WIND TURBINE BREALDOWN: DETECTION FROM SCADA DATA

Abstract

This paper proposes a methodology for the prediction and detection of possible failures of major components of a wind turbine , from data collected by a SCADA system (Supervisory Control And Data Acquisition) monitoring incorporated therein . This artificial intelligence techniques are applied , such as trees or boosted neural networks for modeling the behavior and optimal selection of input parameters. Finally , once defined the methodology is applied to a real case of fault in a gearbox with data from a wind farm in the Aeolian Property Group Riojanas (GER ) located in La Rioja ( Spain ) . The combination of a detailed analysis of the optimal parameters to model the specific behavior of the temperature of the gearbox, along with the development of a neural network model based study to characterize efficiently the normal behavior of the gearbox without deterioration in performance . This allows you to analyze periodically the possible deterioration process of the gearbox and act on it before irreparable damage occurs and requiring replacement.