Hierarchical Modeling for Fault-Controlled Karst Reservoirs Assisted by the Statistical Thresholding Method and Stochastic Attention Mechanisms

The paper delves into the distinctive fault-controlled karst reservoirs situated in China’s Tarim Basin, known for their marked heterogeneity due to the interplay of tectonic movement and dissolution processes. Tectonic cavities and dissolution caverns serve as pivotal hydrocarbon accumulation spaces. These reservoirs manifest as narrow zones following major strike-slip fault lines, occurring at depths exceeding 8000m. While sedimentary facies remain unaffected, reservoir heterogeneity is greatly influenced by tectonic and dissolutive geological elements. This complexity renders conventional geostatistical techniques used in clastic reservoir modeling unsuitable. Furthermore, cavity interiors, prone to loosening, lead to well leaks and rapid bit pressure decline, resulting in scarce well logs. Addressing these challenges, this paper proposes a hierarchical modeling approach. It employs statistical thresholding and deep learning methodologies, bolstered by improved seismic attribute volumes, to sequentially characterize the architectural components of fault-controlled karst reservoirs. A detailed case study focusing on the Shunbei 1 fault zone exemplifies the method’s application, unraveling nuanced insights into this unique geological phenomenon.