Compensation of Parameters Variations in Induction Motor Drives using a Neural Network

In this paper, the possibility of using a neural network (NN) to compensate parameter variations in an indirect field oriented (IFO) controller is studied and presented. In particular, a three-layer NN has been designed and trained offline with a steady state mathematical model of an IFO control scheme in detuning operations. Thus, the trained NN has been added to the controller as a black box to compensate for motor parameters variations. The motor controller behaviour with the NN black box has been studied in tuning and detuning conditions. Complete simulation results for a 4.0 kW induction motor driven by a CRPWM inverter with IFO controller are shown and discussed.