Hybrid genetic algorithm for the identification of high-order synchronous machine two-axis equivalent circuits

Abstract

A hybrid genetic algorithm is used to find high-order equivalent circuits (ECs) of synchronous machines using standstill frequency response (SSFR) data. The algorithm performs satisfactorily despite the great deal of local minima surrounding the optimal solution of high-order ECs. It gives circuit parameters that simultaneously fit the three independent transfer functions given by the d-axis two-port network of the synchronous machine. It is found that as the order of the EC is increased, the optimization index used in the identification procedure is enhanced in a clear fashion. This leads to a new way for determining the right number of rotor branches required to correctly reproduce the SSFR data. The q-axis network is also analyzed with the hybrid algorithm. The so-called Canay's inductances are included in this one-port network to test if the fitting properties of the q-axis EC can be improved. The SSFR data used in this work is generated by a finite element model of a turbine generator, but actual data can also be readily handled.