Neural stator flux estimator with dynamical signal preprocessing

This work presents a novel method of stator flux estimation for induction motor, using artificial neural networks. Proposed estimation scheme can be employed in any vector controlled drive, e.g. DTC (direct torque control) or SFOC (stator field oriented control) drive. It does not exploit pure integration, therefore there is no problem with drift and initial conditions. Moreover, it does not require any stator resistance adaptation algorithm. A multilayer perceptron is trained off-line to approximate stator flux space vector. The approximation space is spanned by stator voltages and currents preprocessed with different low-pass filters. The performance of the SFOC drive with the above stator flux estimator has been investigated in simulation. The results seem to be very promising. Developed estimator is quite insensitive to the stator resistance variations.