In order to improve the real-time reliability of grid-connected wind turbines，optimize the maintenance strategy，and reduce the cost of wind power generation，it is necessary to consider the interaction and coupling between components or subsystems of a wind turbine.An online assessment model for the operation conditions of the whole wind turbine is established by data mining technology.Firstly，after analyzing the shortcomings of the supervisory control and data acquisition（SCADA）warning system of wind turbines，a more robust on-line assessment scheme is proposed based on the cooperation of a regression prediction model and the SCADA warning system.Secondly，the regression prediction model is described in detail that the support vector regression（SVR）algorithm is adopted.The inputs of SVR are part of the monitoring projects of the SCADA system，and the output of SVR is the active power of the wind turbine.Finally，measurement results of a wind farm are used to verify the proposed model.