Abstract:
This study proposes new model-based loss-minimization technique (LMT) for rotor field-oriented control (FOC) in induction machine drives. Compared to existing LMT applied for FOC, the suggested technique incorporates the cross-coupling effects of core losses into induction-machine modeling. The comparison is examined for several induction machines characteristics in order to generalize the functionality of the proposed LMT method. Moreover, based on the weaknesses observed in the conventional and the proposed LMT such as motor stability and machine flux limitations, another loss-minimization technique termed the hybrid LMT is introduced to overcome these issues. In addition, as model-based LMT require the knowledge of motor speed and torque values, this study therefore suggests to use nonlinear Kalman filter as parameters estimator , to adapt these functions; specifically, the square-root extended Kalman filter (EKF), square-root unscented Kalman filters (SRUKF), and square-root cubature Kalman filters (SRCKF). Moreover, to evaluate the filters performance, the study shows a comparison between the nonlinear filter types for Sensorless machine applications. The comparison discusses the machine model uncertainty by conducting a sensitivity analysis of filters estimates against varying noise covariances. The state-space model of induction motor varies due to nonlinearities, saturation and harmonic. Thus, this dissertation explores this variety and establishes that the use of single-model Kalman filter is not sufficient. Therefore, an interactive multiple mode (IMM) filter is used and compared with the single-model filters. Considering the loss minimization, the experimental results show a superior performance of proposed LMT as compared to traditional ones. As for the estimation issue, the experimental results demonstrate a higher performance of multiple model estimation against single model version, with emphasis on operating speed and torque range, estimation error, and stability.