4.Fundamentalsofdifferentialevolutionapproaches
ThissectiondescribestheproposedDECZalgorithm.First,abriefoverviewoftheDEisprovided,and nallytheproposedDECZalgorithmispresented.
4.1.Differentialevolutionalgorithm
DEisapopulation-basedstochasticfunctionminimizer(ormaximizer)relatingtoevolutionaryalgorithms(EAs),whosesim-pleyetpowerfulandstraightforwardfeaturesmakeitattractivefornumericaloptimization.ClassicalDEalgorithmsusea oat-ing-pointencodingEAforglobaloptimizationovercontinuousspaces.LikeotherEAs,DEisaglobaloptimizercapableofworkingreliablyinnon-linearandmultimodalenvironments.Usingafewparameters,DEexhibitsanoverallexcellentperformanceforawiderangeofbenchmarkfunctions.TheadvantagesofDE,suchasasimple,compactstructure,easeofuse,highconvergencechar-acteristics,androbustness,makeitahigh-classtechniqueforreal-valuedparameteroptimization.
DEcombinessimplearithmeticaloperatorswiththeoperatorsofrecombination,mutationandselectiontoevolvefromaran-domlygeneratedstartingpopulationtoa nalsolution.DEuses
mutationwhichisbasedonthedistributionofsolutionsinthecur-rentpopulation.Inthisway,searchdirectionsandpossiblestepsizesdependonthelocationoftheindividualsselectedtocalculatethemutationvalues.Itevolutesgenerationbygenerationuntiltheterminationconditionshavebeenmet.
ThedifferentvariantsofDEareclassi edusingthefollowingnotation:DE/a/b/d,whereaindicatesthemethodforselectingtheparentchromosomethatwillformthebaseofthemutatedvec-tor,bindicatesthenumberofdifferencevectorsusedtoperturbthebasechromosome,anddindicatestherecombinationmecha-nismusedtocreatetheoffspringpopulation.Thebinacronymindi-catesthattherecombinationiscontrolledbyaseriesofindependentbinomialexperiments.
ThefundamentalideabehindDEisaschemewherebyitgener-atesthetrialparametervectors.Ineachstep,theDEmutatesvec-torsbyaddingweighted,randomvectordifferentialstothem.Ifthecostofthetrialvectorisbetterthanthatofthetarget,thetargetvectorisreplacedbythetrialvectorinthenextgeneration.ThevariantimplementedherewasDE/rand/1/bin,whichinvolvedthefollowingstepsandprocedures:
Step1:Initializationoftheparametersetup:Theusermustchoosethekeyparametersthatcontrolthealgorithm,i.e.,pop-ulationsize(N),boundaryconstraintsofoptimization

variables,
Multivariable control ... 暂无评价 15页 免费 ...CHOI AND LEE: ANTIWINDUP STRATEGY FOR PI-TYPE ..., PID Controllers: Theory, Design, and Tuning. ...
(GA) for tuning proportional–integral–derivative (PID) controllers to ...6. Conclusions A multivariable control strategy for simultaneously controlling ...
(GA) for tuning proportional–integral–derivative (PID) controllers to ...6. Conclusions A multivariable control strategy for simultaneously controlling ...
热门文档