A data-driven approach for sensor fault diagnosis in gearbox of wind energy conversion system

Due to the increase in worldwide energy demand, wind energy technology has been developed rapidly over the past years. With a fast growing of wind power installed capacity, an efficient monitoring system for wind energy conversion system (WEC) is required to ensure operational reliability, high availability of energy production and at the same time reduce operating and maintenance (O&M) costs. The state of the art methodologies for WEC condition monitoring are signal analysis, observer-based approach, neural networks, etc. In this paper, an effective and easy adaptable multivariate data-driven method for wind turbine monitoring and fault diagnosis is introduced, which consists of three parts: 1) off-line training process 2) on-line monitoring phase 3) on-line diagnosis phase. The performance of this method is validated for detection of sensor abnormalities that have occurred in real wind turbines.