Prediction and Geomechanical characterization of reservoir in the Jay oil-field of the Niger Delta Basin, Nigeria

The objective of this paper was to predict and geomechanically characterize the reservoirs in the JAY oil-fields (JAY-25, JAY-26, JAY-30, and JAY-33) in the Niger Delta Region of Nigeria using Probability Neural Network (PNN) and Multilayered Feed Forward Network (MLFN) machine learning algorithms which subsequently identified a reservoir of interest. By employing a combination of gamma ray, resistivity logs, volume of shale, porosity, and water saturation logs. Data obtained showed that the reservoir is porous with a high concentration of exploitable hydrocarbons. Well correlation analysis suggests good continuity of sand (reservoir) and shale (seal) packages in the field. Results from these two machine learning algorithms were compared to determine the more reliable technique to predict geomechanical properties for the field from seismic data. The feasibility study results for predicted Young's Modulus using Probabilistic Neural Network (PNN) outperformed Multilayer Feed forward Neural Network (MLFN), showing correlation coefficients of 0.99 and 0.98, respectively. Cross-validation correlations were 0.96 and 0.87 for PNN and MLFN, respectively. Similarly, PNN demonstrated higher performance in predicting Poisson ratio with correlation coefficients of 0.998 compared to MLFN's 0.996, and cross-validation correlations of 0.99 and 0.95 respectively. Therefore, the PNN technique is recommended for predicting both geomechanical properties from seismic in the field. These findings underscore the effectiveness of machine learning models in capturing geomechanical patterns, offering valuable insights for reservoir characterization and exploration in the Niger Delta