Development of smooth torque in switched reluctance motors using self-learning techniques

Control of the switched or variable reluctance motor is complicated by its nonlinear nature, caused by its doubly salient construction and by the fact that it is designed to operate in magnetic saturation. In this paper, the usually laborious characterisation and commissioning procedures are replaced by an offline, self-learning algorithm. Numerical approximations are made to the nonlinear flux/current/position and torque/current/position characteristics of the reluctance motor by determining the coefficients of approximating functions that model such characteristics. Use of this model allows fast, efficient, online torque control suitable for use as an inner loop in velocity and position controllers. Knowledge of the flux-linkage characteristics assimilated by the controller permits the use of feedforward terms to compensate for undesirable nonlinearities in the motor