Lithofacies Analysis Using Unsupervised Self-Organizing Map in the Serpent Field, Offshore Nile Delta, Egypt

Paper ID : 1037-ISCH

Authors

Shaimaa ashraf ahmed eldabaa *

Abstract

Accurate facies analysis and thin gas channel delineation are critical for characterizing and managing reservoirs. Identifying facies and thin layers that are equal to the tunning thickness is challenging using conventional analysis and the most advanced techniques, such as seismic inversion techniques. By exploiting the non-linear relationships between the different seismic attributes, unsupervised self- organizing maps (SOMs) can identify layers thinner than the tuning thickness. Therefore, we introduce a new machine learning (ML) approach for reservoir characterization that is accurate and takes significantly less time than other advanced techniques. Additionally, the SOM approach is non- deterministic and does not demand data to fit physical models. Using the principal component analysis (PCA) dimensionality reduction technique, SOM clusters and maps multidimensional data onto lower- dimensional data. SOM, as a powerful pattern recognition approach, was able to identify all facies and gas channel spatial boundaries based on their geological characteristics without any prior knowledge of the rock type, lithology, porosity, fluid content, facies, and depositional environment. The proposed approach was validated by applying it in the serpent field, West offshore Nile Delta, Egypt. The SOM results were validated by well log analysis, and the facies classification proved better than the model- based inversion.

Keywords

Keywords Serpent Field, seismic inversion, machine learning (ML), self-organized map (SOM), facies analysis,

Status: Abstract Accepted (Poster Presentation)