588 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 30, NO. 2, JUNE 2015 Implementation of a New MRAS Speed Sensorless Vector Control of Induction Machine Idriss Benlaloui, Sa¨ ıd Drid, Senior Member, IEEE, Larbi Chrifi-Alaoui, and Mohammed Ouriagli Abstract—In this paper, a novel rotor speed estimation method using model reference adaptive system (MRAS) is proposed to im- prove the performance of a sensorless vector control in the very low and zero speed regions. In the classical MRAS method, the rotor flux of the adaptive model is compared with that of the reference model. The rotor speed is estimated from the fluxes difference of the two models using adequate adaptive mechanism. However, the performance of this technique at low speed remains uncertain and the MRAS loses its efficiency, but in the new MRAS method, two differences are used at the same time. The first is between rotor fluxes and the second between electromagnetic torques. The adap- tive mechanism used in this new structure contains two parallel loops having Proportional-integral controller and low-pass filter. The first and the second loops are used to adjust the rotor flux and electromagnetic torque. To ensure good performance, a robust vec- tor control using sliding mode control is proposed. The controllers are designed using the Lyapunov approach. Simulation and ex- perimental results show the effectiveness of the proposed speed estimation method at low and zero speed regions, and good robust- ness with respect to parameter variations, measurement errors, and noise is obtained. Index Terms—Induction motor, Lyapunov function, model refer- ence adaptive system (MRAS), sensorless control, speed estimation, vector control. NOMENCLATURE s, r Rotor and stator indices. d, q Direct and quadrate indices for orthogonal compo- nents. ¯ x Variable complex such as: ¯ x = ℜe [¯ x]+ j.ℑm [¯ x] with j 2 = −1. ¯ x It can be a voltage as ¯ u, a current as ¯ i or a flux as ¯ ϕ. ¯ x ∗ Complex conjugate. R s ,R r Stator and rotor resistances. L s ,L r Stator and rotor inductances. T s ,T r Stator and rotor time-constants (T sr = L s,r /R s,r ). σ Leakage flux total coefficient (σ =1 − M 2 /L r L s ). M Mutual inductance. P Number of pole pairs. ω Mechanical rotor frequency (rd/s). Manuscript received May 6, 2014; revised September 29, 2014 and July 16, 2014; accepted October 21, 2014. Date of publication November 20, 2014; date of current version May 15, 2015. Paper no. TEC-00310-2014. I. Benlaloui and S. Drid are with the Laboratory of Induction and Propulsion Systems, Electrical Engineering Department, University of Batna, Batna 05000, Algeria (e-mail: idrissb88@yahoo.fr; s_drid@yahoo.fr). L. Chrifi-Alaoui is with of the University of Picardie Jules Verne, 02880 Cuffies, France (e-mail: larbi.alaoui@u-picardie.fr). M. Ouriagli is with the Laboratoire d’Informatique Math´ ematiques Automa- tique et Opto´ electronique, Facult´ e Polydisciplinaire de Taza Maroc, B.P. 1223 Taza, Morocco (e-mail: omenstz@yahoo.fr). Digital Object Identifier 10.1109/TEC.2014.2366473 Ω Rotor speed (rd/s). ω s Stator current frequency (rd/s). ω r Induced rotor current frequency (rd/s). ω c Injected rotor current frequency (rd/s). J in Inertia. f Coefficient of viscous. Γ Unknown torque. Γ e ,Γ max Electromagnetic torque and maximal torque. ∼ Symbol indicating measured value. Symbol indicating the estimated value. ∗ Symbol indicating the command value. IM Induction motor. MRAS Model reference adaptive system. I. INTRODUCTION S PEED INFORMATION is mandatory for the operation of vector-controlled induction motor (IM) drive. The rotor speed can be measured through a sensor or may be estimated us- ing voltage, current signals, and machine parameters. The use of speed sensor is associated with problems, such as, reduction of mechanical robustness of the drive, need of shaft extension, re- liability reduction, and cost increase. Therefore, a speed sensor- less drive has a clear edge over the traditional vector-controlled drive. Several speed estimators for sensorless vector control of in- duction motor have been proposed as summarized recently in [1]. They can be divided into two groups, the model-based es- timators and signal injection-based estimators. Among the first group, the MRAS, the adaptive Luenberger observers and the extended Kalman-filter. The main drawback of these model- based estimators is their insufficient performance at low speeds and parameters machine sensitivity. In order to overcome these problems, signal injection-based methods [2], [3] were devel- oped. Although these methods perform well at low and zero speed regions, they suffers from, computational complexity, the need of external hardware for signal injection and the adverse ef- fect of injecting signal on the machine performance. Therefore, due to their simplicity, model-based methods and especially MRAS-based methods are, until now, the most widely used. Numerous MRAS-based on rotor flux, back electromotive force, reactive power, and outer product of stator voltage– current (¯ v ∗ s ⊗ ¯ i ∗ s )[4]–[10] have been proposed. However, rotor flux MRAS first introduced by Schauder [11], [5], remain the most popular MRAS strategy, and a lot of effort has been fo- cused on improving its performance [1]. Indeed, the drawbacks of this technique are parameter sensitivity, especially to stator resistance, and pure integration problems [6], [12]–[14], which limit its performance at low and zero speed regions of operation. 1556-6013 © 2014 EU