Enhancing artificial neural network-based direct torque control of a double-fed induction generator

Direct Torque Control finds widespread industrial use due to its advantages; however, torque ripples hinder efficiency because of variable frequency operation caused by hysteresis comparators. Additionally, finite frequency sampling leads to erratic overshoot within the hysteresis band, exacerbated by low-speed operation and generator resistance variations. To tackle these challenges, this study introduces an intelligent direct torque control method for DFIGs, employing a three-level inverter and an ANN to replace switching tables and hysteresis comparators. MATLAB/Simulink simulations reveal substantial enhancements in DFIG performance, with a notable 42.16% reduction in torque ripples and complete elimination of overshoot. However, a slight reduction in the total harmonic distortion is observed, with rotor and stator currents experiencing respective reductions of 0.21% and 0.19%.