Radial Basis Networks for the Simulation of Stand Alone AC Generators during No-Break Power Transfer

摘要

This paper describes the use of an ArtificialrnIntelligence-Electromagnetic modeling approach for thernperformance prediction of stand alone synchronous generatorsrnduring No Break Power Transfer, NBPT, operating conditions.rnThis approach uses Radial Basis Networks, RBN, which have thernadvantage of not being locked into local minima as dornfeedforward Neural Networks. The RBNs are simply linearrnfunction approximators that use Radial Basis Functions whichrnare powerful techniques for interpolation in multidimensionalrnspace. The RBN is used to evaluate the stresses accompanyingrnthis mode of operation which may result in the failure of therndiodes in the rotating rectifier bridge of the generator brushlessrnfield exciter. The modeling approach is applied in a case study ofrntwo standalone synchronous generators system for aerospacernapplications. This study resulted in the prediction of the systemrnperformance characteristics including the peak currents andrnreverse voltages of the rotating diodes. The simulation resultsrnwere validated by comparison to experimental data.