Experimental and simulation study of neural network enhanced direct torque control for induction motors

Because of their easy control, affordability, and high efficiency, induction machines (IMs) are frequently utilized in industrial settings. Robust control strategies are necessary to manage changes in speed, machine parameters, and external disturbances in order to achieve optimal dynamic performance. One of the most widely used techniques for controlling instant messaging is direct torque control (DTC), which is renowned for its quick dynamic response and simplicity of use. Six or twelve sectors are used in traditional DTC strategies of IMs. Despite their effectiveness in dynamic response, they frequently result in high torque and current ripples, particularly when parameters are varied. To address these problems, this paper suggests improved DTC strategies that make use of neural network (NN) techniques. To increase control accuracy and system performance, the modified 6-sector and 12-sector DTC approaches for IM drives use a trained NN technique in place of the traditional switching table. Before being put into practice experimentally using a dSPACE 1104 platform, these strategies were initially created and verified in MATLAB through simulation tests. The simulation results are validated by the experimental results, which also demonstrate the effectiveness of the NN technique-based DTC strategy in improving torque quality and dynamic speed response compared to the traditional DTC technique. The best overall performance was attained by the 12-sector NN technique-based DTC technique.