Taking the SINS/Beidou loose integrated navigation as the object, the processing method of Beidou positioning error at sea is studied. By analyzing the source of Beidou positioning error, it is pointed out that there is a constant deviation of Beidou positioning error in local areas within a short time. This constant deviation can be compensated by the differential data released by the ground differential stations. However, at this stage, the ground differential stations of Beidou system are mainly concentrated in the key areas of China. For overseas areas, especially marine areas, where there is no ground differential station, the common method is to treat the Beidou positioning error as noise, but this method will sacrifice the accuracy of integrated navigation. In this paper, the constant deviation of Beidou positioning is incorporated into the filter state. And the Kalman filter state expansion method is proposed to estimate this constant deviation to improve the integrated navigation accuracy. Finally, the simulation examples show that the method can effectively improve the positioning accuracy.