Prediction of thermal maturity index parameters in organic shales
plays a critical role in defining the hydrocarbon prospect and proper
economic evaluation of the field. Hydrocarbon potential in shales
is evaluated using the percentage of organic indices such as total
organic carbon (TOC), thermal maturity temperature, source potentials,
and hydrogen and oxygen indices. Direct measurement of these parameters
in the laboratory is the most accurate way to obtain a representative
value, but, at the same time, it is very expensive. In the absence
of such facilities, other approaches such as analytical solutions
and empirical correlations are used to estimate the organic indices
in shale. The objective of this study is to develop data-driven machine
learning-based models to predict continuous profiles of geochemical
logs of organic shale formation. The machine learning models are trained
using the petrophysical wireline logs as input and the corresponding
laboratory-measured core data as a target for Barnett shale formations.
More than 400 log data and the corresponding core data were collected
for this purpose. The petrophysical wireline logs are γ-ray,
bulk density, neutron porosity, sonic transient time, spontaneous potential, and shallow
resistivity logs. The corresponding core data includes the experimental
results from the Rock-Eval pyrolysis and Leco TOC measurements. A
backpropagation artificial neural network coupled with a particle
swarm optimization algorithm was used in this work. In addition to
the development of optimized PSO-ANN models, explicit empirical correlations
are also extracted from the fine-tuned weights and biases of the optimized
models. The proposed models work with a higher accuracy within the
range of the data set on which the models are trained. The proposed
models can give real-time quantification of the organic matter maturity
that can be linked with the real-time drilling operations and help
identify the hotspots of mature organic matter in the drilled section.