"Geology-geophysics-data mining" integration to enhance the identification of deep fault-controlled paleokarst reservoirs in the Tarim Basin

Large-scale strike-slip faults play a dual role as conduits for fluid flow, channeling both atmospheric freshwater and deep hydrothermal fluids, thereby creating porous carbonate reservoirs at significant depths. Moreover, these faults serve as vertical conduits, connecting deep source rocks to oil and gas reservoirs capable of generating substantial hydrocarbon fields. In recent years, the Tarim Basin has yielded numerous high-yield reservoirs along large-scale strike-slip faults, exhibiting exceptional hydrocarbon potential. Nevertheless, the intricate heterogeneity of deep-seated traps engendered by these faults presents a formidable challenge in accurately determining reservoir geometry, consequently complicating the drilling of high-yield wells. Currently, enhancing the precision of reservoir identification stands as a paramount issue within the Tarim Basin.This paper introduces an innovative solution termed the "Geology-Geophysics-Data Mining" workflow, designed to enhance the resolution of multiscale geophysics data through the lens of geological insights. Drawing inspiration from the geological model of "fault-controlled paleokarst," and utilizing a fusion of well logs and seismic data, we elect seismic impedance as a discerning parameter to differentiate between porous reservoirs and impermeable host rocks. The proposed methodology unfolds through several steps.Primarily, noise reduction and frequency domain extension of the original seismic data are achieved through spectral shaping and diffusion filtering algorithms. Subsequently, the impedance curve of the logging wave is computed, establishing a non-linear correlation with seismic waveforms using a waveform indication inversion algorithm, ultimately yielding inverse impedance data. In the subsequent phase, the impedance originating from stable sedimentary features is quantified and subsequently extracted from the inversely derived seismic data. This step accentuates the high-quality reservoirs formed via fault-controlled karstification. Ultimately, we validate the identification of reservoir geometry using interpretations from well log data and drilling outcomes. Furthermore, this paper employs the seismic impedance obtained from wellbore logging data as "hard data" and deploys machine learning algorithms to map this information to three-dimensional high-precision seismic data. Following the removal of sedimentary features' influence, we extract the longitudinal connectivity and transverse mutation characteristics inherent in fault-controlled paleokarst reservoirs. This extraction enables the capture of geometric structural information related to fault-controlled paleokarst reservoirs, further substanti-ating the geological distribution of effective reservoirs controlled by strike-slip faults. These findings establish a dependable foundation for directional drilling and studies focused on hydrocarbon accumulation. Notably, the "Geology-Geophysics-Data Mining" approach is extendable to analogous strongly heterogeneous reservoirs.