An LPV Estimation-Based Approach for Fault Diagnosis in Drive Train System

For improving the safety and the reliability of wind turbines and shield them from catastrophic behaviors due to sudden breakdowns, it is important to detect faults as fast as possible. In this paper, a generator speed sensor fault and a fault due to the changed dynamics of the drive train, induced by increased friction will be treated.  For diagnosis, a residual generation design is used based on a comparison between the real generator speed and the estimated one obtained by the use of Linear Parameter Varying (LPV) predictor based subspace identification technique applied at the benchmark system which is modeled as an LPV model considering the wind speed as scheduling variable. The simulation results show the efficiency of the proposed method.