Incorporating Discrete Microfacies Sequences to Improve Permeability Estimation in Sandstone Reservoirs

Several cases have been conducted to address the permeability modeling and estimation, but all were not accurate because of the heteroscedasticity between data. Therefore, integrating the microfacies sequences into permeability modeling became a crucial to obtain accurate prediction and then improve the overall reservoir characterization. The discrete microfacies distribution leads to distinct regression lines given each microfacies type. Therefore, the Random Forest (RF) algorithm was considered in this paper for microfacies classification and Smooth Generalized Additive Modeling (sGAM) was considered for permeability modeling as a function of well logging data and the predicted discrete microfacies distribution. The well logging records that were incorporated into the microfacies classification and permeability modeling: SP, ILD and density porosity logs. These two approaches were adopted in a well in a sandstone reservoir, located in East Texas basin. The effectiveness of using RF and sGAM approaches was investigated by their performance to handle wide ranges of data given the five microfacies types. More specifically, the Random Forest Modeling was super accurate to predict the microfacies distribution at the missing intervals for the same well and other wells. Moreover, the sGAM resulted to obtain accurate modeling and prediction of permeability in high and low permeable intervals.