Neuro-Fuzzy Modeling for the Prediction of Below-Bubble-Point Viscosity

Abstract

Accurate prediction of reservoir fluid is one of the important factors that needs to be determined due to it usefulness in fluid characterizations, material balance calculations, and general management of reserves. Below-bubble-point viscosity is one of the important variables that has been determined either experimentally or empirically.This work focuses on the use of neuro-fuzzy techniques to develop a belowbubble-point viscosity model using 1,693 data obtained from different oil fields in the Niger Delta, Nigeria. The data set was randomly divided into three parts with 56.3% used for training, 18.7% for validation, and 25% for testing. The accuracy of the developed model in this study was compared with some published correlations. The statistical analysis results show that the developed model outperformed existing published correlations.

Examination of Experimental Designs and Response Surface Methods for Uncertainty Analysis of Production Forecast: A Niger Delta Case Study

“…A more accurate response surface at affordable cost was generated when compared with the conventional techniques. Recent studies have shown that the problem of nonlinearity is addressed using ANN [14] and fuzzy inference system [15] that employs hybrid-learning rules for training process.…”