Artificial neural network sensorless direct torque control of two parallel-connected five-phase induction machines

Abstract

induction Conventional machine direct (FPIM). torque Nevertheless, control (DTC) it improves suffers from the significant dynamic performance drawbacks of of high the five-phase stator flux and electromagnetic torque ripples. Moreover, the DTC technique relies on an open-loop estimator for accurate stator flux module and position knowledge. However, this method is subjected to substandard performance, mainly during the low-speed operation range. Therefore, a sliding mode sensorless stator flux and rotor speed DTC based on an artificial neural network (DTC-ANN) for two parallel-connected FPIMs is discussed to tackle the problems above. This approach optimizes the DTC performance by replacing the two hysteresis controllers (HC) and the look-up table. As for the poor estimation drawback, the sliding mode observer (SMO) offers a robust estimation and reconstruction of the FPIM variables and eliminates the need for additional sensors, increasing the system's reliability. The present results verify and compare the performance of the control scheme.