Search

During the last two decades there has been considerable development of sensorless vector controlled induction motor drives for high performance industrial applications. Such control strategies reduce the drive's cost, size and maintenance requirements while increasing the system's reliability and robustness. Parameter sensitivity, high computational effort and instability at low and zero speed can be the main shortcomings of sensorless control. Sensorless drives have been successfully applied for medium and high speed operation, but low and zero speed operation is still a critical problem. Much recent research effort is focused on extending the operating region of sensorless drives near zero stator frequency. Several strategies have been proposed for rotor speed estimation in sensorless induction motor drives based on the machine fundamental excitation model. Among these techniques Model Reference Adaptive Systems (MRAS) schemes are the most common strategies employed due to their relative simplicity and low computational effort. Rotor flux-MRAS is the most popular MRAS strategy and significant attempts have been made to improve the performance of this scheme at low speed. Artificial Intelligence (AI) techniques have attracted much attention in the past few years as powerful tools to solve many control problems. Common AI strategies include neural networks, fuzzy logic and genetic algorithms. The mam purpose of this work is to show that AI can be used to improve the sensorless performance of the well-established MRAS observers in the critical low and zero speed region of operation. This thesis proposes various novel methods based on AI combined with MRAS observers. These methods have been implemented via simulation but also on an experimental drive based around a commercial induction machine. Detailed simulations and experimental tests are carried out to investigate the performance of the proposed schemes when compared to the conventional rotor fluxMRAS. Various schemes are implemented and tested in real time using a 7.5 kW induction machine and a dSP ACE DS 1103 controller board. The results presented for these new schemes show the great improvement in the performance of the MRAS observer in both open loop and sensorless modes of operation at low and zero speed.