Lessons Learned From Energy Models: Iraq's South Rumaila Case Study

Summary Understanding reservoir behavior is the key to reservoir management. This study shows how energy modeling with rapid material-balance techniques, followed by numerical simulations with streamlines and finite-difference methods, aided understanding of reservoir-flow behavior. South Rumaila's long and elongated Zubair reservoir experiences uneven aquifer support from the western and eastern flanks. This uneven pressure support prompted injection in the weaker eastern flank to boost reservoir energy. We learned that aquifer influx provided nearly 95% of the reservoir's energy in its 50-year producing life, with water injection contributing less than 5% of the total energy supply. The west-to-east aquifer energy support is approximately 29:1, indicating the dominance of aquifer support in the west. Streamline simulations with a 663,000-cell model corroborated many of the findings learned during the material-balance phase of this study. Cursory adjustments to aquifer properties led to acceptable match with pulse-neutron capture or PNC-derived-time-lapse oil/water contact (OWC) surfaces. This global-matching approach speeded up the history-matching exercise in that performance of most wells was reproduced, without resorting to local adjustments of the cell properties. The history-matched model showed that the top layers contained the attic oil owing to lack of perforations. Lessons learned from this study include the idea that the material-balance work should precede any numerical flow-simulation study because it provides invaluable insights into reservoir-drive mechanisms and integrity of various input data, besides giving a rapid assessment of the reservoir's flow behavior. Credible material-balance work leaves very little room for adjustment of original hydrocarbons in place, which constitutes an excellent starting point for numerical models. Introduction Before the advent of widespread use of computers and numeric simulators, material-balance (MB) studies were the norm for reservoir management. In this context, Stewart et al. (1954), Irby et al. (1962), and McEwen (1962) presented useful studies. Most popular MB methods include those of Havlena and Odeh (1963), Campbell and Campbell (1978), and Tehrani (1985), among others. Pletcher (2002) provides a comprehensive review of the available MB techniques. In the modern era, classical MB studies seldom precede a full-field numeric modeling, presumably because MB is implicit in this approach. Nonetheless, we think valuable lessons can be learned from analytic MB studies at a fraction of time needed for detailed numeric modeling, preceded by geologic modeling. Of course, the value and amount of information derived from a multicell numeric model cannot be compared to a single-cell MB model. But, an analytic MB study can be an excellent precursor to any detailed 3D modeling effort. Although this point has been made by others (Dake 1994; Pletcher 2002), practice has, however, lagged conventional wisdom. In this paper, we attempt to show the value of a zero-dimensional MB study prior to doing detailed 3D numeric modeling, using both streamline and finite-difference methods. Streamline simulations speeded up the history-matching effort by a factor of three. However, we used the finite-difference approach in prediction runs for its greater flexibility in invoking various producing rules. Initially, the MB study provided key learnings about gross reservoir behavior very rapidly. In particular, energy contributions made by different drive mechanisms, such as uneven natural water influx and water injection, were of great interest for ongoing reservoir-management activities. Estimating in-place hydrocarbon volume and relative strength of the aquifer in the western and eastern flanks constituted key objectives of this study segment. Following the MB segment of the study, we pursued full-field match of historical data (pressure and OWC) with a streamline flow simulator to take advantage of rapid turnaround time. Thereafter, prediction runs were made with the finite-difference model to answer the ongoing water-injection question in the eastern flank of the reservoir. We learned that water injection should be turned off for improved sweep, leading to increased ultimate oil recovery. In addition, the numeric models identified the presence of remaining oil in the attic for future exploitation.