Development and experimental realization of an adaptive neural-based discrete model predictive direct torque and flux controller for induction motor drive

Abstract This paper develops a neural network-based discrete predictive direct torque and flux control (NNPDTFC) for induction motor drive with the space vector modulation (SVM) technique. Moreover, this SVM technique with NNPDTFC is implemented to activate the inverter in the two-level operation and the performance is compared with the conventional PI direct torque and flux control (PIDTFC) technique. The PSO based model predictive control incorporated with the neural network is developed here in the NNPDTFC and is analyzed using MATLAB software. Disturbance reduction, simple control, and real-time implementation are the major features of NNPDTFC and it also enhances the transient performance of the motor drive by reducing settling time and peak overshoot. In addition, the flux and the torque ripples are significantly improved using the proposed NNPDTFC technique which is extensively used for the fast dynamic response of the induction motor drives as compared to PIDTFC. In order to show the potentiality of the proposed controller, a prototype controller is developed and validated with the laboratory setup and the control signals are generated for both NNPDTFC and PIDTFC using a low-cost Digital signal processor (DSP) controller which is fed to the induction motor of 3.7 kW capacity in the real-time platform. It is observed that the results with NNPDTFC are not only found to be extremely satisfactory even with the system and parameter uncertainties and external load perturbations but also, it produces enhanced dynamic as well as steady-state performance along with the reduced ripples in the signal flux, torque, and current compared to that of PIDTFC.