With the increasing complexity of reservoir formation mechanisms and the increasing difficulty of exploration, accurate reservoir prediction is critical for oil and gas exploration. However, traditional methods struggle to simultaneously achieve multi-source data fusion and spatial structure characterization. This study proposes a sequential stochastic fuzzy simulation (SSFS) method that integrates fuzzy recognition and sequential stochastic simulation to fuse well logging and seismic data while preserving geological spatial structure. In order to verify the effectiveness of the method, a tight sandstone reservoir in the D block of the Sulige gas field, Ordos Basin, was taken as the research target. Four gas-sensitive seismic attributes are selected, and the SSFS model is then constructed by fusing well–seismic multi-source data. Validation shows high consistency between predicted and measured gas thickness, with an R2 of 0.955 and an RMSE of 0.866 m, consistent with the dynamic gas testing results of horizontal wells. Compared with conventional geostatistical and machine learning methods, the SSFS method achieves higher accuracy, stronger spatial rationality, and better generalization ability in blind-well validation. Uncertainty analysis (mean, SD, CV, P10-P50-P90) confirms low uncertainty and high reliability. Therefore, the proposed method is reliable and effective, providing new insights for reservoir prediction.
Zhang et al. (Fri,) studied this question.