Abstract

The accurate prediction of porosity and water saturation from well electro log data is crucial in the field of subsurface geology for petroleum exploration and development. Traditional methods for estimating porosity and water saturation involves manual interpretation and empirical correlations, which can be time-consuming and prone to human error. This study focuses on developing a machine learning model to predict porosity and water saturation using electro log curves. For supervised machine learning the dataset was divided into training, validation, and test sets to ensure unbiased model evaluation. Accordingly, 20 wells (ANKL-M1 to M20) were selected having wide lateral spread and covering S-1 to S-5 Pay sands within Hazad Member of Ankleshwar Formation, based on the availability of conventional/advanced logs, core data studies and production performance. Petrophysical processing of these logs has been done with optimized petrophysical parameters. The AIML (Artificial Intelligence and Machine Learning) model has been trained on 15 wells (ANKLM1 to M15) by supervised learning (Porosity and saturation) and validated by 5 (ANKL-M16-M20) blind wells. Cross-validation techniques, particularly k-fold cross-validation, were employed to optimize hyperparameters and prevent over-fitting.