Lithology forms the foundation for evaluating high-quality reservoirs and is crucial for effective hydrocarbon development. Deep tight sandstone reservoirs are characterized by significant burial depth, complex lithology and strong vertical heterogeneity, leading to challenges for traditional well-log lithology identification. This study focuses on the deep tight sandstone reservoirs of the Jurassic Sangonghe Formation within the Taibei Sag, Turpan-Hami Basin. Integrating core, thin section, laboratory analyses, and conventional well logs through core-log calibration, reservoir lithology was classified into siltstone, fine sandstone, medium sandstone, coarse sandstone, and sandy conglomerate based on median grain size. The results demonstrate that the Sangonghe Formation reservoirs exhibit complex and diverse lithologies, dominated by fine, medium, and coarse sandstones. The primary pore type is intragranular dissolution pores. A machine learning-based lithology prediction model was developed using gamma ray (GR), deep resistivity (RD), shallow resistivity (RS), acoustic transit time (DT), bulk density (DEN), and compensated neutron log (CNL) curves as input features, with median grain size as the prediction label. The model achieved a high coefficient of determination (R²) of 0.897 on the test dataset, showing strong agreement with core-measured data and enabling continuous vertical lithology classification in single well. Application to blind wells confirmed the model's generalization capability and prediction reliability, overcoming limitations imposed by scarce core data on reservoir evaluation. Further coupling analysis of lithology and physical properties reveals that lithology significantly controls reservoir quality, with medium and coarse sandstones exhibiting optimal properties and representing the primary lithology for high-quality reservoir development. This study provides theoretical and technical support for sweet spot prediction and efficient gas reservoir development in tight sandstones of the Sangonghe Formation, Taibei Sag.