Neural network based Q-MRAS method for speed estimation of linear induction motor

Abstract

The linear encoder for acquiring the speed information of the linear motor is costly and unreliable. Therefore, a sensorless control system without the speed sensor is designed in this paper. Based on the reactive-power-based model reference adaptive system (Q-MRAS) algorithm, the neural network is applied to replace the reference model in the traditional Q-MRAS algorithm. And the reference voltage instead of the measured voltage is used as the input. The proposed method is robust to noise due to the elimination of the current differentiation term in the traditional Q-MRAS algorithm, and it does not require a voltage sensor. Besides, the neural network is deployed in FPGA to achieve real-time calculation. Through the elaborate design of the hardware implementation, the fast calculation of the neural network is realized with less hardware resource consumption. The feasibility of the proposed method has been verified through simulation and hardware-in-loop tests.