ANN and MRAS based Speed Observer for Sensorless Control of Induction Motor Drive
S. Damkhi, MS.Nait-Said, N. Nait-Said 1. Department of Electrical Engineering Batna, Batna University Central Campus–1, Batna (Algeria). 2026-06-27

Abstract

The Model Reference Adaptive System (MRAS) is indeed the most widely used strategy for speed estimation in
sensorless induction motor drives, thanks to its simplicity and satisfactory performance across a wide speed range. However, this technique has limitations: it is not optimal in medium- and high-speed regions and remains highly sensitive to variations in motor parameters. Therefore, the artificial Neural Network (ANN)-based MRAS scheme is proposed in this paper it replaces the conventional adjustable model in the MRAS structure, using a two-layer Neural Network (NN) rotor flux observer trained online via a backpropagation (BP) algorithm. The error between the reference model and the NN-based adaptive model adjusts the network weights dynamically for robust speed estimation.Speed estimation performance of two different rotor speed observers is studied and compared when applied to a sensorless direct vector control induction motor drive. The ANN-MRAS observer delivers superior response during steady-state operation and at very low speeds, effectively minimizing speed pulsations.

Key words: MRAS, Induction Motor, Neural Network, Sensorless control

References

[1]. J. Holtz, “Sensorless control of IM drives,” IEEE Transactions on Industrial Electronics, vol. 90, no. 8, pp. 1359–1394, Aug. 2002.

[2]. D. Xu, B. Wang, G. Zhang, G. Wang, and Y. Yu, “A review of sensorless control methods for AC motor drives,” CES Transactions on Electrical Machines and Systems, vol. 2, no. 1, pp. 104–115, Mar. 2018

[3].Mohan, H.; Pathak, M.K.; Dwivedi, S.K. Sensorless control of electric drives—A technological review. IETE Tech. Rev. 2020, 37, 504–528.

[4]. P. Kumar, D. Deshpande and M. Dubey, “Model Reference Adaptive system Based Speed Sensorless Vector Control of Induction Motor Drive”, International Journal on Recent and Innovation trends in Computing and Communication, Vol. 2, Issue 6, pp. 1554-1559, 2014

[5]. M. Munshi, S. G. Choudhuri, and S. Member, “Model Reference Adaptive System using Rotor Flux and Back Emf Techniques for Speed Estimation of an Induction Motor operated in Vector Control Mode : A Comparative Study,” pp. 44–49, 2016.

[6]. L. Sbita, D. Zaltni, M.N.Abdelkrim, ”Adaptive Variable Structure Control for an Online Tuning Direct Vector Controlled Induction Motor Drives”, Journal of Applied Sciences, JAS 2007.

[7]. A. Ba-Razzouk, A. Cheriti, G. Olivier, and P. Sicard, "Field oriented control of Induction Motors using Neural Network decouplers", IEEE Trans. Power Electronics, vol. 12,1997, pp. 752-763.

[8]. Baburaj Karanayi, and Muhammed Fazlur Rahman, “Stator and rotor resistance observers for induction motor drive using fuzzy logic and Artificial neural networks,” IEEE Trans. J. Energy Convers., vol. 20, no. 4, pp. 771-780, Jan. 2005.

[9]. J. Amini, "Optimum learning rate in back-propagation neural network for classification of satellite images (IRS-1D)," Scientia Iranica, vol. 15, no. 6, pp. 558-567, 2008.

[10]. Williams, R. J., & Zipser, D. (1995). Gradient-Based Learning Algorithms for Recurrent Networks and Their Computational Complexity. In Back-propagation: Theory, Architectures, and Applications pp. 433-486.

[11]. M Gholami et A. H. Mazouji, "Fuzzy-neural networks controller-based adaptation mechanism for direct field orientation control of induction motor," 2009 International Conference on Computer and Automation Engineering, pp. 361-365, doi: 10.1109/ICCAE, Singapore. 2009.