Conventional rock classification in carbonate reservoirs typically requires considerable amount of core data, which usually may not be available at the depth resolution required for each target interval. In cases of tight carbonate rocks with extremely low porosity (less than 5% in average) and permeability (less than 0.1 md), a reliable rock classification is essential for well stimulation modeling. Such rock classification should take into account depth-by-depth petrophysical, compositional, and elastic properties of the formation. In this paper, we apply an integrated rock classification technique to enhance (a) well-log-based estimates of petrophysical, compositional, and elastic properties and (b) selection of appropriate candidate zones for acid fracturing treatment design in a tight carbonate reservoir in northern slope of Tazhong Uplift, Tarim Basin, China.
We first perform multi-mineral analysis and estimate volumetric concentrations of minerals, porosity, and fluid saturations. Since shear wave sonic logs are not available in most of the wells, we estimate elastic moduli using effective medium models including self-consistent approximation and differential effective medium theory. Corrections including the impact of fluids are developed using Biot-Gassmann fluid substitution. The inputs to the effective medium models include (a) the petrophysical and compositional properties obtained from well logs, (b) bulk and shear moduli for each mineral and fluid component, and (c) shape of rock inclusions (i.e., grains and pores). Core measurements are used for cross validating the well-log-based estimates of elastic moduli and petrophysical properties. Accordingly, we proposed a rock classification technique using unsupervised neural network that integrated depth-by-depth volumetric concentrations of minerals, porosity, and elastic moduli. Finally, we derived permeability models in each rock type and estimated the permeability in the target depth intervals. Variogram analysis on well-log-based estimates of permeability provides correlation lengths as inputs to acid fracturing treatment modeling.
We successfully applied the technique introduced here to a challenging tight gas interval of Tarim field in China. The estimated porosity and permeability were in good agreement with laboratory core measurements. The identified rock classes were verified by core samples and thin sections. We estimated elastic moduli with average relative errors of approximately 13% compare to the core measurements. The estimated elastic moduli were used as a key input for modeling of acid-fracturing treatments and improved stimulation success.
The rock classification technique introduced here provides important input parameters for well stimulation modeling, gives insight into evaluation of acid fracturing in tight carbonate reservoir, and helps with selection of best candidate zones for acid fracturing treatment design.