Integrated Technique Predicts Permeability Accurately in Carbonate Reservoirs

_ This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper IPTC 23411, “Integration of Rock Typing and Neural-Network Techniques for Accurate Permeability Prediction in Heterogeneous Carbonate Reservoirs: A Case Study From Abu Dhabi Offshore Field,” by Islam Elabsy, SPE, Ahmed Soliman, and Lichuan Deng, SPE, ADNOC, et al. The paper has not been peer reviewed. Copyright 2024 International Petroleum Technology Conference. _ Permeability prediction or calculation in heterogeneous carbonate formations is a challenging task because of the complexity of rock properties and pore systems that are difficult to characterize accurately. In the study detailed in the complete paper, an efficient approach was developed and used to overcome this challenge by combining rock typing and machine-learning neural-network (MLNN) techniques to predict the permeability of heterogeneous carbonate formations accurately. The approach is applicable to a wide range of carbonate formations and has the potential to improve reservoir characterization and production optimization significantly. Introduction The reservoir under investigation offshore Abu Dhabi represents a complex amalgamation of sedimentary environments, ranging from tidal flat to lagoon and inner ramp, occasionally transitioning to midramp deposits. This intricate hierarchy of sedimentary cyclicity within the reservoir reflects the influence of diverse depositional processes and water-depth fluctuations, ultimately shaping the reservoir’s heterogeneity and dynamic fluid-flow pathways. A notable geological feature within this setting is the prevalence of dolomitization, exerting a pronounced effect, primarily in the middle and lower sections of the reservoir. Dolomitization, while contributing to enhanced reservoir quality in specific zones, introduces further complexity to the prediction of permeability by altering the pore structure and modifying the petrophysical attributes of the rock. In response to the challenges posed by this complex geological environment, this study proposes an integrated approach that synergizes rock typing and MLNN techniques. The depicted study workflow delineates a comprehensive process encompassing data preparation, detailed formation evaluation, advanced rock-typing methodologies, machine-learning model training, and accurate permeability prediction. Tailored specifically to the unique challenges posed by the offshore oilfields of Abu Dhabi, this workflow embodies a holistic and adaptive approach that can advance the understanding and management of reservoir heterogeneity, leading to improved efficiency and productivity in hydrocarbon extraction. Methodology Data Preparation. The initial phase of the workflow involved a systematic data-selection process, emphasizing the comprehensive integration of core data from the targeted reservoir zones. Four wells were selected based on their extensive core and openhole logging data, enabling a comprehensive understanding of the reservoir’s geological and petrophysical properties. Additionally, the selection of three nearby deviated wells, primarily based on the availability of nuclear magnetic resonance (NMR) data and their proximity to the core wells, facilitated the calibration of the NMR permeability index using the corresponding core data. Following the data selection, rigorous log quality control (LQC) was implemented, prioritizing the validation of data integrity across varying wellbore conditions. The meticulous evaluation of NMR data was of particular significance, given its susceptibility to artifacts influenced by shallow-depth investigations. The LQC process was supplemented by an extensive data-cleaning step, aimed at rectifying any inconsistencies or irregularities within the data set, ensuring that the data was in the best available condition for future analysis in this study. Moreover, before conducting detailed analyses, a thorough depth-matching process was executed to establish a coherent alignment between different measurements and the corresponding core data. This critical step ensured the accurate integration of data from various sources, enabling a comprehensive analysis of the reservoir’s properties.