A sensor fault detection scheme of DFIG-based wind turbine using deep auto-encoder approach

Abstract

The reliability of the wind turbine doubly-fed induction generator (DFIG) is of paramount concern for adequate power production. This paper investigates an effective fault detection scheme for DFIG using the deep auto-encoder (DAE) structure. The methods contain three main steps: first, the measurement of the stator currents and voltages directly presented to the DAE to capture the characteristics of the signals effectively. Second, using those features, a neural network model is used to detect faults affecting the stator immediately. Then, a binary decision logic proposed for isolation. The results confirm the method efficiency, rapidity, robustness against the occurrence of multiple faults in the presence of measurement noise and unknown inputs.