Sensorless Position Control of Switched Reluctance Motors Based On Artificial Neural Networks

This paper presents a new artificial neural network method for position control of switched reluctance motors, in this method the flux observer is not needed and the data sources used for training are voltages and currents of two phases, so the designed neural network has four inputs and one output which is the rotor position. The advantage of this method is its low torque ripple in comparison with shaft encoder observers and also because of not having any flux observer, any uncertainties of phase resistance does not affect training process. In this paper the development and operation of an ANN-based position estimator for a three-phase SRM is presented. The experimental process has been implemented on a 6/4, 4 kW SRM