Characterization of Stand Alone AC Generators During No-Break Power Transfer Using Radial Basis Networks

This paper describes the use of an artificial intelligence-electromagnetic modeling approach for the performance prediction of stand alone synchronous generators during no break power transfer (NBPT) operating conditions. This approach uses radial basis networks (RBNs), which have the advantage of not being locked into local minima as could do feedforward neural networks. The RBNs are simply linear function approximators that use radial basis functions which are powerful techniques for interpolation in multidimensional space. The RBN is used to evaluate the stresses accompanying this mode of operation which may result in the failure of the diodes in the rotating rectifier bridge of the generator brushless field exciter. The modeling approach is applied in a case study of two standalone synchronous generators system for aerospace applications. This study resulted in the prediction of the system performance characteristics including the peak currents and reverse voltages of the rotating diodes. The simulation results were validated by comparison to experimental data