Identification ofInduction Machines Stator Currents withGeneralized Neurons
Jing Huang · 2007
g Abstract -A newapproach toidentify thenonlinear modelofan induction machineusingtwogeneralized neurons(GNs)is presented inthis paper. Compared tothemultilayer perceptron feedforward neural network, a GN hassimpler structure and lesser requirement intermsofmemorystorage whichismakesit attractive forhardware implementation. Thismethodshowsthat withless numberofweights, GN isabletolearn thedynamics of an induction machine. Theproposed modelismadebytwo coupled networks. A modified particle swarmoptimization algorithm isdesigned tosolve this distinctive GNtraining problem. Apseudo-random binary sequence signal injected totheinduction machine operating atitsrated value waschosen asthetest input signal. Forvalidation, thetrained GN modelisapplied onthe different operating conditions ofthesystem.