ANewRepresentation inEvolutionary Algorithms fortheOptimization ofBioprocesses
Miguel Rocha, Isabel Rocha, Eugénio Campos Ferreira · 2005
Evolutionary Algorithms (EAs)havebeenusedto achieve optimal feedforward control inanumberoffedbatch fermentation processes. Typically, theoptimization purpose istosettheoptimal feeding trajectory, beingthefeeding profile overtimegiven byapiecewise linearfunction, inorder toreduce thenumberofparameters totheoptimization algorithm. Inthis work,anovel representation scheme forthe encoding ofthefeeding trajectory overtimeisproposed. Eachgeneinthevariable sized chromosome hastwo components: atimelabel andthereal value ofthevariable. Thenewapproach iscompared withatraditional real-valued EA,withchromosomes ofconstant size and fixed discretization steps. Threedistinct casestudies are presented, taken fromprevious workfromtheauthors andliterature, allconsidering theoptimization offedbatch fermentation processes. Theexperimental results showthattheproposed approach iscapable ofresults better oratthesamelevel ofquality ofthebesttraditional EAsandisable toautomatically evolve thebest discretization steps foreachcase, thussimplifying the