Скачать или смотреть 1040 Pena Repsol Reservoir Petrophysical Properties estimation from drill cuttings using analytic ML

2019 Force Symposium machine learning and subsurface data
Presentations https://www.npd.no/en/force/archive/h...


Reservoir Petrophysical Properties estimation from drill cuttings using
advanced data analytics.
Pena1, A.C., Santos2 C., Caja M.A. 2, Garcia2,L., Fernandez2, P., Perez2, J., Blazquez2, V.
(1) Universitat de Barcelona. (2) Repsol Technology Lab.
Conventional workflows to obtain petrophysical properties based on core/plug
analysis as a means to characterize reservoirs are usually time-consuming and
expensive, implying specific core sampling of the well. Contrarily, drill cuttings
are widely available, relatively cheap and with untapped potential. This study
proposes the estimation of petrophysical properties with the use of advanced
data analytics on multiphysics information from such lithological material. The
aim is to build models that allow the prediction of Routine Core Analysis's
(RCA) properties, based on more simple information, that can be quickly
obtained from different features extracted from the rock cuttings. For instance,
light photography images can be taken to cuttings, from which RG B properties
can be derived; from the radioactivity field, Gamma Ray (GR) measurements
can be obtained and from the X-ray Fluorescence (XRF) a compositional
analysis.
Porosity (the target for the regression model) was measured following RCA
techniques in different standard sandstone cores. Moreover, GR, XRF, and
RGB properties were obtained from the cuttings of those different standard
sandstones to create the feature matrix. Exploratory Data Analysis was carried
to evaluate the distributions and correlations within the data and also the
discriminative capacity of the properties have been analyzed to understand the
sensitivity of the properties to the different rock types. As a first approximation,
a linear model to predict porosity values was developed, achieving encouraging
results. Improvements of this model adding new features and the development
of new models for permeability and irreducible water saturation prediction will
be addressed in further studies. The results of the model will be presented.
The aforementioned multiphysics data (features), extracted from drill cuttings
and used to predict porosity values, can be obtained in short times, leading to
the possibility of acquiring and calculating those features automatically on the
well site (in situ). This proposed methodological approach could allow to obtain
relevant petrophysical information continuously during drilling.