Torque and speed control of induction motors using ANN observers

Two types of observers (based on the linear and nonlinear model of the machine) have been used in torque and speed control of induction motor control schemes [K. B. Nordin et al. (1985), A. Bellini et al. (1985), D. S. Wijesundera et al. (1992)]. The reduced-order linear model of observer [D. S. Wijesundera et al. (1992)] is easy to implement in real time, but it does not give an accurate estimation of the rotor m.m.f. vector angle, β, since the induction motor normally operates in the region of saturation. The nonlinear observer model which incorporates this effect of magnetic saturation of the induction motor cannot be practically implemented by using normal methods as it takes too long a time to estimate the angle β. The implementation of the real-time torque/speed controller discussed in this paper is based on artificial neural networks (ANN) which take into account the effect of saturation and estimate the angle β in a few microseconds which is well within the real time deadline.