Neural Network Model-based Direct Torque and Flux Predictor for Induction Motor Drive

Abstract

This paper describes the proposed neural network (NN) predictive control based on ant colony optimization (NNPC-ACO) implemented to a direct torque and flux controlled (DTFC) induction motor drive (IMD) using space vector modulation (SVM) technique and is compared with that of an ant colony optimized model predictive control (MPC-ACO) in order to show its superior performance. Since MPC has a major drawback in computational complexity due to iterative computations at each time step, NNPC has been introduced as a powerful control method for IMD. ACO technique has been implemented efficiently with the predictive controller to solve the nonlinear optimization problems with the consideration of system constraints like output and states. Further, the proposed NNPC-ACO for torque and flux controller must be tuned carefully to obtain satisfactory performance at different states of operation. Moreover, the complexity in NNPC-ACO is reduced as compared to MPCA-CO and therefore, the optimum generated control signal of NNPC-ACO gives better dynamic performance and settling time compared to the MPC-ACO. The model using both controllers is observed using MATLAB/Simulink software as well as a hardware prototype with a 3.7 kW IMD. The NNPC-ACO provides improved performance and better robust operation than that of the MPC-ACO-based drive.