Neural network inverse control of current-fed induction motor

Abstract

The decoupling and linearize (D&L) control of induction motor is an important approach to improve the performance further. The analytical inverse system can realize D&L of nonlinear system when the model is exactly known, but for the induction motor with parameters varying and disturbance, the D&L is destroyed. So the neural network inverse system (NNIS) theory was adapted to approximate the analytical inverse system in order to weaken the couple of rotor flux and speed, the NNIS was designed for the induction motor in the synchronous rotating (dq) reference frame in this paper. Through the analytical inverse system expression we pointed out that the D&L effect is unrelated to the position of d axis. Subsequently, the neural network inverse control (NNIC) structure was proposed. As a special case, the NNIS of induction motor in rotor field oriented (MT) reference frame was also given, the comparison of this NNIC with direct rotor field oriented control (DRFOC) was done and we conclude that it is an improved method of DRFOC. At last, the simulation and experiment were done to test the proposed structures.