Sensorless position estimation of switched reluctance motors using artificial neural networks

In this paper a model for sensorless position estimation of switched reluctance motor (SRM) is developed. This artificial neural network (ANN) based model is ultimately developed for nonlinear modeling of SRM. The nonlinear characteristics of SRM, which are mainly due to the magnetic saturation of the phase winding, are considered. This model is developed together with a set of measured data, which comprises of magnetization data for the SRM with flux linkage and phase currents as inputs and the corresponding rotor position as output. ANN forms a very efficient mapping structure for the nonlinear SRM. Given a sufficient large training data set, the ANN can build up a correlation between flux-linkages and rotor angle for an appropriate network architecture. The resultant model allows the determination of rotor estimation without any implementation of empirical equation to determine the unknown parameters in SRMs. This paper presents the development, implementation, operation and results of an ANN-based position estimator for any type of SRM.