Continuous Rock Strength Measurements On Core And
Neural Network Modeling Result In Significant Improvements In Log-Based
Rock Strength Predictions Used To Optimize Completion Design and Improve
Prediction of Sanding Potential and Wellbore Stability
Roberto Suárez-Rivera, SPE, TerraTek, Gary Ostroff, BHP Billiton
Petroleum (Americas) Inc., KaiSoon Tan, BHP Billiton Petroleum (Americas)
Inc., Bill Begnaud, SPE, BHP Billiton Petroleum (Americas) Inc., Wesley
Martin, TerraTek and Tony Bermudez,TerraTek.
This paper was prepared
for presentation at the2003 SPE Annual Technical Conference and Exhibition
held in Denver, Colorado, USA, 5-8 October 2003.
Copyright 2003, Society of Petroleum Engineers Inc.
Abstract
The scaling-up of laboratory rock mechanical measurementsfrom sample-scale
to reservoir-scale is fundamental to evaluation of wellbore stability,
sanding potential, reservoir compaction or casing failure. Understanding
rock heterogeneity is fundamental for adequate scaling-up laboratory
measurements to core- and reservoir-scales and thus, to predictions
of mechanical failure. Historically, scaling-up from core scale to
reservoir scale has been dependent on calibration of log-based models
to a sparsely sampled data set of rock mechanical property measurements
made on core plugs. Such a sparsely sampled data set of core plug measurements
alone may inadequately characterize the range heterogeneities in the
reservoir, resulting in less than optimum log-based predictive models.
With the introduction of continuous, high resolution, rock strength
(UCS) measurements on core via scratch testing, an excellent calibration
reference for producing robust log-based predictions of rock strength
now exist.
In this study, high-resolution measurements of strength heterogeneity
were obtained as a function of core length and were correlated with fundamental
textural and compositional parameters from petrographic analysis. Using
adaptive learning neural networks, fundamental relationships between
log measurements and rock strength were obtained. This methodology was
adequate for characterizing the intrinsic rock heterogeneity at appropriate
scales for mechanical analysis of completion design and sanding (0.25
ft). The methodology is also potentially applicable to the scaling-up
of other fundamental mechanical properties such as in-situ strength,
compressibility and thick-walled cylinder strength.
Results show that intrinsic textural heterogeneity and strength heterogeneity
are strongly related in sedimentary rocks. Recognizing the importance
of rock heterogeneity and being able to scale-up this property to core
and reservoir scales via log measurements results in significant improvements
in the predictive capacity for sanding potential and wellbore stability.
For example, thin layers of considerably weaker-strength than the surrounding
rock, undetectable from conventional log-based rock strength predictions,
were detected and included in the mechanical model. In addition to possessing
high sanding potential, these weaker sections are also regions of fluid
loss during drilling. Results can be used for selection of competent
rock across the field (based on LWD measurements) for multilateral junction
placement, and for selection of optimum completion strategies.
|
