Three-Phase Induction Motor Speed Estimation Using Recurrent Neural Network

Abstract

In induction motor speed control method, the development of the field-oriented control (FOC) algorithm which can control torque and flux separately enables the motor to replace many roles of DC motors. Induction motor speed control can be done by using a close loop system which requires a speed sensor. Referring to the speed sensor weaknesses such as less accurate of the measurement, this is due to the placement of the sensor system that is too far from the control system. Therefore, a speed sensorless method was developed which has various advantages. In this study, the speed sensorless method using an artificial neural network with recurrent neural network (RNN) as speed observer on three-phase induction motor has been discussed. The RNN can maintain steady-state conditions against a well-defined set point speed, so that the observer is able and will be suitable if applied as input control for the motor drives. In this work, the RNN has successfully estimated the rotor flux of the induction motor in MATLAB R2019a simulation as about 0.0004Wb. As based on speed estimation error, the estimator used has produced at about 26.77%, 8.7% and 6.1% for 150rad/s, 200rad/s and 250rad/s respectively. The future work can be developed and improved by creating a prototype system of the induction motor to get more accurate results in real-time of the proposed RNN observer.