Abstract: Considering the Earth’s curvature effect, the positioning accuracy of very low frequency(VLF) emission sources is studied by using a long-distance multi-station time difference positioning algorithm based on a spherical model. By the Monte-Carlo simulation method, the spatial distribution of the root mean square error (RMSE) of positioning results is calculated for the two typical station geometries of the square- and star-shaped stations in the four-stations receiving system. The influence of three factors of station geometry, baseline length, and master station selection on positioning accuracy is analyzed. The results show that increasing the baseline length can effectively improve the positioning accuracy, but the improvement effect decreases as the baseline increases. Compared with the square-shaped stations, the star-shaped geometry has more advantages. The overall positioning error in the detection area is lower without a directional distribution, so it is more suitable for blind source positioning. For the star-shaped configuration, the positioning accuracy of the master station at the center of the triangle is better than that at other positions. The research results have a significant reference value for the actual receiving site setting of the long-distance VLF source positioning system