Why is it so Difficult to Predict Polymer Injectivity in Chemical Oil Recovery Processes?

Polymer injection to improve and/or accelerate oil recovery is a widespread technique with numerous ongoing and successful projects. In recent years, many field cases have been reported with injected polymer viscosity ranging from 5 to 160cP, producing large incremental oil volumes, without major injectivity issues. These field results often contradict pessimistic predictions of injectivity from prior studies. Despite abundant publications on the subject, there is no standard explanation of the reasons for discrepancies between forecast and actual behavior, and many questions are not yet fully answered. Will it be possible to inject the polymer solution at target viscosity? How much to inject? How fast? Will high pressures lead to fracturing or polymer degradation? Should the polymer solution be pre-treated, pre-sheared? What should be done if planned injection rates are not achievable? Will injectivity decline over time? These questions are very topical when it comes to building a business case for EOR, using 3D reservoir simulation models for forecasting production and calculating the economics of the project.
In this paper, we present a critical review of selected field cases from the literature, analyzing reservoir characteristics and development history as well as properties of the injected solution. We discuss the mechanisms which can affect injectivity, including polymer solution rheology, near-well flow regimes, reservoir heterogeneity and geomechanical effects, and how these mechanisms can be represented in reservoir simulation models. Based on this investigation, we propose appropriate methodologies for dynamic modeling of polymer injection, considering the impact on predicted flow behavior of assumptions about polymer physics, selection of key parameters for sensitivity studies and the issues of upscaling from core experiments to the field. We suggest guidelines for using laboratory measurements and field observations, and for implementing forecasting workflows. Finally, we make recommendations on designing a practical field injection and monitoring program, to obtain data for calibrating models and improving future predictions.