Condition Monitoring of Wind Turbines Using Adaptive Control Charts

Due to the harsh working environments, variable speeds and alternating loads, wind turbines are likely to breakdown or suffer damage. Effective condition monitoring methods for wind turbines are essential for maintenance decisions which aim to reduce O&M costs. A typical supervisory control and data acquisition (SCADA) system records comprehensive wind turbine condition parameters, which would be fault informative. Thus, a framework for condition monitoring of wind turbines is introduced based on adaptive control charts and SCADA data. The adaptive exponential weighted moving average (AEWMA) is proposed for abnormal state alarm of wind turbines. Random forest (RF) is used for feature selection and regression prediction to establish the normal condition prediction model (NCPM) of wind turbine with fault-free SCADA data. The performance and robustness of various control charts are compared comprehensively. Compared with the exponential weighted moving average (EWMA) control charts, the AEWMA control chart behaves more sensitive to the abnormal state, and thus has more effective anomaly identification ability and better robustness.