In the processing of Ocean Bottom Node (OBN) seismic data, suppressing surface multiple is one of the key factors affecting the quality of imaging. In OBN acquisition, due to the special geometry where seismic sources are positioned near the free surface and receivers are deployed on the seafloor, conventional surface multiple suppression methods based on wave theory (such as Surface-Related Multiple Elimination, SRME) are not applicable, as they generally assume that sources and receivers are coplanar and evenly spaced, and rely on full wavefield data. To overcome these limitations, this paper proposes a surface multiple prediction method based on demigration theory. By incorporating information such as migrated sections and velocity fields, the method uses one-way wavefield extrapolation in common node gathers to simultaneously predict all orders of surface multiples generated between the seabed, sub-seabed interfaces, and the free surface. Additionally, Phase Shift Plus Interpolation (PSPI) is introduced to handle the lateral velocity variations of complex subsurface geology. The results from synthetic models and field data examples demonstrate that the surface multiple models predicted by this method are highly consistent with those observed in the original data, highlighting the method's significant practical value for the suppression of multiples in OBN seismic processing.