Stator Fault Classifier in a Doubly-Fed Induction Genera-tor DFIG

Abstract

Fault detection is one of the most common studies on wind turbines. In this case, the doubly fed induction generator (DFIG) is specifically analyzed. The fault cases analyzed are: inter-turn short circuit and open circuit, in addition to normal operating conditions. The K-means algorithm was used for analysis and classification. The data set is obtained from multiple simulations in MATLAB/Simulink, in which the stator resistance (Rs) and stator inductance (Ls) were varied. From these simulations, the current and voltage signals are processed using tools such as the Park transform, stator current imbalance, and harmonic analysis to obtain relevant characteristics for the classification of each case. It is known that Rs is a determining parameter in fault detection: in a short circuit, Rs tends to fall below its nominal value due to the appearance of a low-impedance path, while in an open circuit, Rs tends to rise above the nominal value due to the interruption of the conductor. The results obtained indicate that the K-means algorithm, together with the proposed methodology, are efficient in classifying the different stator states. This suggests that the proposed solution, based on the results, could be effective and economical for wind turbine monitoring.