Incorporated artificial intelligence and digital imaging system for unconventional reservoirs characterization

The unconventional reservoirs (UCRs) play a key role in global oil and gas supply. However, their reservoir characterization is difficult because of complex pore structure and low permeability-viscosity ratio. Usually, traditional techniques hardly can be used for determination of pore structure and estimation of reservoir properties. In this case, digital rock analysis (DRA) shows the potential for capturing detailed pore structure information and simulating rock properties, such as porosity, permeability, electrical properties and elastic properties. Recently, artificial intelligence (AI) techniques have presented an ever-increasing trend in a wide variety of research and commercial fields. Many AI applications can free man from the labor of complicated works in some way. Machine leaning (ML), which is a subdivision of AI, has attracted researchers' attention and been widely used in geoscience and reservoir characterization, such as feature extracting, rock type prediction and reservoir property estimation. The incorporation of AI and DRA is becoming an inevitable development trend for future reservoir study. In this paper, firstly, DRA workflow for reservoir characterization is introduced; secondly, the commonly used ML algorithms in DRA study is reviewed; finally, a case study of characterization of a tight carbonate reservoir with ML algorithm and DRA is presented. The analysis shows that ML can be applied in any part of DRA progress such as image segmentation, feature detection, rock image classification, numerical simulation and result analysis. Compared with traditional DRA algorithm, ML-based methods can reduce manual operation that has greatly impact on the results. The combination of ML and DRA provides a new insight in UCRs characterization and outlook the future opportunities of AI to solve the oilfield problems.