I. Introduction
The induction motor drive fed by a matrix converter is superior to the conventional inverter because it does not have bulky dc-link capacitors with limited lifetime and offers bidirectional power flow capability, sinusoidal input/output currents, and an adjustable input power factor. Furthermore, because of the high integration, the matrix converter topology is recommended for extreme temperatures and critical volume/weight applications [1], [2]. The direct torque control (DTC) scheme for matrix converter drives was initially presented in [3]. The generation of the voltage vectors required to implement the DTC of induction motors under a unity input power factor constraint was allowed. However, the DTC scheme using a switching table has some fatal drawbacks. Switching frequency varies according to the motor speed and the hysteresis bands of torque and flux, a large torque ripple is generated in a low speed range because of the small back electromotive force (EMF) of the induction motor, and high control sampling time is required to achieve good performance [4]. Although several methods to solve these problems have been presented [5]–[7], these methods are designed for a conventional inverter drive system. On the contrary, research results to solve these problems for matrix converter drives have not been reported in the literature.