Neural network based estimation of feedback signals for a vectorcontrolled induction motor drive

摘要

Neural networks are is recently showing good promise forapplication in power electronics and motion control systems. So far,they have been applied to a few cases, mainly in the control of powerconverters and drives, but their application in estimation ispractically new. This paper explores the application of neural networksfor estimation of feedback signals in induction motor drive systems. Afeedforward neural network receives the machine terminal signals at theinput and calculates flux, torque and unit vectors (cos θ_e and sin θ_e) at the output which are then usedin the control of a direct vector-controlled drive system. Thethree-layer network has been trained extensively by the NeuralWorksProfessional II/Plus program to emulate the DSP-based computationalcharacteristics. The performance of the estimator is good and iscomparable to that of DSP-based estimation. The drive system has beenoperated in wide torque and speed regions independently with a DSP-basedestimator and a neural network-based estimator, and are shown to havecomparable performance. The neural network estimator has the advantagesof faster execution speed, harmonic ripple immunity and fault tolerancecharacteristics compared to the DSP-based estimator