Low Switching Frequency Model Predictive Control of Three-Level Inverter-Fed IM Drives With Speed-Sensorless and Field-Weakening Operations

This paper proposes a single-vector-based model predictive control (MPC) for an induction motor drive supplied by a three-level neutral-point-clamped inverter operating at a low switching frequency. Unlike the torque and flux control in the conventional MPC, the proposed MPC tries to minimize the error between the applied voltage vector and a reference voltage vector obtained based on the principle of deadbeat control, thereby reducing the number of weighting factors in the cost function. To reduce the computational burden and restrict the high jumps in both phase and line voltages, two switching tables are proposed and compared for the preselection of candidate voltage vectors. The neutral point potential fluctuation and switching frequency are also included in the cost function to achieve the balance between the upper and lower dc voltages and a relatively low switching frequency. Furthermore, a speed adaptive stator flux observer with a novel gain matrix is proposed to achieve speed-sensorless operation, which has a higher speed and flux estimation accuracy than conventional fixed gains. Finally, the proposed MPC is extended to the field-weakening operation by adjusting the torque and stator flux references online, which significantly widens the speed range and improves the practical value of the MPC. The effectiveness of the proposed method is confirmed by the experimental results obtained at an average switching frequency of less than 600 Hz.