چكيده به لاتين
Matrix Converters are emerging as an attractive alternatives for voltage source inverters due to their advantages such as the absence of DC link capacitor which makes their design more compact and increase the reliability, bidirectional power flow, sinusoidal input/output current and with controllable input power factor.
On the other hand, predictive control has been attracting growing interests in recent years because its concept and implementation are very intuitive and straightforward and the system constraints and nonlinearities can be simply involved in the control technique. However, the use of predictive control for matrix converters is computationally expensive because of the high numbers of the possible switching states of these converters. Since there is no modulation stage in predictive control its switching frequency is variable and to reach the similar performance of the modulation based control methods, a shorter control interval is needed. But, the high computational efforts of the predictive control id an obstacle for reducing the control time interval. Hence, the real time implementation of the predictive control requires powerful hardware which limits its industrial applications
In this dissertation the predictive control is implemented for a permanent magnet synchronous motor drive fed by a matrix converter. Since in a drive system different variables can be controlled, two conventional predictive control algorithms including predictive current control and predictive torque control are introduced and their experimental results are presented for comparison. Then, different strategies are proposed to reduce calculations in predictive control and their experimental results are presented to verify the validity of the proposed methods.
Furthermore, since the predictive control directly uses the system model to predict the behavior of the control objectives, the model parameter mismatch with their actual values results in inaccurate prediction and affects the performance of the algorithm. In this dissertation a simple and novel method is proposed to improve the performance if the predictive algorithm at the presence of parameter mismatch, without imposing high calculations.
