Application of local linear neuro-fuzzy model in estimating reservoir water saturation from well logs

The objective of the present study is to show an application of local linear neuro-fuzzy (LLNF) model in estimating reservoir water saturation (S-w) from well logs in a carbonate reservoir. A locally linear model tree (LOLIMOT), as an incremental tree-based learning algorithm, is actively used for optimizing the rule premise parameters. To show the relative strength of the proposed method in estimating Sw, the results obtained from LLNF are compared with three conventional equations. Moreover, the results are also evaluated on the basis of two criteria including "determination coefficient (R-2)" and "root mean square error (RMSE)." The R-2 for the estimated Sw is 0.93 using the LLNF model, while it is approximated about 0.81 by the " Dual-water (Clavier et al. 1984) equation," as the best-fitted empirical equation. Also, the RMSE obtained is 0.023 by LLNF model, while it is 0.042 by Dual-water.