This paper deals with control and diagnosis of the synchronous generator (SG) used in a wind energy conversion system under inter turn short-circuit fault. In the first part of this paper, a speed sensorless control of the synchronous generator is presented. In this case speed estimation is carried out using model reference adaptive system (MRAS). In the second part, the MRAS observer is associated with the Luenberger observer in order to estimate the stator resistances according d and q axis. The stability of the system is proved by using Lyapunov theory. Finally, the proposed new hybrid observer MRAS-Luenberger is used to detect the turn-to-turn short circuit faults. The Park's vector approach (PVA) is adopted to take decision if there are faults. The proposed technique is tested on dSPACE DS1103 and the results confirm the efficacy to detect the fault by observing the shape. This method gives us the right information for the fault isolation.