Abstract:Based on the analysis of large amounts of time series data collected by supervisory control and data acquisition system (SCADA) in wind turbines, a wind turbine online condition monitoring approach based on generative adversarial network (GAN) is proposed. Firstly, a data set that has the same dimension with the SCADA data is generated with the generative model. Secondly, the generated SCADA data and the real SCADA data are used to optimize and train the GAN model. After training, the obtained discriminative model in GAN is used for distinguishing the health condition of wind turbines. Finally, the proposed approach was used to analyze the SCADA data of a healthy and a faulty wind turbines. The result shows that the GANbased approach can effectively monitor the online operation condition of the wind turbines, it can detect the anomalies of the faulty wind turbine 5 days earlier than the SCADA system. When the wind turbine works normally, the number of false alarms reported by the GAN approach is less than other approaches (such as Mahalanobis distance, principal component analysis, deep neural network and support vector machine). When the wind turbine fails, the GANbased approach can detect more abnormal samples than other approaches.