PETROPHYSICAL ROCK TYPING BASED ON PORE GEOMETRY IMPROVES PERMEABILITY AND BOUND FLUID VOLUME ESTIMATION IN HETEROGENEOUS SANDSTONE FORMATIONS

摘要

In complex reservoirs, variations in pore geometrical attributes define distinct hydraulic units, which must be accounted for in a permeability prediction model. In this paper, we study an offshore siliciclastic brownfield where the reservoirs are highly heterogeneous with matrix permeability ranging from 0.1 mD to over 1 Darcy. Logging-While-Drilling (LWD) triple combo and NMR T2 relaxation data were acquired while drilling. Core samples were taken in the target sands and conventional core analysis was performed. We characterize the pore geometry variations by classifying core samples into a number of petrophysical rock types using a novel scheme.The novel classification is done by applying carefully designed cutoffs to the pore throat radii of the core samples and is propagated to the entire well with an artificial neural network. For each rock type, the parameters in the Timur-Coates permeability equation are calibrated with core measurements and a continuous permeability is computed using the calibrated parameters. The workflow consists of these detailed steps:1. Compute pore throat radius from core porosityand permeability using established equations.Classify the samples by defining a set of porethroat radius cutoffs based on statisticalanalysis, modified Lorenz plot and porosity vs.permeability (poroperm) crossplot.2. For each rock type, calibrate the T2 cutoff valueand the multiplier (the A value) in the Timur-Coates equation by minimizing the differencebetween measured and predictedpermeabilities.3.With supervised machine learning, learn rocktype classification from cored intervals andpropagate the classification to un-coredintervals using selected log curves.4. Compute a continuous Timur-Coatespermeability for the entire well using thecalibrated parameters pertinent to eachpetrophysical rock type.The proposed workflow significantly improves the match between core and predicted permeability, as demonstrated in two development wells. By comparison, a conventional permeability model is unable to capture the permeability variations seen in the core data. An additional deliverable is a calibrated T2 cutoff curve that varies with the rock types. The variable T2 cutoffs can be applied to an offset well containing the same rock types to improve the accuracy of bound fluid volume (BFV) estimation from the T2 distribution.