods, InTech Education and Publishinghweb.org, 2013 (Chapter published in a Book (Power Quality 2)).Test System. http://www.ee.washington.edu/research/pstca/an RD, Murillo-Sanchez CE, Gan D. MATPOWER: a matlab power sys-lation package. http://www.pserc. cornell.edu/matpower/3.2d, K. Srairi, Multi objective large power system planning under severondition using learning DE-APSO-PS strategy, Int. J. Energy Convers.lsevier) 87 (2014) 338350.

    Security optimal power flow considering loading margin stability using hybrid FFAPS assisted with brainstorming rules1 Introduction2 Security optimal power flow2.1 Objective functions2.1.1 Minimization of total fuel cost2.1.2 Minimization of voltage deviation2.1.3 Minimization of power losses

    2.2 Constraints2.2.1 Equality constraints2.2.2 Inequality constraints2.2.2.1 Generator constraints2.2.2.2 Transformer constraints2.2.2.3 FACTS controllers constraints2.2.2.4 Security limits

    2.2.3 Constraints management

    2.3 Loading margin stability2.4 Voltage stability index

    3 Modeling of static VAR compensator (SVC)3.1 Optimal location strategy of SVC controllers

    4 Overview of FFA5 Overview of pattern search (PS)6 The proposed dynamic power system planning strategy6.1 Methodology of the proposed combined algorithm6.2 Coordination based on brainstorming process6.2.1 Overview6.2.2 Brainstorm optimization algorithm6.2.3 Proposed brainstorming rules

    7 Case studies and numerical results7.1 Test system 1: IEEE 14-Bus7.2 Test system 2: IEEE 30-Bus: without considering SVC controller

    8 Verification of limit violations9 Discussions and perspectives10 ConclusionAcknowledgementReferences