Abstract: The existing cluster products of ensemble system are produced by the classification on the ensemble members and are linked with certain weather process. An automatic identification forecast model for cold wave is proposed and the application of TIGGE data is improved. Based on the cold wave activity data provided by China Central Meteorological Office and reanalysis data from NCEP/NCAR, several different cluster analysis methods cooperated with subjective analysis have been used to find the typical synoptic pattern of 500 hPa geopotential height field of cold wave in China. When the cluster analysis is made on the 500 hPa geopotential height field, the results dont match subjective analysis. The cause is that the disturbance of 500 hPa geopotential height is weak compared to the basic circulation, so the different distance used by the cluster method mainly reflects the seasonal variation rather than the dissimilarities of the synoptic feature. Then the cluster based on the abnormal field of 500 hPa geopotential height is made to reduce the influence of seasonal variation and give prominence to the circulation disturbance of the cold wave. The cluster analysis results in three typical abnormal distributions of 500 hPa geopotential height during cold wave: Meridional positive-negative-positive distribution, zonal negative-positive-negative-positive and positive-negative-positive distribution. Based on the typical synoptic pattern and the threshold of the physical characteristic during the cold wave process, the objective cold wave forecast model is established. The capability of the objective cold wave forecast model are confirmed by the forecast experiments using the data from 1991 to 2006, although the result also shows the objective method makes false alarm by forecasting weak cold process sometimes. Finally two forecast experiments are conducted using the TIGGE ensemble data of the cold wave process in Jan 2008 and Jan 2009. The result indicates that by using the objective cold wave forecast model, the information of the ensemble products is concentrated, and the method could provide the forecaster with probability of the cold wave occurrence, and it consequently builds a direct relation between the ensemble forecast system and the cold wave in China. Due to the limitation of the TIGGE data, just two forecast experiments are carried out. More cases are needed in order to study the efficiency of the objective cold wave forecast model and the prediction performance of TIGGE ensemble data more thoroughly.