Rev. Roum. Sci. Techn.– Électrotechn. et Énerg. Vol. 64, 4, pp. 323–330, Bucarest, 2019 1 Laboratoire de Génie Energétique et Génie Informatique (L2GEGI), Ibn Khaldoun University of Tiaret, Algeria, tahriahmed89@gmail.com 2 Laboratoire d’Informatique et d’Automatique pour les Systèmes(LIAS), University of Poitiers, France. EXPERIMENTAL VERIFICATION OF A ROBUST MAXIMUM POWER POINT TRACKING CONTROL FOR VARIABLE SPEED WIND TURBINE WITHOUT MECHANICAL SENSOR AHMED TAHRI 1 , SAID HASSAINE 1 , SANDRINE MOREAU 2 Key words: Permanent magnet synchronous generator, Wind turbine, Sensorless control, Maximum power point tracking (MPPT), Backstepping control, Sliding mode observer. This paper presents a robust sensorless control scheme of permanent magnet synchronous generators (PMSG) wind turbine with two full-scale controllable three phase converter connected to the grid. The main aim is to drive wind turbine at an optimal rotor speed in order to perform maximum power point tracking (MPPT) control of the wind generation system. The proposed strategy combines a nonlinear backstepping control of the PMSG which is based on both feedback laws and Lyapunov theory and a non-linear sliding mode observer (SMO) to estimate the PMSG rotor position in order to perform a sensorless control. The PMSG wind turbine is integrated with the grid through a three phase inverter which is controlled through classical proportional integral (PI) to control the active and reactive power. The proposed method was successfully tested with both simulation and experimental setup using MATLAB and a dSPACE 1104 platform. The obtained simulation and experimental results show clearly the accuracy and the effectiveness of the proposed method. 1. INTRODUCTION The petroleum crisis and the increasing demand of energy, coupled with the possibility of a reduced supply of conventional fuels, have motivated progress in renewable energy research and applications. Among renewable energy sources, wind energy is currently considered to be the most useful natural energy source because it is abundant and clean. Despite these advantages, the efficiency of wind energy conversion is currently low, and the initial cost for its implementation is still considered high. Therefore, the wind energy cost reduction is essential for the rapid spread of the wind power generation through more efficient, reliable and cost-effective wind energy conversion systems (WECS)[1]. WECS, based on permanent magnet synchronous generator (PMSG) drive, is a promising technology due to its high efficiency and the elimination of gearboxes and the external excitation system [2]. Various wind generator systems based on modern wind turbine concepts are developed and described in literature [3–5], which are all about reducing the cost to the minimum, power maximization and improvement and protection of the environment. The direct-drive PMSG is the attractive solution/technology thanks to its high efficiency and reliability as well as its high power density and torque-inertia ratio [5]. However, the performances of the PMSG system depends on synchronous generator and its control technique. The principle of the vector control introduced by Blaschke [6] gave good dynamic performances. Therefore, many researches are currently focusing on improving this control type. The main difficulty of the vector control is the speed measurement. Therefore, the present research work focuses on the elimination of such sensor. Indeed, the use of mechanical sensors such as tachometer generators, incremental encoders and resolvers to measure mechanical speed not only reduce reliability but also increase the maintenance cost of the control. Consequently, vector control without a mechanical sensor has become an increasingly important part in industry and research. Several techniques have been developed for the PMSM sensorless control technology. They can be classified into two types: the estimation method based on observer [7,8] and the high-frequency injection method using the salient effect of the motor [9,10]. The flux linkage estimation method, an extended Kalman filter and the model reference adaptive method are widely used to estimate the rotor position. However, the estimation accuracy and stability of the last method depends on motor model. In [11,12], an adaptive neuro-fuzzy method is provided to improve the accuracy of the position estimation under varying speed with variable PMSG parameters, whereas adaptive neuro-fuzzy inference system (ANFIS) estimator shows better performance and immunity against parameters variation. However, in practice, the problem with fuzzy controllers is that the calculation time and complexity increase strongly with the number of variables. In [13] an adaptive interconnected structure with online parameter estimation is adopted to improve the accuracy of the position estimation. It is based on the interconnection between several observers, satisfying some required properties. However, the convergence of the observer depends on the gains related to the state estimation and the parameter identification of the system. Considering the circumstances mentioned above, sliding mode observer (SMO) is widely used thanks to its robustness and simplicity, which compensate the dependence of the observer on the model [14,15]. This paper presents an improvement and experimental verification of the sensorless control strategy for a PMSG based wind turbine. The control is based essentially on the backstepping control [10,11,16] of the wind turbine speed by imposing the wind profile while SMO is introduced to estimate the speed of the wind turbine generator. In