High-Performance Computing Using FPGAs for Improving the DTC Performances of Induction Motors

Introduction

Recently, the Direct Torque Control (DTC) method can be considered as the most used control technique for electrical motors (Lin et al, 2006). This type of torque and flux control was firstly proposed as direct self control by Depenbrock (1988) and direct torque control by Takahashi (1986). This control strategy presents an outstanding dynamic performance as well as good robustness under variations in the motor parameters. It seems well suited to applications in traction or electric vehicle. However, this control strategy presents many disadvantages, such as the ripples in the torque and flux, and the distortion in the stator current waves; due to the presence of the hysteresis comparators used to control the torque and the stator flux (Luis et al, 2003; Deng et al, n.d). In fact, the basic structure of the conventional DTC was developed by Takahashi in 1986 and integrates two hysteresis controllers in order to control the torque and the flux in an independent way. The ripples of the torque and flux are limited by the bands of the hysteresis controllers in order to satisfy the torque and flux demand. Yet, in practice, it is difficult to guarantee this operation conditions due to the time delay between the sampling instant of the torque which follows a discrete computation approach and the located time to calculate the inverter switching states (Jidin et al, 2010). In this case, the torque ripples surpass the hysteresis bands, which leads to select a voltage vector causing a rapid increase or decrease of the torque (Idris et al, 2004). This consequently increases the torque ripples, the stator current distortions and degrades the conventional DTC performances. To overcome these problems, novel DTC structures have been developed, like the combination between the DTC and the Space Vector Modulation (SVM), this novel structure is known by DTC-SVM (Kyo-Beum et al, 2008). The DTC-SVM reduces considerably the ripples and guarantees an operation with fixed switching frequency. Generally, the DTC-SVM uses Proportional-Integral (PI) controllers to control the torque and the rotor speed. The gains of the PI controllers are estimated with the parameters of the motor model which cannot work well in practice and affect the robustness of the control strategy. Other researchers have replaced the PI controllers by Sliding Mode Controllers (SMCs) to improve the robustness of the control strategy. Hence, the combination of the DTC-SVM with SMCs increases the algorithm complexity (Saber et al, 2017). Alternatively, other DTC structure based on predictive controller has been developed by (Zhu et al, 2012; Preind et al, 2013), deadbeat direct torque control (Xu et al, 2014), duty ration control (Xia et al, 2014) and root mean square criteria (Shyu et al, 2010). The cited control techniques reduce the torque ripples and improve the DTC performances, but the algorithm uses more parameters of the motor and it is more complex. In paper (Singh et al, 2013), the authors propose a DTC with a modified switching table. Thus, this method cannot be considered as a definitive solution to prevent the flux ripples caused by sector exchange. In paper (Zhang et al, 2012; Geyer et al, 2012), the authors have used multilevel inverters with more levels of the hysteresis controllers and more sectors number in order to select an appropriate voltage vector in each sampling time. This method reduces the torque ripples, but the multilevel inverter increases the control system cost and the commutation losses. Another method is based on the intelligent technique like the neural networks and the fuzzy logic, but in this method the switching frequency remains variable and does not provide a good performance in terms of torque and flux ripples (Uddin et al, 2012; Abbou et al, 2009) Moreover, the conventional DTC offers good performances in terms of ripples reduction if the sampling frequency is higher which require processors with high processing speed.