Machine learning-driven classification of hydraulic flow units for enhanced reservoir characterization

• Published in: Physics of fluids (1994) Vol. 37; no. 3
• Main Authors: Hussain, Wakeel, Ali, Muhammad, Kasala, Erasto E, Ali, Sajid, AL-khulaidi, Ghamdan, Sadiq, Izhar, Nyakilla, Edwin E, Hussain, Saddam, Abelly, Elieneza Nicodemus
• Format: Journal Article
• Language: English
• Published: Melville American Institute of Physics 01.03.2025
• Subjects: Classification; Cluster analysis; Clustering; Fluid flow; Indicators; Machine learning; Particle swarm optimization; Reservoirs; Sandstone; Shales; Spatial distribution; Vector quantization
• ISSN: 1070-6631; 1089-7666
• DOI: 10.1063/5.0259689

This study focuses on the classification of Hydraulic Flow Units (HFUs) within the Lower Goru reservoir using a hybrid modeling approach for a more precise and data-driven reservoir characterization. The methodology begins with K-means clustering, which groups the reservoir into distinct HFUs based on reservoir properties. To enhance the accuracy of this classification, Particle Swarm Optimization (PSO) is employed to optimize the clustering process. The flow capacity and rock quality of each HFU are then assessed using two key indicators: the flow zone indicator (FZI) and the rock quality index (RQI). The results reveal four distinct HFUs: Clean Sandstone, Clayey Sandstone, Shaly Sandstone, and Shale. Among these, HFU 1 (Clean Sandstone) exhibits the highest FZI and RQI values, indicating excellent rock quality and flow capacity, while HFU 2 (Clayey Sandstone) demonstrates moderate FZI and RQI values, suggesting good reservoir potential. In contrast, HFUs 3 (Shaly Sandstone) and 4 (Shale) show progressively lower FZI and RQI values, reflecting poorer rock quality and reduced flow potential. This integrated approach significantly improves the precision of reservoir characterization by combining K-means clustering, PSO optimization, and petrophysical indicators such as FZI and RQI. The study's findings not only provide valuable understanding of reservoir dynamics and fluid flow potential but also enhance our comprehension of the spatial distribution and petrophysical properties of each HFU, offering a solid foundation for optimizing hydrocarbon recovery and enhancing reservoir management approaches.