Advancing Reservoir Understanding: Integrating Deep Learning into Automated Formation Micro-Imaging Log Interpretation for Enhanced Reservoir Characterization and Management

Abstract

Abstract The Formation Micro-Imager (FMI) produces high-resolution borehole images, crucial for identifying geological features like bedding and fractures, aiding in reservoir characterization. Its sensitivity to mineralogy and fluid content variations enhances its utility in subsurface analysis for optimizing hydrocarbon extraction strategies. However, interpreting FMI logs requires expertise and experience due to the complexity and nuances of the images, posing challenges in obtaining accurate geological information. This research harnesses the capabilities of automated Formation Micro Imager (FMI) interpretation combined with expert calibration to innovate the analysis of subsurface stress fields, striving to improve the accuracy of geo-mechanical modeling and optimize water-flooding projects, particularly those necessitating hydraulic fracture stimulations. It revolves around the creation of accurate local stress maps using FMI data, specifically drilling-induced fractures, to determine the present-day maximum horizontal stress (SHmax) orientations, thereby informing wellbore stability and drilling strategy optimization. This study utilizes advanced logging techniques to identify borehole breakouts and other enlargements, which are critical for accurate in-situ stress estimation. This process is supplemented by a comprehensive analysis of multiple wells, revealing the variability and predominant orientations of SHmax with the regional tectonic framework. The study uncovers the significant variability of SHmax orientations, aligning predominantly with major fault trends within the Egyptian Western Desert's complex geological structure. The results from this analysis facilitate an updated regional stress map, indicating the prevalence of normal and strike-slip faulting regimes. Additionally, the research extends to reservoir management, where automated FMI log interpretation aids in optimizing hydraulic fracture directions in Field A, contributing to more efficient hydrocarbon extraction and better reservoir pressure maintenance strategies. The novel integration of FMI data with deep learning applications presents a groundbreaking approach to subsurface analysis. It offers a new perspective on managing reservoirs by strategically placing water injector wells to support hydrocarbon production while maintaining reservoir integrity. This method enhances the predictive capability of simulation models, ensuring more accurate performance matching and better-informed decision-making in production strategies. The study's approach combines geological insights with advanced technology, offering a substantial advancement in the field of geo-mechanics and reservoir characterization.