Quasi and fully sensorless speed control of indirect RFOC induction motor drives for low speed operation
For high performance drive applications, the accuracy of speed estimation from a digital shaft encoder is reduced in the low speed range. Drive performance can be further impaired at low speed when low cost low-resolution digital encoders are used. This thesis contributes to the development of a quasi-sensorless solution for AC induction motor drives, where instead of a conventional position sensor and MH" or closed loop observer, an artificial intelligence based 64PPR SKF sensor bearing is proposed. Furthermore, the performance of speed-controlled IM drives is sensitive to rotor time constant mismatch. A newly developed closed loop RMLE-GNA observer is proposed as a new adaption mechanism with a Model Reference Adaptive Scheme for rotor flux estimation and on-fine rotor time constant adaptation. For high performance sensorless speed controlled IM drive applications, the IM is required to operate at low speed in all four-quadrants in a stable manner whilst maintaining constant air-gap flux and torque. Such low speed operation fails with conventional open loop observers and MRAS, and closed loop observers such as EKF suffer from a tuning problem and extensive computing time. Parameter uncertainty is a further factor limiting drive performance. A new joint state estimation and parameter adaptation closed loop Recursive Maximum Likelihood Estimator with an iterative Gauss Newton Raphson Algorithm (RMLE-GNA) is developed for the sensorless speed estimation and on-line parameter adaptation of IM drives. Two case studies are presented: correct and stable convergence of the closed loop observer within two iteration local loops utilising a full order IM model, and correct and stable convergence within a single iteration local loop utilising a simplified full order IM model. In both cases it is shown that the closed loop observer can provide satisfactory and stable convergence for 4-quadrant operation. Of particular interest are the correct convergence, robustness against parameter mismatch and the stability of the sensorless speed controlled IM drive at zero speed and zero stator flux frequency. Thus, a small signal transfer function of the closed loop observer with the sensorless speed vector controlled IM drives is derived. The stability analysis shows the robustness and the convergence of the observer for 4-quadrant operation for high and low speed' and up to the rated load torque. The effect of various machine model parameters on the stability and convergence of the RMLE-GNA-based sensorless drive system is investigated, providing an insight into the inherent dynamic characteristics of the proposed closed loop observer-based sensorless speed-controlled IM drives.