Comparison of different state models in direct torque control of induction machines operating without speed sensor

This paper reports on the application of different state models to the direct torque control (DTC) of induction machines operating without a speed sensor. A full-order and two reduced-order state models are developed. The estimation of internal system variables is performed by means of an original extended Kalman filter operating in two distinctive time frames allowing significant reduction of the calculation cost of the filter in the DTC framework. These “fast” and “slow” tasks are physically implemented on two standard processors operating in parallel. The complexity and, hence, cost of calculation is a first element to compare the different state models. Significant experimental results including slow speed reversals under load demonstrate the interesting performances obtained with each of the different models