Parameter estimation of induction machines from nameplate data using particle swarm optimization and genetic algorithm techniques

This article presents an optimization-based methodology to estimate the six equivalent circuit parameters of three-phase induction machines from its nameplate data for steady-state analysis. The optimization problem is based on minimizing the normalized square error between the computed performance of the equivalent circuit and that supplied by the manufacturer through the nameplate data. The problem is solved by using two routines that belong to the evolutionary computation family, namely, the particle swarm optimization (PSO) and the genetic algorithm (GA). A comparison between the functioning of the two routines is conducted. The motor performance computed through the PSO/GA parameters is compared to that computed by classical parameters obtained via machine testing, as well as the measured performance. Results show the superiority of the PSO/GA parameter set over the classical one, besides the distinct gain of eliminating the need to carry out lab testing in order to obtain the machine parameters.