Extracting Textural Parameters from Image Logs to Enhance Reservoir Rock Typing in Diagenetic Controlled Siliciclastic Reservoir

Abstract

Abstract Image logs are appreciated by the geoscientists community due to they recognize important geological features such as sedimentological structures and structural patterns leading to enhance reservoir descriptions. Nevertheless, their quantitative use for petrophysical purposes may be underestimated. Incorporating high resolution information is vital for reservoir description whether depositional environment causes high heterogeneity, diagenetic process, and thin reservoir layers (reducing vertical permeability and acting as flow barriers reducing vertical sweep efficiency). Usually, thin sand layers are under conventional logs vertical resolution therefore they are not able to capture reservoir properties. On the other hand, cores offer direct high-resolution information, but they are expensive and only a limited number of samples are available. In this context, it is a crucial step to relate geological and petrophysical features (i.e.: sedimentary facies, lithology, permeability measurements…) from core to log domain using appropriate machine learning algorithms for clustering analysis. In this paper, an automated high-resolution classification process based on textural parameters extracted from microresistivity image logs was successfully performed. Machine learning (ML) algorithm chosen to recognize and order texture was MRGC (Ye, et al., 2000; Rabiller, et al., 2001). Textural model is defined by first and second order moments of image. Due to its high resolution, microresistivity borehole imaging provides valuable information involving grain size, sorting and porosity distribution. Moreover, in laminated reservoir zones with diagenetic complexity, where conventional logs overlooked some thin prospective intervals, this approach revealed reservoir characteristic that directly impacts on reservoir management decisions. As a result of the application of this approach, improved spatial resolution of the rock type classification yielding a more detailed and comprehensive characterization of the reservoir and enhancing geocellular model. Thus, this approach has the potential to locate bypassed oil, maximize oil recovery reserves, minimize risks and thereby saving significant costs.