Abstract:China Meteorological Administration (CMA) is one of the data exchange centers of global ensemble forecasting system. Global ensemble forecast systems (GEFS), including the European Center for Medium-Range Weather Forecasts (ECMWF), and the National Centers for Environmental Prediction (NCEP) have been used in real-time operation of the National Meteorological Centre (NMC) of CMA since 2010. The application of multi-model has promoted the progress of tropical cyclone forecasting operation of CMA. Based on the data of ECMWF and NCEP ensemble models from 2010 to 2019, we evaluate the performance and systematic deviation of their track forecast performance in following aspects: intensity, moving speed, month, ocean area, landing and turning. The ensemble mean forecast errors of the two models generally show a downward trend, and their systematic deviations are almost in the opposite directions. Separately, ECMWF ensemble forecast has a southwest bias, while NCEP ensemble tends to be northeastward, and is larger than the former in number. At present, the track prediction performance of ECMWF ensemble model is better than that of NCEP, especially for weak TCs and landing TCs. The results of the performance evaluation and deviation analysis could also provide a reference for actual TC forecasts when there are obvious differences between the products of the two ensemble models.