Abstract: Owing to low dominant frequency and narrow band of seismic data acquired in the Xunyi area, southern Ordos Basin, it is difficult to identify thin gravity-flow sands in the 7^th Member of the Yanchang Formation (Change 7 Member). To accurately identify single-layer thin sands and internal barriers within thick sands, this paper proposes a method and workflow based on harmonic spectrum extension. Based on single-well sands interpretation, frequency gathers obtained from sinusoidal decomposition are used to analyze the variation of seismic signals, the degree of which is characterized using information entropy to determine the characteristic frequencies of sand bodies with different thicknesses. Leveraging the multi-resolution characteristics of continuous wavelet transform, the fundamental wave signal is then extracted from the dominant frequency band. Harmonic signals are calculated and added to the fundamental signal to obtain stacked signals. The time-frequency spectrum within the effective frequency band is calculated iteratively to achieve synchronous expanding of seismic data at both low and high frequencies. Based on the broadband data, high-resolution seismic meme inversion integrating well and seismic information is performed to predict thin gravity-flow sand bodies in the study area. Drilling results demonstrate an accuracy rate of prediction over 85%, providing robust support to the deployment of horizontal wells for production and efficiency improvement in the exploration and development of tight sandstone oil. This technique is applicable to gravity-flow sandstone identification in semi-deep to deep lacustrine sedimentary settings.