It is important to identify the location of gravelly sandstone in Taoerhe Sag, an oil target area in the Chezhen Depression. To date, most attention has focused on sedimentary characteristics; hence, information regarding a clear sensitivity analysis of seismic attributes is incomplete. To address this, we used well analysis and seismic attribute interpretation to find sensitive seismic attributes and sand bodies. Based on the well analysis and core data, a geologicald model of mudstone and sandstone was established, and the electrical characteristics of three kinds of lithology were identified. We used correlation analysis to select the optimal logging parameters and seismic attributes to identify sandstone. We used multi-attribute fusion technology and stratigraphic slices to characterize the spatial distribution characteristics of gravelly sandstone in the target area and used genetic algorithm inversion volume to evaluate the prediction results. The results indicate that the three main lithologies of the lower submember of Shahejie Formation 3 (P2S3X) in the Taoerhe Sag can be distinguished by natural gamma (GR), acoustic time difference (AC) and saturated hydrocarbon content (SH) logging curves. The seismic attribute characteristics of gravelly sandstone of P2S3X are high root mean square amplitude (RMS) values, high instantaneous bandwidth (BW) values, low 3D curvature (Curv) values, medium-high instantaneous phase (Phase) values and high instantaneous frequency (Freq) values. In this study, we found two nearshore underwater fans in P2S3X. Gravelly sandstone is located along the slope zone and the lake bottom, with a total sedimentary area of 57.9 km2. The method summarized in this paper can be applied to other similar deep-water basins.