Ensemble Learning Strategies for Enhanced Porosity Prediction in Complex Carbonate Reservoirs Using Well Data

Accurate estimation of porosity in carbonate rocks is a fundamental requirement for effective hydrocarbon reservoir characterization, yet it remains a significant challenge due to the inherent complexity of carbonate formations. Conventional approaches to porosity estimation rely heavily on well data, a process that is both time-consuming and costly. This limitation underscores the potential of modern machine learning (ML) techniques to enhance the efficiency and precision of porosity prediction. In this study, the comparative performance of ensemble learning models and single-model ML algorithms for porosity estimation was investigated. Three advanced techniques were employed: Convolutional Neural Networks (CNNs), Fuzzy Logic (FL), and Support Vector Regression (SVR). The models were trained and tested using a dataset obtained from an Iranian oil field, which included well log measurements such as Bit Size (BS), Caliper (CAL), Gamma Ray (GR), Neutron Porosity (NPHI), Bulk Density (RHOB), and Bulk Density Correction (DRHO). Each algorithm’s performance was rigorously evaluated with particular emphasis on its ability to capture non-linear relationships and manage irregular or noisy data. The findings indicate that while individual ML models provide valuable contributions to porosity prediction, ensemble learning frameworks consistently deliver superior accuracy and reliability. The comparative analysis highlights the advantages of ensemble approaches, demonstrating their potential to integrate diverse learning algorithms into a unified model with enhanced predictive capabilities.

Keywords:

Carbonate Reservoirs , Porosity Prediction , Well Log Data , Ensemble Learning , Machine Learning Algorithms