Model Predictive Direct Torque Control of induction machines using a two-fold state approximation strategy

Abstract

This paper presents a novel method for adopting the concept of Model Predictive Control (MPC) in Direct Torque Control (DTC) of Electrical machines. The proposed algorithm enhances the performance of a DTC controller by keeping the motor's electromagnetic torque and stator flux magnitude within predefined hysteresis bounds while minimizing the switching power losses. The MPC controller predicts the output trajectories using an explicit model of the drive. A two-fold state approximation policy limits the quantity of the admissible outputs in drawing the tree of feasible trajectories over the prediction horizon. The chains of switching sequences and relevant switching losses are identified and a dynamic programming algorithm chooses the chain of switching sequences that minimizes a cost function on power losses in the inverter. Using receding horizon policy, only the first component of this chain is applied to the machine as the input signal at every sampling instant. The simulations are performed a small-sized induction motor-drive unit. The outcomes verify the advantages of this method in comparison with classic DTC.