The simulation and design of polymer flooding

• Main Author: AlSofi, Abdulkareem Mohamad
• Other Authors: Blunt, Martin ; LaForce, Tara
• Published: Imperial College London 2011
• Subjects: 622
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.537728

Most polymers used in Enhanced Oil Recovery exhibit shear-thinning behaviour. An in-house streamline simulator was modified and used to study the effects of both shear-thinning and shear-thickening on oil recovery. First, we describe how to simulate Newtonian and non-Newtonian polymer flooding. In contrast to current simulators, our methodology: (1) implements an iterative approach to solve the pressure field opposed to the common approach where this viscosity-pressure interdependence is ignored, (2) defines non-Newtonian viscosities to be cell-centred while current simulators use a face-approach and (3) uses a physically-based rheological model where non-Newtonian viscosities in two-phase flow are those exhibited in single-phase flow at the same pressure gradient not the same flow rate. To validate the simulator, we constructed one-dimensional analytical solutions for waterflooding with a non-Newtonian fluid. We then compared our results to those from commercial simulators illustrating the significance of current assumptions and their effects on the simulation results as well as the design of polymer flooding. The simulator was also used to investigate non-Newtonian effects on sweep and recovery. The results of this work prove the importance of taking polymers’ non-Newtonian behaviour into account for the successful design and evaluation of polymer flooding projects. Shear-thinning impairs sweep through exacerbated fingering and channelling. Increased channelling is due both to an overall reduction in viscosity as well as local viscosity variations, the latter factor being less significant. In addition, our results illustrate the potential of using shear-thickening agents for channelling reduction. The numerical simulation of such processes is also a challenge as the solutions should minimise numerical dispersion. Traditional numerical simulations of polymer flooding give excessive front smearing compared to pure waterfloods, requiring many thousands of gridblocks in one dimension to resolve the fronts adequately and rendering the predictions from three-dimensional simulations dubious at best. Investigating numerical dispersion in simulations of polymer flooding, we showed that these erroneous predictions occur because of the coupling of compositional dispersion with fractional flow. Small errors in composition alter the fractional flow, causing the development of incorrect wavespeeds. Rather than implementing a higher-order discretisation method, we propose a simple scheme based on segregated-flow within a gridblock. Compared to current mixing schemes, it differs in that segregation not only affects fluid properties but the transport, too. The scheme was shown to reestablish self-sharpness across the trailing shock. After validating the approach in one dimension, we performed multi-dimensional simulations demonstrating that traditional simulation methods can vastly overestimate recovery, potentially leading to poor injection design and management decisions. Finally, we illustrated the extendibility of this technique to low-salinity flooding as well as compositional simulations of miscible and near-miscible gas injection processes. At the end, since one of the main purposes of reservoir simulation is to optimally design the exploitation and production of petroleum resources, we investigated various aspects of the design of polymer flooding processes. First, we investigated the design of such processes in terms of finding the optimal solution and the characteristics of optimal strategies. The results suggest that polymer-flooding design – in terms of concentration, slug size and initiation – is more intuitive than expected previously. In terms of optimisation, polymer flooding is unimodal. In terms of optimal design, we found that: (1) it is always beneficial to start polymer flooding as soon as possible preferably before any waterflooding; (2) optimal slugs are very close to being continuous and (3) shear-thinning floods require higher polymer concentrations to compensate for losses in mobility control. Second, we quantified the impact of uncertainty on both the optimal design and profitability. The uncertainty results provide a quantitative ranking of the various factors affecting polymer flooding. This serves as a guide to associated data-acquisition efforts. Pre-polymer flooding initiation efforts can be focused on reducing uncertainties of high impact factors, thereby increasing the probability of success. Finally, we investigated upscaling effects. The main limitation to the use of upscaled models was found to be injectivity-related. For cases where polymer-flooding injectivity was not a factor, the results illustrate the potential utility of upscaled models for the preliminary design of polymer floods in terms of optimal polymer concentrations. This is despite the significant mismatch in polymer flooding predictions obtained with the different upscaled models.