-Wind turbine generators (WTGs) are usually equipped with mechanical sensors to measure wind speed and rotor position for system control, monitoring, and protection. The use of mechanical sensors increases the cost and hardware complexity and reduces the reliability of the WTG systems. This paper proposes a mechanical sensorless maximum power tracking control for wind turbines directly driving permanent magnetic synchronous generators (PMSGs). In the proposed algorithm, the PMSG rotor position is estimated from the measured stator voltages and currents by using a sliding mode observer (SMO). The wind turbine shaft speed is estimated from the PMSG back electromotive force (EMF) using a model adaptive reference system (MRAS) observer. A back propagation artificial neural network (BPANN) is designed to generate the optimal shaft speed reference in real time by using the estimated turbine shaft speed and the measured PMSG electrical power. A control system is developed for the PMSG wind turbine to continuously track the optimal shaft speed reference to generate the maximum electrical power without using any wind speed or rotor position sensors. The validity of the proposed control algorithm is shown by simulation studies on a 3-kW PMSG wind turbine and experimental results on a practical 300-W PMSG wind turbine.Index Terms-artificial neural network, model adaptive reference system (MRAS), permanent magnet synchronous generator (PMSG), sensorless control, sliding mode observer (SMO), wind turbine.