This research contributes with a reliable approach to detect incipient stator winding inter-turn short-circuits in induction generators applied in wind turbines. Using a wind turbine test-bench, we inserted different types of short-circuit in the generator. The electrical current is acquired to build a fault database. We propose the use of four feature extraction techniques with three classifiers. The MLP identified 100% of the generator’s Normal conditions with less than 1% false positives and negatives. Using different topologies of MLP, it was possible to identify incipient short-circuits in 1.41% turns with 99.33% accuracy. The combination Fourier-MLP is more useful for fault detection, since it obtained 84.48% of accuracy, with 99.98% of Normal conditions correctly classified.
Among the renewable energy sources, wind energy has become the most effective and accepted solution for electricity generation worldwide, contributing 486.7 GW to global demand . This energy production represents only 3% of the world’s energy demand, but it is estimated that in 2030 the wind power will be able to supply 17–19% of the global demand .
Despite the growing exploitation of this energy source at 17% per year  the technologies are not consolidated, there are still engineering and science challenges to be solved to support this expansion.
Operational problems directly impact on the cost of energy, according to Polinder et al.  only with a reliable and available wind turbine system that the cost of energy can be mitigated. Moreover, in the light of reliability are the maintenance operating costs, which account for up to 30% of the cost of energy .
Hahn, Durstewitz and Michael have already evidenced the concern with maintenance in wind farms . They exhibited records of fault types in wind turbines. The compilation of data from a set of wind farms installed in Europe has shown that the most costly faulty component for the wind farm is the electric generator.
Among the areas of study surrounding wind turbines, the present text focuses on electric generators, especially the Squirrel Cage Induction Generator (SCIG). The importance of the SCIG is based on its robustness, consolidated technology and future trends. Also, Yaramasu et al.  forecasts this generator will dominate the market of wind turbine in the next years.
Despite the versatility of SCIG, it is not immune to faults, and has limitations. In general terms, these faults are associated with: overheating, electrical’, dynamic and mechanical effects . Bonnet and Soukup  emphasizes that the short-circuit between turns is the most incipient fault and the most difficult to detect.
So, this work focus on detecting incipient short-circuits in SCIG, to provide a means to improve wind turbine reliability. Moreover, this could be potentially good for giant wind turbines. Because, detecting an incipient short-circuit before total degradation could guide corrective maintenance to make repairs on site, without having to remove the generator from the wind turbine and decrease the machine availability.
Since wind turbines are a sensor-based equipment, monitored and controlled by a Supervisory Control and Data Acquisition (SCADA) system, the incipient short-circuit detector could be deployed inside the hardware responsible for data acquisition. So, one could feasibly replicate the methodology of this paper in real wind turbine systems.
This paper is organized as follows: in Section 2 state of the art in fault detection for reliability improvement in electrical machines is presented; in Section 3 the feature extractions methods used in this work are explained; in Section 4 the procedures to emulate the fault emulation, acquire data and create a dataset for pattern recognition are presented; in Section 5 the machine learning methods used in this work are described, as well as its configurations and metrics for evaluations; in Section 6 our results are analyzed and conclusions presented in Section 7.