Correlation and complexity analysis of well logs via Lyapunov, Hurst, Lempel–Ziv and neural network algorithms

Abstract

Well logs produce a wealth of data that can be used to evaluate the production capacity of oil and gas fields. These data are usually concerned with depth series of petrophysical quantities such as the sonic transient time, gamma emission, deep induction resistivity, neutron porosity and bulk density. Here, we perform a correlation and complexity analysis of well log data from the Namorado’s school field using Lyapunov, Hurst, Lempel–Ziv and neural network algorithms. After identifying the most correlated and complex series, we demonstrate that well log data estimates can be confidently performed by neural network algorithms either to complete missing data or to infer complete well logs of a specific quantity.