Objective In ultra-deep formation environments, structural complexity escalates remarkably. Traditional numerical simulation methods are hard-pressed to accurately predict the distribution of in-situ stress fields, directly compromising the feasibility and safety of subsequent drilling and production operations.

Methods To address this challenge, high-resolution 3D pre-stack seismic data and rock mechanics logging data were fully leveraged. Probabilistic neural network prediction technology effectively resolved the issue of missing shear wave curves in wells. Pre-stack AVO inversion technology was adopted to construct large-scale 3D models of P-wave velocity, S-wave velocity, and density covering the entire 600 km² work area, achieving refined characterization of elastic parameters in the target region. Based on the Eaton equation accounting for the impact of abnormal high pressure in ultra-deep layers, accurate 3D formation pressure prediction for the entire stratigraphic sequence was realized. Finally, a combined spring model integrated with structural curvature and cokriging interpolation technology were employed to efficiently establish a high-precision 3D in-situ stress model for the entire structure from the surface to the target layer.

Results The seismic data-driven in-situ stress field modeling method demonstrates remarkable advantages in applications to ultra-deep complex formations, with a prediction accuracy of 93.793%. Three key insights were further revealed: the barrier effect of imbricate structures on in-situ stress, the intensification of the neutral plane with the increase of structural deformation, and obvious stress release at faults.

Conclusions This method lays a technical foundation for constructing a "transparent basin" and provides effective support for wellbore stability analysis, well path/fracturing optimization, fracture development and stability, as well as seismic source rupture research. It serves as a critical basis for optimizing oilfield development plans and mitigating production risks.