Modeling and parameter identification of switched reluctance motors from operating data using neural networks

An appropriate model of switched reluctance motor is essential to its control implementation. A simple model of SRM consists of a nonlinear inductance and a resistance that can be estimated from standstill test data. However, this model may not represent the characteristics of SRM during operation because of the saturation and losses in phase windings and rotor body. To model this effect, one or two damper windings can be added into the model structure. This paper proposes an application of artificial neural network to identify the nonlinear model of SRMs from operating data. A recurrent neural network has been adopted to estimate the damper(s) currents from operating data. Then the damper(s) parameters can be identified using maximum likelihood estimation techniques.