An anisotropy-dynamic ant colony optimization with probabilistic fracture uncertainty quantification for sub-seismic fault detection

The precise characterization of subsurface fracture systems, especially sub-seismic fractures below seismic resolution, is critical for developing complex hydrocarbon reservoirs. While ant colony optimization (ACO) introduced “ant tracking” for seismic fracture detection, traditional methods rely on isotropic post-stack attributes, ignoring azimuthal anisotropy—a key indicator of fracture orientation and density. The azimuth-aware anisotropic bayes ACO (Ani-Bayes ACO) integrated pre-stack anisotropy via Bayesian priors but suffered from deterministic constraints and static heuristics, limiting its ability to model conjugate fracture systems or parameter uncertainty. To resolve these limitations, we propose the anisotropy-dynamic ACO (ADACO) algorithm. ADACO replaces deterministic constraints with probabilistic, dynamically evolving fracture parameter distributions: von Mises for orientation and log-normal for density, both parameterized by elliptical fitting credibility. During optimization, a Hidden Markov Model (HMM) globally evaluates path consistency, while elite-path feedback iteratively focuses the distributions. This enables uncertainty-quantified fracture prediction, multi-set tracking, and autonomous adaptation to fracture clustering. Validation in a complex shale gas reservoir showed 85% consistency with drilling data—a significant improvement over Ani-Bayes ACO (46%). ADACO thus provides a robust tool for sub-seismic fracture characterization.