Abstract: [Objective]To address the difficulties in three-dimensional lithofacies modeling of shale reservoirs under sparse-well conditions, including insufficient lateral constraints between wells, difficulty in constructing representative three-dimensional training images, and non-orthogonal spatial distribution of interwell structures, a three-dimensional lithofacies modeling method constrained by non-orthogonal connected-well sections is proposed.[Methods] Taking the F2-2 Member shale reservoir in the Qintong Sag as an example, well lithofacies data were first discretized into a three-dimensional model. Two-dimensional interwell sections were then constructed by extracting profiles between arbitrary pairs of wells and written back into the three-dimensional model. On this basis, slice-by-slice simulation was carried out along non-orthogonal slicing paths consistent with the section directions. In addition, planar guiding points were introduced to supplement the intra-layer distribution information in weakly constrained areas, thereby forming the NCWDS three-dimensional lithofacies modeling workflow.[Results]The case study shows that the NCWDS model achieves a hard-data matching rate of 100.0% at well locations and a facies-proportion L1 deviation of 1.8736 percentage points. Compared with SIS and SNESIM, the NCWDS model has the lowest weighted mean absolute error of the vertical transition probability matrix, with a value of 0.056454. The same-facies adjacency ratios in the X, Y, and Z directions reach 98.5359%, 98.5419%, and 75.8044%, respectively. The total number of connected components, the proportion of small patches, and the proportion of isolated cells are all lower than those of the comparison models[Conclusion]The NCWDS method can better preserve the vertical stacking patterns and spatial continuity of shale reservoirs while strictly honoring well-control constraints. It also reduces the fragmentation of lithofacies distributions in weakly constrained areas, providing a methodological reference for three-dimensional lithofacies modeling of sparsely drilled reservoirs with strongly developed lamination.