Training Hybrid Neuro-Fuzzy System to Infer Permeability in Wells on Maracaibo Lake, Venezuela

TL;DR

This paper presents a hybrid Neuro-Fuzzy System to accurately infer permeability in wells using porosity and water saturation data, improving upon traditional empirical methods in the Maracaibo Lake region.

Contribution

The study introduces a hybrid Neuro-Fuzzy System (ANFIS) for permeability inference, tailored to well data from Maracaibo Lake, enhancing accuracy over existing empirical equations.

Findings

NFS provided high-accuracy permeability estimates.

The model was validated with data from neighboring wells.

ANFIS effectively integrated porosity and water saturation data.

Abstract

The high accuracy on inferrring of rocks properties, such as permeability ($k$), is a very useful study in the analysis of wells. This has led to development and use of empirical equations like Tixier, Timur, among others. In order to improve the inference of permeability we used a hybrid Neuro-Fuzzy System (NFS). The NFS allowed us to infer permeability of well, from data of porosity ($\phi$) and water saturation ($Sw$). The work was performed with data from wells VCL-1021 (P21) and VCL-950 (P50), Block III, Maracaibo Lake, Venezuela. We evaluated the NFS equations ($k_{P50,i}(\phi_i,Sw_i)$) with neighboring well data ($P21$), in order to verify the validity of the equations in the area. We have used ANFIS in MatLab.