Advanced Pattern Recognition Approach for Fault Diagnosis of Wind Turbines

The production of maximum amount of electrical power from wind requires the improvement of wind turbine reliability. The life duration and the good functioning of the wind turbine depend heavily on the reliability of its blades. Thus, a critical task is to detect and isolate faults, as fast as possible, and regain optimal functioning in the shortest time. In this paper, a pattern recognition approach is proposed for fault diagnosis of a wind turbine, in particular the pitch system composed of actuators and sensors. To achieve this task, feature and decision spaces have been defined. The aim of the pitch system is to adjust the pitch angle of a blade in order to optimize the generated electrical power according to the wind speed. Thus, a fault in the pitch system can reduce the wind turbine performance. Pitch system fault diagnosis is a challenging task because the pitch system feedback compensates the effect of the fault in the pitch actuator. In addition, the observation of the pitch actuator performance is very hard due to the strong variability of the wind speed. A wind turbine simulator is used to validate the performance of the proposed approach.