Advanced AI Approach for Enhanced Predictive Modeling in Reservoir Characterization Within Complex Geological Environments
Wakeel Hussain,Muhammad Ali,Rakhshanda Sadaf,Harith F. Al-Khafaji,Izhar Sadiq,Chuanyu Hu,Hamza Daud,Salah Alshareef Alkfakey Ahmed
DOI: https://doi.org/10.1007/s40808-024-02049-5
2024-01-01
Modeling Earth Systems and Environment
Abstract:Presently, various machine learning techniques are commonly used to enhance reservoir properties, including fuzzy logic, neural networks, cluster analysis, self-organizing maps, principal component analysis, genetic algorithms, and more. While each method possesses its own set of strengths and weaknesses, a key challenge faced by most existing techniques is the optimum handling of the diverse, dynamic ranges evident in petrophysical input data. This paper introduces Domain Transfer Analysis (DTA), an advanced method developed from extensive Computational Fluid Dynamics (CFD) research, designed to enhance the integration of diverse petrophysical data types. DTA, distinct from traditional statistical techniques in the petroleum industry, employs a nonlinear partial differential equation solver to more effectively handle input variables that exhibit both logarithmic (e.g., resistivity) and linear (e.g., gamma ray) variations. The study is particularly important as it addresses the critical challenge of optimally handling dynamic ranges in petrophysical input data, which has been a significant limitation of existing methods. The performance of this technique is demonstrated through case studies that examine the prediction of porosity and permeability using a dataset from wells in the Lower Goru sand reservoir, Southern Indus Basin, Pakistan. The comparative analysis of results from different approaches shows the reliability of DTA. The paper concludes with recommendations for optimizing the use of DTA in real-time prediction of porosity and permeability. Furthermore, it shows DTA’s perspective and continuous applications in petrophysical evaluation, including essential parameters such as saturation, total organic carbon (TOC), mineral volumes, and brittleness. These valuable findings can be acquired from data during various phases of the drilling and completion process, enabling a comprehensive evaluation of subsurface characteristics. The present study employed the Domain Transfer Analysis for Optimal Prospectivity Evaluation. The successful application of DTA establishes a solid foundation for reservoir characterization and modeling research. The results show that the Domain Transfer Analysis provides better results than other statistical methods. This approach can be used globally to provide a strong correlation between real and predicted reservoir properties.