Improving Torque and Speed Estimation Accuracy by Conjoint Parameter Identification and Unscented Kalman Filter Design for Induction Machines

Any model-based control of induction machines (IM) is sensitive to incorrect model parameters and measurement errors when utilized in an open-loop torque or speed-sensorless drive. Thus, various publications are available which address the IM's parameter identification as well as the flux- and speed-observer design. Classically, this is handled as a separated two-step procedure. However, measurement uncertainty negatively affects the IM identification result and, consequently, the uncertainty regarding the IM's parameter is then adversely affecting the observer design. In contrast to that, this contribution is proposing a conjoint observer design and machine identification process by one global optimization. As the observer tuning and machine identification have the same objective, i.e. being able to control a given drive with highest speed and torque accuracy, there is no compelling reason to separate them from each other. Hence, more degrees of freedom are given to the optimization process. Whereas an unscented Kalman filter (UKF) is applied as a flux- and speed-observer, sensor noise is taken into account due its internal working principle. Experimental test bench evaluations prove the feasibility of the proposed method, however, the achievable estimation accuracy is worse compared to a similar approach using an extended Kalman filter instead of the UKF.