Qualitative diagnosis of wind turbine system based on wavelet transform

In this paper we present a qualitative evaluation of generated residual signal using wavelet transform in purpose of fault diagnosis for wind turbine benchmark model. The fault detection is based on generating residual signal by comparing the real and an estimated behavior. The `Takagi-Sugeno' (TS) fuzzy identification and modeling is considered to approximate the non linearity presented in this system. Due to noise in the wind speed, the generated residual signal has to trade of the risk of false alarms to the risk of undetected faults. Occurrence of false alarms is largely dictated by the quality of the model of which the design of the Fault Detection and Isolation (FDI) system relies. Therefore, the proposed method using wavelet transform is considered to remedy the problem of false alarms. The treated signal of the residue with wavelet gives significant results which are validated with the wind turbine simulator.