Correlation analysis based relevant variable selection for wind turbine condition monitoring and fault diagnosis

Wind turbines' fault diagnosis under complex environments and disturbances is significant to maintaining high reliability and secure operation over a prolonged period of time. Due to the difficulty of installing additional sensors, the supervisory control and data acquisition system is the only path for condition monitoring and fault diagnosis. However, the complexity of numerous variables bogged down the situation of diagnosis. Hence, this paper proposes a correlation analysis method to filter the variables for maximizing redundant data suppression first. Secondly, a data utility maximization method based on a prior-posterior support vector machine is proposed. Finally, a series of parallel support vector machines are used to realize multi-condition monitoring and fault diagnosis. Experiment results illustrate the effectiveness, robustness, and generality of the method.