Abstract: In order to evaluate the health degree of the gearbox of wind turbine timely and accurately,this paper proposes a performance monitoring and evaluation method based on KECA-GRNN. This method is divided into three stages:condition monitoring,fault prediction and health assessment. In the condition monitoring stage,KECA algorithm is applied to the performance monitoring of wind turbines,and SPE statistics data are used to monitor the state of gearboxes. In the stage of fault prediction,the master metadata extracted by KECA algorithm is used as the input of GRNN model to establish the KECA-GRNN prediction model,and the change trend of prediction residual is used to define the alarm limit to realize the early warning of fault. In the stage of health assessment,the multivariate prediction residuals are integrated to enhance the reliability of assessment. Finally,the method is applied to the SCADA data of a 1.5 MW wind turbine in a wind farm nearly two months before the failure. The results show that the abnormal condition of the gearbox can be known two weeks in advance,which realized the accurate assessment of the healthy state of a wind turbine gearbox.