The second member of Xujiahe Formation in Xinchang western Sichuan Depression

a tight sandstone reservoir

exhibits strong heterogeneity and thin reservoir characteristics. Previous interpretations of lithology and reservoir properties obtained directly through seismic attributes and inversion techniques were limited by individual understanding and interpretation accuracy. Consequently

the interpretation results often fell short of meeting the requirements for detailed reservoir development in gas fields.To address these challenges

a novel approach was implemented. Sensitivity parameters related to pre-stack lithology and reservoir properties were used as learning samples. Pre-stack inversion techniques were combined with machine deep learning algorithms to construct an interpretation model. This innovative method ultimately achieved quantitative predictions of sandstone thickness and reservoir properties. This method not only improves the resolution of reservoir prediction

but also greatly improves the prediction accuracy. The prediction results provide effective support for sedimentary microfacies research

reservoir formation analysis

and well location deployment

and have achieved good application results in the development of the second member gas reservoir of Xujiahe Formation in Xinchang. This article introduces a new interpretation approach for quantitative prediction of lithology and reservoirs

and serves as a valuable reference for gas field development in other regions.