Sensorless switched reflectance motor drive with torque ripple minimization

Position sensorless torque ripple minimization techniques are presented to deal with the issues of rotor position sensor requirement and high torque ripple production in a switched reluctance motor (SRM) drive. In the proposed methods, multilayer perceptron (MLP) neural networks have been applied to learn the nonlinear electrical characteristics of an SRM. The nonlinear model of an SRM is used in the simulation which takes into account the magnetisation saturation effect. The model is verified with experimental flux linkage, inductance and torque data taken from a 7.5 kW SRM. Simulation results have shown that torque ripple minimization can be achieved without a rotor position sensor or torque sensor. Experimental work has been undertaken to show the effectiveness of the torque prediction by the neural network