Abstract: Tight sandstone reservoirs deposited in braided river delta systems are generally characterized by low porosity, low permeability, complex pore-throat structures, and strong heterogeneity. Conventional well logging interpretation methods mostly rely on single parameters or empirical criteria, making it difficult to simultaneously reflect reservoir storage capacity and seepage capacity, thus limiting the accuracy and interpretability of reservoir classification. To address these issues, this study proposes an intelligent well logging classification method for tight sandstone reservoirs based on flow unit index (FZI) constraints. First, the core-based FZI is calculated using rock physics experimental data from the study area, and the cumulative frequency method is applied to classify the core FZI and establish the classification criterion for the reservoir indicator parameter F. Then, reservoir porosity is calculated using the neutron–density crossplot method, and a permeability model is established based on porosity–permeability experimental data to estimate reservoir permeability. Based on the calculated porosity and permeability, the logging-derived FZI and classification indicator parameter F are further computed for the study area. Finally, conventional logging curves, including GR, AC, DEN, CNL, and RD, together with the model-derived porosity, permeability, FZI, and F, were used as input data. By incorporating F into the input feature vector along with conventional logging curves, the clustering process was constrained by seepage-related physical properties. A Gaussian Mixture Model (GMM) clustering algorithm was then employed to achieve intelligent classification of tight sandstone reservoirs and establish corresponding classification criteria. The proposed method is applied to a typical braided river delta depositional area in the western margin of Well Pen-1 in the Junggar Basin, achieving fine classification of low-porosity and low-permeability reservoirs and clarifying the relationship between reservoir types and gas-bearing properties. The results demonstrate that this method effectively improves the accuracy of reservoir classification and shows strong potential for application to other tight sandstone reservoirs with low porosity and low permeability.