A Data Driven Workflow Enables Successful Fluid Characterization and Modeling of a Near-Critical Fluid System

Abstract

A fundamental step towards accurate reservoir characterization and hydrocarbon in-place estimation involves correct fluid type identification. Precise modeling is a challenge for gas condensates due to their complex near-critical behavior. Moreover, this becomes an uphill task when commingled production from a compartmentalized reservoir is going on. This work presents a data-driven workflow to solve the fluid type identification challenge, for a compartmentalized and layered field that was previously characterized as having shallow gas condensate and deeper oil layers. Consequently, a single, robust and coherent EOS model was obtained to successfully predict field-wide hydrocarbon properties. Fluid analysis of seven different separator samples of five formations from three wells were analyzed in this study. The development and implementation of the new workflow includes multiple stages. An extensive data analysis is carried out, which includes compositional comparative analysis of the samples, comparison between lab and producing gas-oil-ratios, and establishment of compositional grading utilizing key PVT parameters. Next, a decision tree methodology was utilized to prioritize samples based on lab adjustments along with saturation pressure and reservoir pressure comparison and apply the compositional gradient approach on the most representative dataset. Finally, a common EOS model for the field was developed, tuned, and validated through quality checks by a simulation model. Comparative compositional analysis helped in rectifying PVT lab's fluid classification for one of the samples based on C7+ composition. Evidence of compositional variation with depth is observed; an increase in C7+ percentage and reduction in API shows heavier components in the fluid column from top to bottom. The high-priority data points exhibited a gradient without contact (representative of near-critical fluids), in agreement with the comparative analysis. Furthermore, about 30% liquid dropout for a pressure change of 200-300 psi for the oil sample is also pointing towards a near-critical system. The compositional gradient modeling enabled a single coherent solution to be applied, which accurately predicts both oil and gas properties. Finally, the EOS model was tested and validated through a simulation model-based workflow whose results are consistent with the measured lab parameters. This final test was important, as there was always the alternative hypothesis: a gas-condensate underlain by a volatile oil with a distinct GOC. Some of the data points were consistent with this hypothesis, and it was not immediately obvious as to which type of fluid existed in the reservoir. Other evidence included unusual GOR versus time behavior, which was inconsistent with a sharp GOC. The resulting PVT model enabled the prediction of gas vs. condensate fluids field-wide for a highly complex fluid system. An attempt to unravel the enigma through integrated data analysis has helped in providing a new perspective of near-critical fluid system for field.