To realize fault prediction and intelligent condition maintenance for wind turbine generator system（WTGS）， a non-equidistant grey prediction of the operating condition parameters is investigated. Firstly， the distinction of the indicating trends and variable rules of the historical data based on different time intervals is considered， by extracting the multi-group unequal interval time sequence from the monitored data and using the average weakening buffer algorithm， a non-equidistant grey prediction model GM（1，1） of the selected data time sequences is established， respectively. Secondly， by introducing the concept of relational degree， the prediction result with the maximal relational degree is selected by comparing with the actual measured values. In addition， the operating condition parameters of the rotor speed and the temperatures of main components of a 850 kW variable speed constant frequency WTGS are forecasted by the proposed model. Finally， the rotor speed of a 2 MW WTGS is used to forecast and the prediction results are compared with that of back propagation（BP） neural network and support vector machine（SVM） methods， the results show that the prediction results by the non-equidistant grey model has a higher accuracy.