To improve low-amplitude structural reservoir trap identification and reservoir prediction accuracy, a new method of constructing low-frequency model combining constrained relative acoustic impedance with geostatistics is put forward. In this method, relative acoustic impedance information is extracted from seismic data and used as a constraint for the low-frequency model. By employing geostatistical methods such as Kriging interpolation and sequential Gaussian simulation interpolation optimization is performed based on the spatial correlation and variation characteristics of geological parameters. A low-frequency model is subsequently constructed and used for inversion research. The effectiveness of this method is demonstrated through its application in an actual work area. The results show that, compared with traditional methods, the low-frequency model constructed in this study has significant advantages in matching well logging data, depicting structural morphology, and predicting reservoir development. Its agreement with the actual geological conditions is over 85%. Moreover, the present study also reveals the limitations of this method. For example, due to the influence of seismic data resolution, the prediction accuracy for thin layers needs to be improved.