Abstract: In numerical simulation of the air flow around Daya Bay nuclear power station (NPS),a simplified data assimilation method is used to restrain forecasting errors that are caused by the uneven distribution of observation data in time and space.The effect of data assimilation on the initialization field model is analyzed.It shows that the method restrains errors in initialization field and in model significantly.

Key words: data assimilation, nuclear power station, air flow distribution