A Review of Data-Driven Machine Learning Applications in Reservoir Petrophysics

Reservoir petrophysical characterization stands as an essential initial step in petroleum production and gas storage operations. It involves the use of scientific and engineering tools to understand and explore the nature of the reservoir formation, its fluid content, and the most effective and efficient way of producing it. This involves determining the wetting behavior (wettability), the pore storage capacity (porosity), the quantity of individual fluids (saturation), and the ability of the reservoir to deliver its fluid to the wellbore (permeability). Conventional methods for determining these petrophysical properties such as the special core analysis laboratory (SCAL) and geophysical/petrophysical logs are being practiced. However, traditional SCAL, seismic, and logging methods are time-consuming and costly. Machine learning techniques are faster and help in analysis and better understanding of SCAL and logging methods, and it also provides reliable estimations of reservoir petrophysical properties. Therefore, this review provides a comprehensive overview of recent advancements in machine learning (ML) applied to reservoir petrophysics, covering applications in hydrocarbon exploration, enhanced recovery, and carbon dioxide (CO₂) and hydrogen (H₂) storage. Techniques for reservoir petrophysical characterization are explored, focusing on ML applications in rock typing, porosity/permeability estimation, fluid identification, and wettability assessment. Challenges and limitations associated with ML algorithms in petrophysical analyses are discussed, with insights into future research directions. The review encompasses a broad range of ML algorithms such as artificial neural networks, support vector machines, decision trees, and ensemble methods. Structured sections discuss ML-based petrophysical characterization, ML in CO₂/H₂ storage, integrated workflows combining ML with traditional methods, and challenges of ML applications in petrophysics. The review aims to illuminate the transformative impact of ML on reservoir petrophysics and its potential in CO₂ and H₂ storage, offering valuable insights for researchers and industry professionals. Promising results have been achieved with ML in predicting petrophysical properties, lithology classification, and fluid identification. Opportunities for further research and development in ML applications for reservoir petrophysics are identified, emphasizing the integration of ML with physics-informed models and conventional analysis methods. This review uniquely covers both laboratory and field data, making it a comprehensive resource for understanding ML techniques in reservoir petrophysics, spanning oil and gas reservoirs as well as CO₂ and H₂ subsurface storage operations.