Predictive Direct Torque and Flux Control of an Induction Motor Drive Based on Multilevel Converter Approach

The paper presents a predictive direct flux and torque control of an induction machine based on finite states space model. The proposed control algorithm selects the switching state of the inverter that minimizes the quadratic error between torque and flux predictions to their computed values for all different voltage vectors. The optimal voltage vector that minimizes a cost function is then applied to the terminal of the induction machine. The proposed predictive control uses only one sample time and it is associated at first to a two-level voltage inverter then extended to a three-level converter. The strategy is very intuitive since it is very simple and provides best performances compared to other control laws. Induction motors are widely used in industrial applications due to their low maintenance, simplicity and relatively low cost compared to other machines. However, their dynamical model is multivariable, highly coupled, non-linear and the states are not all measurable for feedback control purposes. Therefore, they are more difficult to control than DC motors. The conventional hysteresis current controllers are generally used for their fast dynamic response, robustness and simplicity implementation by comparing measured load currents with references using hysteresis comparators and each comparator determines the switching state of the corresponding leg of the converter such that the load currents are forced to remain with the hysteresis band. In fact, these controllers ensure a good control of the current without necessity of having any knowledge of the system model or its parameters. Some drawbacks appears like variable switching frequency, thus, the current waveform contains numerous harmonics, causes resonance problems, switching losses restrict the application of this technique to lower power levels. To cope with these disadvantages, derived methods using a variable hysteresis bandwidth are used but this solution needs the knowledge of the system parameters and it is difficult to implement. PWM current control methods are very popular and are discussed extensively in the literature, their basic principle is to compare an isosceles triangle carrier wave with the fundamental frequency sinusoidal modulating wave generated by a modulator, and the points of intersection determine the switching points of power devices. The error between the reference and measured current is processed by a proportional-integral PI controller which assures zero steady state error for continuous reference but might fail for sinusoidal references causing undesired results where high performances are needed. Model Predictive Control (MPC) is a very powerful control strategy that uses the model of the system to precalculate the behavior of the system for a predefined horizon in the